Detection and quantification of circulating tumor cells in mouse models of human breast cancer using immunomagnetic enrichment and multiparameter flow cytometry

Authors

  • Alison L. Allan,

    1. London Regional Cancer Program, London, Ontario, Canada
    2. Department of Oncology, University of Western Ontario, London, Ontario, Canada
    3. Department of Pathology, University of Western Ontario, London, Ontario, Canada
    Search for more papers by this author
  • Sharon A. Vantyghem,

    1. London Regional Cancer Program, London, Ontario, Canada
    2. Department of Oncology, University of Western Ontario, London, Ontario, Canada
    3. Department of Pathology, University of Western Ontario, London, Ontario, Canada
    Search for more papers by this author
  • Alan B. Tuck,

    1. London Regional Cancer Program, London, Ontario, Canada
    2. Department of Oncology, University of Western Ontario, London, Ontario, Canada
    3. Department of Pathology, University of Western Ontario, London, Ontario, Canada
    4. London Health Sciences Centre, London, Ontario, Canada
    Search for more papers by this author
  • Ann F. Chambers,

    1. London Regional Cancer Program, London, Ontario, Canada
    2. Department of Oncology, University of Western Ontario, London, Ontario, Canada
    3. Department of Pathology, University of Western Ontario, London, Ontario, Canada
    Search for more papers by this author
  • Ian H. Chin-Yee,

    1. London Regional Cancer Program, London, Ontario, Canada
    2. London Health Sciences Centre, London, Ontario, Canada
    3. Lawson Health Research Institute, London, Ontario, Canada
    Search for more papers by this author
  • Michael Keeney

    Corresponding author
    1. London Health Sciences Centre, London, Ontario, Canada
    2. Lawson Health Research Institute, London, Ontario, Canada
    • Technical Specialist, Hematology/Coagulation, Associate Scientist, Lawson Health Research Institute, London Health Sciences Centre, 800 Commissioners Road East, London, ON N6A 4G5, Canada
    Search for more papers by this author

Abstract

Background

Circulating tumor cells (CTCs) in the peripheral blood of breast cancer patients may be an important indicator of metastatic disease and poor prognosis. However, the use of experimental models is required to fully elucidate the functional consequences of CTCs. The purpose of this study was to optimize the sensitivity of multiparameter flow cytometry for detection of human tumor cells in mouse models of breast cancer.

Methods

MDA-MB-468 human breast cancer cells were serially diluted in whole mouse blood. Samples were lysed and incubated with a fluorescein isothiocyanate–conjugated anti–human leukocytic antigen antibody and a phycoerythrin-conjugated anti-mouse pan-leukocyte CD45 antibody. Samples were then immunomagnetically depleted of CD45-positive leukocytes, fixed, permeabilized, and stained with propidium iodide before flow cytometric analysis.

Results

Human breast cancer cells could be differentiated from mouse leukocytes based on increased light scatter, cell surface marker expression, and aneuploid DNA content. The method was found to have a lower sensitivity limit of 10−5 and was effective for detecting human breast cancer cells in vivo in the circulation of experimental mice carrying primary human mammary tumors.

Conclusions

This technique has the potential to be a valuable and sensitive tool for investigating the biological relevance of CTCs in experimental mouse models of breast cancer. © 2005 Wiley-Liss, Inc.

Over the past several decades, significant research advances have been made in the realms of cancer prevention, detection, and management of primary tumors. Despite this, breast cancer remains among the leading causes of morbidity and mortality in women (1, 2), primarily due to the failure of effective clinical detection and management of metastatic disease in distant sites such as lymph nodes, lungs, liver, brain, and bone (3–5).

The metastatic process is comprised of a series of sequential steps, and cancer cells must successfully complete each step to give rise to a metastatic tumor. These steps include dissemination of cancer cells from the primary tumor into the bloodstream (intravasation), survival in the circulation, arrest and extravasation into the secondary site, and initiation and maintenance of growth to form clinically detectable micrometastases (4, 6–10). Breast cancer cells may also disseminate from the primary tumor through the lymphatic system, although the lack of direct flow from the lymphatic system to other organs means that tumor cells escaping by this route must still enter the arterial system to be distributed to distant organs (4, 10, 11).

Given the multistep nature of the metastatic cascade, there should be several opportunities for early identification and therapeutic targeting of metastatic cells before they become a clinical problem. In breast cancer patients with metastatic or apparently localized disease, circulating tumor cells (CTCs) have been identified in the peripheral bloodstream using cytometric techniques (12–14) or nucleic acid–based techniques (15, 16). There is growing evidence that the presence of CTCs in breast cancer patients may be an important indicator of metastatic disease and poor prognosis (14, 17–21), although the functional relevance of CTCs remains poorly understood. Successful and consistent enumeration and tracking of CTCs in breast cancer patients would have tremendous clinical utility in terms of identifying the potential for metastatic disease at very early stages, managing risk stratification in the adjuvant setting, monitoring response to treatment, monitoring disease recurrence, and the prospective development of targeted therapies based on molecular characterization of CTCs (21–23). Although a recent study by Cristofanilli et al. (14, 21) have demonstrated encouraging results involving quantitation of CTCs in the clinical setting, most techniques to detect and analyze single cells in the peripheral blood of breast cancer patients are imperfect, especially if such cells are present as rare events (22, 23). Similar challenges in detecting and quantifying rare metastatic cells in mouse models of human breast cancer has hindered the ability to quantify early steps in metastasis and to determine the timing and location of metastatic spread of cells.

Experimental investigation of tumor growth and metastasis of human cancer often involves xenotransplantation of human tumor tissue or tumor cells into immunocompromised animals such as athymic nude mice (nu/nu), SCID (severe combined immunodeficiency) mice (xid), or beige (bg) mice (24–28). After orthotopic injection into the mammary fat pad and subsequent growth of a primary breast tumor, human breast cancer cells may spontaneously metastasize in a way that is thought to simulate many aspects of the natural metastatic process, most commonly involving the lymph nodes and lungs (29–31). The standard endpoint of this “spontaneous” metastasis assay is the detection of metastases in sites distant from the primary breast tumor grossly at necropsy or by histopathologic analysis. However, longitudinal assessment of the occurrence and kinetics of the metastatic process is difficult in this model system, particularly at early stages when individual disseminated tumor cells in the peripheral bloodstream or distant organs may be present only in very small numbers and must be identified against a large background of normal host cells.

The goal of the present study was to develop and optimize a flow cytometric method to detect and quantify rare circulating human breast cancer cells in mouse models of metastatic breast cancer. The unique characteristics of a xenograft model allowed us to take advantage of several differences between mouse leukocytes and human breast cancer cells, including differences in cell size, expression of species-specific and/or cell type-specific proteins, and differences in DNA content. Human breast cancer cells could be identified against a background of mouse leukocytes based their larger size (increased light scatter), positive staining with a fluorescein isothiocyanate (FITC)-conjugated anti–human leukocytic antigen (HLA) antibody, negative staining with a phycoerythrin (PE)-conjugated anti-mouse pan-leukocyte CD45 antibody, and aneuploid DNA content based on propidium iodide (PI) staining. These differentiation criteria were found to be effective for detecting a range of different human breast cancer cell lines in whole mouse blood, including MDA-MB-468, MDA-MB-231, MDA-MB-435, MCF-7, and 21NT cells. The lower detection limit for sensitivity of the flow cytometric method was 0.01% or 10−4 (one human tumor cell per 10,000 mouse leukocytes), and the sensitivity of the technique could be increased to 10−5 by using an immunomagnetic method for depletion of the CD45-positive mouse leukocytes before flow cytometric analysis. The technique was validated by detection of circulating tumor cells in experimental blood samples taken from mice that had developed mammary tumors after mammary fat-pad injection with MDA-MB-468 human breast cancer cells. Our findings suggest that this novel application of immunomagnetic enrichment and multiparameter flow cytometry has the potential to be a valuable tool for studying CTCs in experimental mouse models of human metastatic breast cancer.

MATERIAL AND METHODS

Cell Culture

MDA-MB-468 and MDA-MB-435 human breast carcinoma cells (a gift from Dr. Janet Price, MD Anderson Cancer Center, Houston, TX, USA) (29) were maintained in Minimal Essential Medium (αMEM) medium supplemented with 10% fetal bovine serum (FBS). MDA-MB-468 cells previously transfected to stably express enhanced green fluorescent protein (EGFP) were maintained in αMEM medium supplemented with 10% FBS and 500 μg/ml G418 to maintain selective pressure on the EGFP-transfected cells. MDA-MB-231 human breast carcinoma cells (a gift from Dr. Danny Welch, University of Alabama at Birmingham, Birmingham, AL, USA) (32) were maintained in Dulbecco's Minimal Eagle's Medium and F12 medium supplemented with 10% FBS. MCF-7 human breast carcinoma cells (a gift from Dr. James Koropatnick, London Regional Cancer Program, London, ON, Canada) (33) were maintained in Dulbecco's Minimal Eagle's Medium supplemented with 10% FBS. 21NT human mammary carcinoma cells (a gift from Dr. Vimla Band, Dana Farber Cancer Institute, Boston, MA, USA) (34) were maintained in αHE medium (αMEM supplemented with 10% FBS, L-glutamine [2 mM], insulin [1 mg/ml], epidermal growth factor [EGF; 12.5 ng/ml], hydrocortisone [2.8 mM], HEPES [10 mM], sodium pyruvate [1 mM], non-essential amino acids [0.1 mM], and gentamycin [50 mg/ml]) as previously described (35). All media were obtained from Invitrogen Corporation (Carlsbad, CA, USA). FBS was obtained from Sigma Chemical Company (St. Louis, MO, USA). All plastic ware was obtained from Nalge Nunc International (Rochester, NY, USA).

Cells were grown in T75 filter-top flasks in a 37°C humidified incubator with 5% CO2. Adherent cells were harvested at 80% confluency by washing twice with cold Dulbecco's phosphate buffered saline (PBS; Invitrogen Corporation) and exposure to 0.25% trypsin (Invitrogen Corporation) for 5 min at room temperature. Cells were counted with a hemocytometer (six replicates), washed twice with cold flow buffer (PBS + 2% bovine serum albumin [Bioshop Canada Inc., Burlington, ON, Canada] + 2 mM ethylenediaminetetraacetic acid [Bioshop Canada Inc.]), and resuspended in cold flow buffer at a concentration of 1 × 106 cells/ml.

Blood Collection

Fresh whole blood was collected from female athymic NCr nude mice (nu/nu; ages 10 to 20 weeks; Harlan, Indianapolis, IN, USA) via terminal cardiac puncture of the right ventricle. Blood was collected using 22-gauge needles and 1-ml syringes precoated with 30 μl fresh heparin (10,000 IU/ml; Leo Pharma, Thornhill, ON, Canada) and anticoagulated in potassium ethylenediaminetetraacetic acid blood collection microtubes. Needles, syringes, and microtubes were obtained from Becton Dickinson & Company (Franklin Lakes, NJ, USA).

Mammary Fat-Pad Injections

MDA-MB-468 cells stably transfected to express EGFP were used for the in vivo experiments to allow for confirmation of the presence of tumor cells in the collected blood samples by fluorescence microscopy secondary to detection and enumeration by flow cytometry. MDA-MB-468-EGFP cells were grown in 150-mm tissue culture dishes to approximately 80% confluency, trypsinized, washed twice with sterile PBS, and resuspended in cold PBS at a concentration of 1 × 107 cells/ml. For each mouse (n = 6), 1 × 106 cells were injected into the second thoracic mammary fat pad of 9- to 10-week-old female nude mice as described elsewhere (29). Primary mammary tumors developed at the site of injection, and tumors were allowed to grow for 8 to 12 weeks. At the time of death, tumor size was measured and whole blood was collected by terminal cardiac puncture as described above. Animals were housed and cared for in accordance with the recommendations of the Canadian Council on Animal Care, under a protocol approved by the University of Western Ontario Council on Animal Care.

Antibodies and Reagents

Mouse anti-human HLA antibody (clone W6/32) conjugated to FITC was obtained from Sigma Chemical Company. Rat anti-mouse pan-leukocytic CD45 antibody (clone 30-F11) conjugated to PE was obtained from Caltag Laboratories (Burlingame, CA, USA). Red blood cells were lysed using a 1× ammonium chloride lysing solution (Beckman Coulter, Fullerton, CA, USA). The EasySep PE Selection Kit (StemCell Technologies, Vancouver, BC, Canada) was used for immunomagnetic depletion of CD45-PE–positive mouse leukocytes. Samples were fixed and permeabilized using the IntraPrep Fix/Perm Kit (Beckman Coulter). Propidium iodide (PI; DNA-Prep Stain) was obtained from Beckman Coulter.

Test Sample Preparation

The cell concentration of human breast cancer cell suspensions and the leukocyte concentration of whole mouse blood samples were determined with a Coulter LH 750 Analyzer (Beckman Coulter). Human tumor cell concentration was normalized to the leukocyte count, and serial dilutions (50% to 0.005%) of human tumor cells were prepared by using whole mouse blood as the diluent. Because the focus of the study was to optimize detection of rare circulating human tumor cells in mouse blood by using flow cytometry, a second set of experiments was performed to test whether magnetic depletion of CD45-PE–positive cells before flow cytometric analysis could increase the detection sensitivity. In this set of experiments, tumor cells were spiked into mouse blood at levels of 10%, 1%, 0.1%, 0.01%, and 0% before the immunomagnetic enrichment procedure (described below).

To confirm the utility of the method in vivo, experimental blood samples from mice that had developed primary mammary tumors after mammary fat-pad injection with MDA-MB-468-EGFP breast cancer cells (tumor size 1–1.5 cm3, n = 6) were labeled and magnetically depleted of CD45-PE cells as described below to concentrate any circulating tumor cells. As a control, 100 μl of blood from each mouse was also subjected to gentle red blood cell lysis with 2 ml of 1× NH4Cl for 10 min, washed, centrifuged, and cultured on T25 tissue culture flasks with G418 selective medium. Forty-eight hours later, flasks were examined under fluorescence microscopy to confirm the presence or absence of tumor cells.

Labeling Procedure

Aliquots (100 μl) of whole mouse blood, breast cancer cells in suspension, or serial diluted samples of mixed blood and tumor cells were labeled with anti-human HLA-FITC (5 μl) and anti-mouse CD45-PE (3 μl) for 15 min at room temperature in the dark. Samples were subjected to red blood cell lysis with 2 ml of 1× NH4Cl for 10 min at room temperature, washed with an equal volume of PBS, centrifuged, and resuspended in 100 μl of flow buffer. Samples were fixed and permeabilized with the IntraPrep Fix/Perm Kit (Beckman Coulter) according to the manufacturer's instructions. Samples were washed with an excess volume of PBS, centrifuged, resuspended in 500 μl of PI (50 μg/ml) containing 4 KU/ml RNAse (DNA-Prep, Beckman Coulter), and incubated for 15 min at room temperature followed by 45 min at 4°C before flow cytometric analysis.

Immunomagnetic Separation

For immunomagnetic separation studies, the starting sample volume was 500 μl. Samples were subjected to red blood cell lysis in 10 ml of 1× NH4Cl for 10 min at room temperature, washed with an equal volume of PBS, centrifuged, and resuspended in 100 μl of flow buffer containing murine FcR blocker (StemCell Technologies) before antibody labeling with anti-human HLA-FITC (10 μl) and anti-mouse CD45-PE (15 μl). Samples were washed in an excess volume of PBS, resuspended in 100 μl of flow buffer, and immunomagnetically separated using the EasySep PE Selection Kit according to the manufacturer's instructions. Briefly, samples were labeled with 10 μl of the EasySep PE Selection Cocktail for 15 min followed by addition of 5 μl of EasySep magnetic nanoparticles and incubation for another 10 min. The total sample volume was brought up to 2.5 ml with flow buffer, and samples were incubated for 2 × 5 min in the EasySep magnet. After each incubation, the fraction containing the tumor cells (supernatant) was collected by pouring off, centrifuged, fixed, permeabilized, and stained with PI as described above.

Flow Cytometry

Prepared samples were kept on ice in the dark until analysis by flow cytometry. A four-color XL-MCL (Beckman Coulter) was configured to detect the HLA-FITC signal in FL1 (525-nm bandpass filter), CD45-PE in FL2 (575-nm bandpass filter), and PI in FL3 (625-nm bandpass filter). Compensation was adjusted on a 50:50 mixture of mouse leukocytes and MDA-MB-468 human breast cancer cells. Figure 1 shows the standard acquisition setup. A threshold (region A) was set on DNA content based on PI fluorescence equivalent to a diploid mouse leukocyte (cells with lowest DNA content). By plotting integral PI signal against the ratio of the PI peak/PI integral signal, it is possible to identify debris, cell doublets, and clumps that can be excluded from further analysis (Fig. 1A). Figure 1B shows HLA-FITC versus CD45-PE gated on region A in histogram 1A. Events that fell within region B (blue) were counted as meeting the criteria for mouse leukocytes, and events that fell within region C (red) were counted as meeting the criteria for MDA-MB-468 human tumor cells. Figure 2 shows light scatter (Fig. 2A), HLA-FITC (Fig. 2B), CD45-PE (Fig. 2C), and DNA content (Fig. 2D) gated on mouse leukocytes. Figure 2E–H shows the same parameters gated on MDA-MB-468 human breast cancer cells. Figure 2I–L shows expression based on the 50:50 mix of mouse leukocytes and human breast cancer cells from Figure 1. A minimum of 25,000 PI-positive events gated in region A was collected per sample. For samples with low numbers of tumor cells, up to 200,000 total events were collected.

Figure 1.

Standard acquisition setup. Samples were lysed, incubated with anti-human HLA-FITC (5 μl) and anti-mouse CD45-PE (3 μl), fixed, permeabilized, and stained with PI (50 μg/ml) before flow cytometric analysis. Compensation was adjusted on a 50:50 mixture of mouse leukocytes and MDA-MB-468 human breast cancer cells. A: Representative histogram shows a threshold (region A) set on DNA content based on integral PI signal versus the ratio of the PI peak/PI integral equivalent to a diploid mouse leukocyte (lowest DNA content). B: Paired representative histogram shows HLA-FITC versus CD45-PE fluorescence gated on region A in histogram A. A minimum of 25,000 PI-positive events gated in region A were collected per sample. Events that fell within region B (blue) were counted as meeting the criteria for mouse leukocytes, and events that fell within region C (red) were counted as meeting the criteria for MDA-MB-468 human tumor cells.

Figure 2.

Several parameters can be applied to differentiate mouse leukocytes from human MDA-MB-468 tumor cells in mixed samples. Samples were lysed, incubated with anti-human HLA-FITC (5 μl) and anti-mouse CD45-PE (3 μl), fixed, permeabilized, and stained with PI (50 μg/ml) before flow cytometric analysis. A minimum of 25,000 PI-positive events gated in region A (from Fig. 1) were collected per sample. Representative histograms are shown of individual criteria gated on mouse leukocytes (AD), MDA-MB-468 human tumor cells (EH), or the 50:50 mix of mouse leukocytes and human breast cancer cells from Figure 1 (IL). Criteria included differential size based on light scatter (A, E, I), differential staining with HLA-FITC in FL1 (B, F, J), differential staining with CD45-PE in FL2 (C, G, K), and differential DNA content based on nuclear staining with PI in FL3 (D, H, L).

Data Analysis

Titration experiments were performed on three separate occasions. In some situations the 50:50 mix of mouse leukocytes and human tumor cells did not produce the expected recovery of tumor cells, probably due to excessive clumping in the concentrated tumor sample. In this case, the actual percentage recovered in the 50% sample was taken as the starting point to assess recovery and linearity in the subsequent dilutions. Correlation, regression analysis, and a Mann-Whitney rank sum test were performed on each dataset. The lowest level of detection was determined according to two criteria: the value obtained had to correspond to the expected value, and HLA-FITC events in the test sample had to be higher than events in the negative control (mouse blood without tumor).

RESULTS

Specificity

The HLA-FITC antibody (clone W6/32; Sigma Chemical Co.) was found to be very specific for the human cells tested, with little or no nonspecific binding to mouse leukocytes (Fig. 2B,F,J). Similarly, the mouse CD45-PE antibody (clone 30-F11; Caltag) was found to be highly specific for mouse leukocytes and did not stain the human tumor cells (Fig. 2C,G,K). Light scatter properties (Fig. 2A,E,I) and PI fluorescence (Fig. 2D,H,L) provided little overlap between the smaller mouse cells and the larger human tumor cells. The addition of PI took advantage of the fact that mouse leukocytes have a diploid chromosome number of 40, whereas the normal diploid chromosome number for humans is 46. Similar to many human tumors and tumor cell lines (36), the MDA-MB-468 breast cancer cells used in this study also had hyperdiploid aneuploid DNA content, providing further separation from the mouse leukocytes.

To determine the applicability of the method to different breast cancer cell lines, the ability to detect MDA-MB-435, MDA-MB-231, 21NT, and MCF-7 breast tumor cells in mouse blood (∼30%) was also tested. In all cases the tumor cells exhibited similar staining to the MDA-MB-468 cells and could be clearly resolved from mouse leukocytes (Fig. 3). Some variability in HLA-FITC expression was observed, with the MCF-7 cells expressing the dimmest staining (Fig. 3E). Because all human breast cancer cell lines tested were found to be distinguishable from mouse leukocytes by using the same criteria, the remainder of the experiments were carried out with MDA-MB-468 cells only.

Figure 3.

The differentiation criteria are effective for detecting a range of different breast human cancer cell lines in whole mouse blood. Samples consisting of approximately 30% human tumor cells in mouse blood were lysed, incubated with anti-human HLA-FITC (5 μl) and anti-mouse CD45-PE antibody (3 μl), fixed, permeabilized, and stained with PI (50 μg/ml) before flow cytometric analysis. A minimum of 25,000 PI-positive events gated in region A (from Fig. 1) were collected per sample. Representative HLA-FITC versus CD45-PE histograms of (A) MDA-MB-468 cells, (B) MDA-MB-435 cells, (C) MDA-MB-231 cells, (D) 21NT cells, and (E) MCF-7 cells. Events that fell within region B (blue) were counted as meeting the criteria for mouse leukocytes, and events that fell within region C (red) were counted as meeting the criteria for human tumor cells.

Sensitivity

Analysis of serial dilutions (50% to 0.005%) of human MDA-MB-468 tumor cells in whole mouse blood demonstrated that the lower detection limit for sensitivity of the method was 0.01%, or 10−4, equivalent to one human tumor cell per 10,000 mouse leukocytes (Fig. 4A–F). Below this level, background events overlapped with tumor events in an unpredictable fashion. Recovery and linearity were highly reproducible across three separate experiments (Fig. 4G), and the number of tumor events recovered could be positively correlated with the number of tumor events expected based on the serial dilutions (R2 = 0.96). The percentage of tumor cells recovered was not found to be significantly different from the percentage of tumor cells expected (P > 0.70, Mann-Whitney rank sum test).

Figure 4.

The ability to detect human tumor cells in mouse blood by flow cytometry is titratable down to a sensitivity of 0.01%. Human tumor cell concentration was normalized to the leukocyte count, and serial dilutions (50% to 0.005%) of human tumor cells were prepared using whole mouse blood as the diluent. Samples were lysed, incubated with anti-human HLA-FITC (5 μl) and anti-mouse CD45-PE (3 μl), fixed, permeabilized, and stained with PI (50 μg/ml) before flow cytometric analysis. A minimum of 25,000 PI-positive events gated in region A (from Fig. 1) were collected per sample. For samples with small numbers of tumor cells, up to 200,000 total events were collected. Representative HLA-FITC versus CD45-PE histograms are shown for (A) 50% tumor cells, (B) 10% tumor cells, (C) 1% tumor cells, (D) 0.1% tumor cells, and (E) 0.01% tumor cells. F: Mouse blood alone was used as a negative control. Events that fell within region B (blue) were counted as meeting the criteria for mouse leukocytes, and events that fell within region C (red) were counted as meeting the criteria for MDA-MB-468 human tumor cells. G: Correlation and regression analysis of recovered versus expected number of positive tumor events at dilutions of lower than 2% (R2 = 0.96, three separate experiments). The percentage of tumor cells recovered was not significantly different from the percentage of tumor cells expected based on the serial dilutions (P > 0.70, Mann-Whitney rank sum test).

Immunomagnetic Enrichment

To increase the sensitivity of the flow cytometric method, human MDA-MB-468 tumor cells diluted in whole mouse blood (10%, 1%, 0.1%, 0.01%, and 0%) were magnetically depleted of CD45-PE events before flow cytometric analysis using the EasySep PE Selection Kit (StemCell Technologies). Analysis of the enriched tumor cell fractions (Fig. 5) demonstrated that it was possible to increase the sensitivity by approximately 1 log from the unseparated method (mean fold increase across all samples = 11.8 ± 2.6), allowing for a lower detection limit of 10−5, or one tumor cell in 100,000 mouse leukocytes (n = 3 separate experiments). The immunomagnetic enrichment seemed to be most efficient at very low tumor cell concentrations because the highest mean fold increase (22.7 ± 6.9) was observed in samples with the lowest starting concentration of tumor cells (0.01%).

Figure 5.

Detection sensitivity can be increased by immunomagnetic enrichment before flow cytometric analysis. Human breast cancer cell suspensions and whole mouse blood were analyzed with a Coulter LH 750 Analyzer. Human tumor cell concentration was normalized to the leukocyte count, and serial dilutions (10%, 1%, 0.1%, and 0.01%) of human tumor cells were prepared with whole mouse blood as the diluent. Samples were lysed, incubated with anti-human HLA-FITC (10 μl) and anti-mouse CD45-PE (15 μl), immunomagnetically depleted of CD45-positive cells using the EasySep PE Selection Kit, fixed, permeabilized, and stained with PI (50 μg/ml) before flow cytometric analysis. A minimum of 100,000 PI-positive events gated in region A (from Fig. 1) were collected per sample. Data are presented as the mean ± standard error of the mean for three separate experiments.

In Vivo Studies

To confirm the utility of the method in vivo, experimental blood samples from mice (n = 6) that had developed primary mammary tumors after mammary fat-pad injection with MDA-MB-468-EGFP tumor cells were labeled, magnetically depleted of CD45-PE cells, and analyzed (Fig. 6). In mice that had primary tumors of 1 cm3 or smaller (Fig. 6A), very few human tumor cells were detected in blood samples. However, in mice that had primary tumors of 1.5 cm3 or larger (Fig. 6B), circulating human tumor cells could be detected in the range of 0.2% to 2% in immunomagnetically enriched samples. For each mouse, an aliquot of the blood sample was also lysed, washed, centrifuged, and cultured for 48 h in selective medium before examination under fluorescence microscopy. In all cases, the presence or absence of tumor cells was consistent with the flow cytometric data for each corresponding sample (data not shown).

Figure 6.

Utility of immunomagnetic enrichment and multiparameter flow cytometry for detecting CTCs in vivo. MDA-MB-468 cells stably transfected to express EGFP were injected into the second thoracic mammary fat pad of 9- to 10-week-old female nude mice (1 × 106 cells/mouse, n = 6). Primary mammary tumors developed at the site of injection, and tumors were allowed to grow for 8 to 12 weeks. Whole blood was collected via terminal cardiac puncture. Samples were lysed, incubated with anti-human HLA-FITC (10 μl) and anti-mouse CD45-PE (15 μl), immunomagnetically depleted of CD45-positive cells using the EasySep PE Selection Kit, fixed, permeabilized, and stained with PI (50 μg/ml) before flow cytometric analysis. A minimum of 100,000 PI-positive events gated in region A (from Fig. 1) were collected per sample. Representative HLA-FITC versus CD45-PE histograms are shown from (A) mice bearing tumors of 1 cm3 or smaller at the time of blood collection and (B) mice bearing tumors of 1.5 cm3 or larger at the time of blood collection. Events that fell within region B (blue) were counted as meeting the criteria for mouse leukocytes, and events that fell within region C (red) were counted as meeting the criteria for MDA-MB-468 human tumor cells.

DISCUSSION

The study of rare metastatic tumor cells in mouse models of human breast cancer has been complicated by an inability to detect and resolve individual tumor cells against a background of large numbers of host cells. Several methods have been proposed to detect disseminated breast cancer cells in human blood or bone marrow using immunocytochemistry, polymerase chain reaction analysis, or flow cytometric approaches (12–16, 37–39). Most of these rely on the expression of epithelial-specific markers such as cytokeratins, which are expressed on epithelial cells but not on leukocytes (10, 22, 40). However, several reports have indicated not all cytokeratin-positive cells are tumor derived, and not all epithelial tumor cells are cytokeratin positive (41–47). In particular, cells of higher metastatic potential may lose expression of epithelial-specific markers during the course of metastatic progression (46–49). It has also been observed that many immortalized human breast cancer cell lines that are commonly used to study breast cancer metastasis demonstrate low or absent expression of epithelial-specific markers, particularly highly metastatic cell lines such as MDA-MB-231 and MDA-MB-435 (46, 47, 50, 51 and our unpublished observations).

To avoid possible false-positive and/or false-negative results associated with the use of epithelial-specific markers in our experimental system, we chose instead to take advantage of the unique characteristics of a xenograft model and use markers that were highly specific for human cells (HLA) or mouse leukocytes (CD45). In the present study, we demonstrated that HLA and CD45 are highly expressed and highly specific markers for identifying human breast cancer cells and mouse leukocytes, respectively, by multiparameter flow cytometry in individual or mixed samples of tumor cells and whole blood. Further, differences in cell size and DNA content between human breast cancer cells and mouse leukocytes provide additional parameters with which to distinguish between the two cell populations.

The differentiation criteria were found to be effective for detecting a range of different human breast cancer cell lines in whole mouse blood, including MDA-MB-468, MDA-MB-231, MDA-MB-435, MCF-7, and 21NT cells. Some variability in HLA expression was observed across the different cell lines, and these differences could be due to a number of factors. For example, because HLA class I antigens are believed to be involved in immunosurveillance against tumors and have been shown in some cases to have variable expression levels in primary breast tumors and disseminated tumor cells (52–56), the HLA expression status of the cells at the time of their initial isolation from breast cancer patients would most likely influence the HLA expression levels in the resulting established cell lines. It is also possible that differences in HLA expression could be due to specific genotypic/phenotypic alterations that can occur as a result of the inherent genetic instability of tumor cells and the long-term maintenance of immortalized cell lines in simplified culture conditions (51, 57). Similar to the discussed limitations in using epithelial markers, it has been suggested that variable expression of HLA class I antigens could be a limitation in using HLA as a tumor-specific marker for detecting disseminated tumor cells in patient samples (56). However, despite the observed variability in HLA expression across the different cell lines tested in this study, in all cases the human tumor cells could be clearly resolved from mouse leukocytes, indicating that HLA is a suitable marker for identifying human breast cancer cells in xenograft mouse models of breast cancer.

Rare-event detection is affected by many parameters, including quality of the starting sample, specificity and expression level of the chosen markers, sample preparation, and analysis. The largest potential source of error is typically the identification of true-negative events as false-positive ones (58). Several steps were taken in the present study to minimize the effect of this issue. To prevent carryover from a previous sample, two wash cycles with distilled water were processed through the flow cytometer between each sample. Flow rates were kept below 350 events/s to minimize the effect of coincidence events, although these events were visible at some of the higher concentrations (5% to 50%) of tumor cells in blood. Coincidence is caused by two cells passing in front of the laser intercept point at the same time, resulting in the appearance of dual-positive cells. Alternatively, the dual-positive cells could represent a mouse and tumor doublet pair expressing CD45-PE and HLA-FITC. Because the target population in prospective test samples is anticipated to be in the range of less than 0.1%, coincidence is unlikely to play a major role in detecting tumor events by this method. By using all the information generated by the three fluorescence parameters and light scattering, it was possible to decrease the likelihood of a nonspecific event fulfilling the criteria of a target event to very low levels.

Immunomagnetic methods for enrichment of tumor cells in blood samples can use two basic approaches: positive selection using anti-tumor antibodies conjugated to magnetic beads or negative selection using anti-leukocyte antibodies conjugated to magnetic beads (13, 59, 60). In this study, we chose to negatively select the tumor cells by specifically removing the CD45-positive mouse leukocytes. Although negative selection versus positive selection can compromise the purity of the target population (i.e., mouse leukocytes were still observed in the tumor cell fraction), it does ensure high recovery of the tumor cells. Another advantage of the negative-selection approach is that the target population of enriched tumor cells is subjected to less manipulation and is not labeled with magnetic beads or purifying antibodies. This is beneficial because it will allow a pure population to be used for future studies involving molecular and morphologic characterizations of these CTCs. In the present study, samples that had been magnetically depleted of CD45-positive leukocytes showed decreased differential light scatter between mouse and tumor compared with unseparated samples and exhibited higher background staining (data not shown), possibly due to the presence of bound and/or unbound magnetic nanoparticles. However, it was still possible to increase the flow cytometric detection sensitivity by approximately 1 log from the unseparated method, and the immunomagnetic enrichment seemed to be most efficient at very low tumor cell concentrations. This will be advantageous for our future experimental studies because we expect that in most cases our target population will be rare (<0.1%). Continuing studies will test whether the addition of a Ficoll gradient spin of the initial sample to remove granulocytes will help to decrease background events and increase sensitivity even further.

The applicability of the method for detecting rare circulating tumor cells in our experimental mouse models in vivo was validated by analyzing blood from mice that had developed primary mammary tumors after mammary fat-pad injection with MDA-MB-468-EGFP tumor cells. CTCs were only detected above background levels in mice bearing tumors of 1.5 cm3 or larger. It is possible that tumor cells are present in the circulation at levels lower than 10−5 and are thus undetectable by our method. However, because selective culturing of the blood for 48 h confirmed the flow cytometric analysis with regard to the presence or absence of tumor cells, it is likely that factors other than detection sensitivity may be contributing to the observed levels of CTCs in our experimental models. In addition to tumor size, other influencing factors might include time between tumor cell implantation and CTC analysis, viability of disseminated tumor cells in the bloodstream, and the metastatic chronology and/or metastatic potential of particular breast cancer cell lines. This last factor might be of particular importance in determining the rate and occurrence of tumor cell shedding from the primary tumor and is supported by the findings of Schmidt et al. (61) who found that mice injected with the highly aggressive 435/HAL breast cancer cell line (derived from MDA-MB-435 cells) had detectable CTCs as soon as 2 weeks postinjection, although these CTCs did not appear to be quantitatively correlated with increasing tumor size. The MDA-MB-468 cell line used in the present study is much less aggressive than MDA-MB-435 in terms of occurrence of distant metastasis and time to metastasis development (29, 30). Interestingly, the mice in our study that did have detectable levels of CTCs also had gross evidence of regional and distant metastases, whereas mice with undetectable CTCs had no observable metastases (data not shown). Although the animal experiments presented in this report are very preliminary and are intended as proof-of-principle studies only, our findings suggest that there may be a reliable relation between circulating tumor cells and metastatic progression in this spontaneous breast cancer metastasis model. Further studies are clearly needed to delineate the importance and functional relevance of CTCs in mouse models of human breast cancer, including examination of a greater number of animals, examination of other tumor injection routes (i.e., intravenous, intracardiac), comparison of different cell lines, comparison of different blood sampling sites, investigation of the appearance and kinetics of CTCs in relation to metastatic progression, and quantitative correlation with metastatic burden.

In summary, the combined use of immunomagnetic enrichment and multiparameter flow cytometry is a useful and sensitive approach for detecting and quantifying human breast cancer cells in mouse models of metastatic breast cancer. This technique has the potential to be a valuable tool for monitoring the kinetics of metastatic progression in vivo and for investigating the biological relevance of CTCs in experimental mouse models of metastatic breast cancer. Although previous studies have demonstrated that cytometric methods can be used to detect rare breast tumor events in human blood at frequencies approaching 10−7 (62–64), these methods involved the use of a modified multilaser cell sorter and an automated “genetic” algorithm to eliminate false-positive events (62), the use of a laser scanning cytometer (64), and/or the use of epithelial-specific markers unsuitable for our model system (62–64). The smaller blood volume of a mouse combined with our preliminary animal studies indicate that the observed lower detection limit of 10−5 of our technique should be more than sufficient for our studies in mouse models and should be applicable for use by any group with access to a standard flow cytometer. In addition, the use of complementary methods such as laser scanning cytometry might allow us to not only quantitate rare metastatic events in our mouse models but to also investigate the cytomorphology and molecular characteristics of CTCs. Such analysis of CTCs in experimental models will be highly complementary to recent encouraging findings regarding the value of CTC analysis in the clinical setting (14, 21) and will be important to fully elucidate the functional consequences of CTCs during breast cancer progression.

Acknowledgements

We thank Carl Postenka for expertise and assistance with the cardiac punctures, and Wendy Brown, Leslie Gray-Statchuk, Karin Weir, Jan Popma, and Sylvia Bamford for technical advice and helpful discussions. A.L.A. is supported by the H. L. Holmes Postdoctoral Award from the National Research Council of Canada. S.A.V. is the recipient of a Rethink Breast Cancer Career Development Award. This work was supported in part by grant 42511 to A.F.C. from the Canadian Institutes of Health Research.

Ancillary