Division of Drug Discovery and Development, Shizuoka Cancer Center Research Institute, Shizuoka, Japan
Correspondence to: Yasuhiro Koh, Drug Discovery and Development Division, Shizuoka Cancer Center Research Institute, 1007 Shimonagakubo, Nagaizumi-cho, Suntou-gun, Shizuoka 411-8777, Japan. E-mail: email@example.com
Circulating tumor cells (CTCs) are cells that have detached from the primary or metastatic tumor and entered the peripheral circulation . CTCs have been of interest to the medical and research communities for over a century . Recent advances in technology have enabled reproducible CTC detection and enumeration, and these techniques have been revealed as minimally invasive, real-time, and blood-based “liquid biopsy” [3, 4].
The CellSearch platform (Veridex, LLC, Raritan, NJ), the only US Food and Drug Administration-approved CTC enumeration system, has been used to demonstrate the prognostic significance of CTC numbers in patients with metastatic breast, prostate, or colorectal cancer [5-13]. The clinical significance of CTCs was also evaluated in patients with small-cell lung and gastric cancers [14-16]. One of the major limitations of the CellSearch system is its dependence on anti-epithelial cellular adhesion molecule (EpCAM) antibody-based enrichment. Recent studies demonstrated that tumor cells undergo the epithelial–mesenchymal transition (EMT) to gain migratory capacity . In this process, mesenchymal cells lose their epithelial characteristics and gain mesenchymal properties . CTC populations that have undergone EMT have been reported, and these CTCs lack or express low levels of epithelial markers such as EpCAM and cytokeratin (CK) [18-20]. Therefore, CTC enumeration using the EpCAM-based system may produce misleading findings due to the lower expression of this antigen. In fact, limited numbers of CTCs were observed using the CellSearch platform in normal-like breast tumor cells  as well as in patients with metastatic non-small cell lung cancer (NSCLC) [22, 23]. Thus, there exists an urgent need to address sensitivity issues related to EpCAM-based enrichment for the clinical use of CTCs in patients with tumors expressing low levels of EpCAM. Moreover, the ability to capture tumor cells with mesenchymal phenotypes in peripheral blood may be useful to improve the understanding of the clinical significance of CTCs in the metastatic process.
In a recent study, we developed a flow cytometry protocol for detecting CTCs using a novel flow cytometry system, FISHMAN-R . The FISHMAN-R system, which uses a microfluidic chip, features cross-contamination-free measurement and quantification of the entire sample and permits the collection of the measured sample . As the processing for positive enrichment of CTCs could have potentially led to a loss of CTCs in our prior study, we tested a negative enrichment protocol of CTCs that features the immunomagnetic depletion of CD45+ white blood cells (WBCs). After validation of the new protocol, we demonstrated the suitability of using markers of EMT to detect EpCAM−/CK− cells and EMT-induced cells in peripheral blood. Thus, our multicolor CTC detection system provides a new method for the highly efficient detection of previously unrecognized populations of CTCs.
Materials and Methods
Buffer, Reagents, and Antibodies
MACS buffer contains 0.5% bovine serum albumin (NACALAI TESQUE, Kyoto, JAPAN) and 2 mM ethylenediaminetetraacetic acid (EDTA, SIGMA–ALDRICH, St. Louis, MO) in phosphate-buffered saline (Invitrogen, Carlsbad, CA). T-buffer contains 0.5% Immuno TOKUI (On-Chip Biotechnologies, Tokyo, Japan) in MACS buffer. Information regarding Immuno TOKUI was described by Takao and Takeda . BD Pharm Lyse™ lysing solution (10×) and Lyse/Fix Buffer (5×) were purchased from BD Biosciences (San Jose, CA). The FcR blocking reagent and CD45 Microbeads (Dynabeads®, DYNAL) were purchased from Miltenyi Biotec (Bergisch-Gladbach, Germany) and Invitrogen, respectively.
The fluorescein isothiocyanate (FITC)-conjugated anti-CK mAb CK3–6H5 and FITC-conjugated mouse IgG1 isotype control were purchased from Miltenyi Biotec. The phycoerythrin (PE)-conjugated anti-CD326 (EpCAM) mAb 9C4 and PE-conjugated mouse IgG2b, κ isotype control were purchased from BioLegend (San Diego, CA). Alexa Fluor 700-conjugated anti-CD45 mAb F10–89-4 was purchased from AbD Serotec (Oxford, UK). 7-Aminoactinomycin D (7-AAD) was purchased from Beckman Coulter (Marseille, France). The PE-conjugated anti-vimentin mAb and PE-conjugated rabbit IgG isotype control obtained from Cell Signaling (Danvers, MA). The transforming growth factor (TGF)-β1 was obtained from R&D Systems (Minneapolis, MN).
The non-small cell lung cancer (NSCLC) cell line A549 and the gastric tumor cell line KATO-III were obtained from the American Type Culture Collection (Manassas, VA) and RIKEN BioResource Center (Tsukuba, Japan), respectively. The NSCLC cell line PC-14 was kindly provided by Dr. Y. Hayata (Tokyo Medical College, Tokyo, Japan). The breast tumor cell lines MCF-7 and Hs578T were kindly gifted by Dr. Tohru Mochizuki (Shizuoka Cancer Center Research Institute, Japan). KATO-III, A549, and PC-14 cells were cultured in RPMI-1640 (Invitrogen) containing 10% fetal bovine serum (FBS, GIBCO, Life Technologies, Grand Island, NY). MCF-7 and Hs578T cells were cultured in Dulbecco's Modified Eagle's Medium (Invitrogen) containing 10% FBS. Cell lines were cultured under a humidified 5% CO2/95% atmosphere at 37°C. For spiking experiments, cells were harvested using 0.25% trypsin/EDTA (GIBCO) and then suspended in T-buffer immediately before the experiment.
Negative Enrichment of Tumor Cells
For the blood samples obtained from healthy donors, red blood cells were removed by applying 10 volumes of 1× Pharm Lyse™ lysing solution and incubating the mixture on ice for 20 min. After a 5-min centrifugation at 400g, the cell pellet was washed with T-buffer once and then resuspended in T-buffer containing FcR blocking reagent. After a 15-min incubation at 4°C, CD45-conjugated microbeads were added. The cell-microbead mixture was incubated at 4°C for 30 min with rotation, and the tube was placed on a magnet for several minutes. The supernatant was transferred to a new tube and then washed with T-buffer once. The cell pellet was resuspended in an appropriate volume of 1× Lyse/Fix buffer for fixing and then incubated for 20 min at room temperature.
The tumor cell-enriched samples were subsequently centrifuged at 400g for 5 min. After washing with T-buffer once, the cell pellet was dissolved in a staining solution containing CK-FITC, EpCAM-PE, 7-AAD, and CD45-Alexa Fluor 700. Samples were incubated overnight at 4°C in the dark. Next, unbound antibodies were removed via washing with 2 mL of T-buffer followed by centrifugation.
Flow cytometry was performed using a FISHMAN-R system (On-Chip Biotechnologies). The details of the FISHMAN-R system were described previously  and on manufacturer's web site (http://www.on-chip.co.jp/en/index.html). Briefly, this system is a disposable microfluidic chip-based flow cytometer, which realize absolute contamination free measurement, whole volume measurement, and sample collectivity after the measurement. In this study, wavelength range of four detection channels (FL1, 2, 3, and 4) was modified for the CTC detection. The ranges are 509–552 nm for FL1, 565–605 nm for FL2, 658–695 nm for FL3, and >700 nm for FL4. Signals for FITC, PE, 7-AAD, Alexa Fluor 647, and Alexa Fluor 700 were collected through these detection channels. Data analysis was performed using FlowJo software v7.6.5 (Tree Star, Ashland, OR).
Detection of Spiked Tumor Cells
Tumor cells were harvested by incubation with 0.25% trypsin–EDTA solution for several minutes at 37°C. Cells were washed and resuspended in T-buffer at a density of 1 × 104 cells/mL. This suspension was used to serially dilute the cells as needed for spiking experiments. After spiking an expected number of cells into 1 or 4 mL of blood obtained from a healthy donor, samples were processed according to the aforementioned protocol. For the detection of Hs578T cells, staining for vimentin-PE and CD45-Alexa700 was performed. This study was approved by the independent institutional review boards of Shizuoka Cancer Center and National Cancer Center.
In the enumeration comparison of the FISHMAN-R system versus the CellSearch platform, four tubes (two CellSave tube and two regular 5-mL blood collection tubes containing EDTA) of blood were collected from a healthy donor. After tumor cells were spiked into 7.5 mL of blood in the CellSave and blood collection tubes, the CellSave tube was delivered to an independent medical laboratory (SRL), and tumor cell enumeration was then performed using a CellSearch CTC kit. For the FISHMAN-R platform, the overall assay detection was interpolated between two tubes of blood.
Wound Closure Assay
The details of the wound closure assay were described previously . Briefly, A549 cells were grown in 12-well plates until confluence was achieved. After overnight culture in 1% serum medium, a scratch was made in each well using a sterile 200-µL pipette-tip and cellular debris was removed by washing with 1% serum medium twice. Cells were treated with or without 10 ng/mL TGF-β1 for 48 h. Phase-contract images of the scratch were taken under a microscope immediately after scratch creation and after 48 h.
After an overnight culture in 1% serum medium, A549 cells were treated with (post-EMT) or without (pre-EMT) 10 ng/mL TGF-β1 for 48 h. Pre- and post-EMT cells were spiked into 1 mL of blood obtained from healthy donors and processed according to the previously described negative enrichment protocol. The enriched cells were stained with CK-FITC, vimentin-PE, and CD45-Alexa700 for 12 h at 4°C in the dark.
Microsoft Excel 2010 software (Microsoft, Redmond, WA) was used for statistical analyses.
CTC Detection After CD45 Depletion: Assay Development
Discrimination of CTCs from the bulk of blood cells was achieved via a negative enrichment method using anti-human CD45 microbeads as described in Materials and Methods section.
Identification of the CTCs by a FISHMAN-R system using a PE against FITC density plot (Fig. 1A) was achieved after staining CTCs with anti-EpCAM-PE and anti-CK-FITC. DNA was stained with 7-AAD (Fig. 1B). Although negative selection using CD45 microbeads did not completely remove WBCs, CTCs that appeared in the FL1/FL4 density plot were easily distinguished from the WBC population (Fig. 1C).
It is assumed that CTCs with an epithelial phenotype will be stained with anti-CK-FITC and anti-EpCAM-PE antibodies, whereas cell nuclei are stained with 7-AAD. To identify a tumor cell, the cell must be positive for CK-FITC and/or EpCAM-PE (Fig. 1A), positive for 7-AAD (Fig. 1B), and negative for CD45-Alexa700 (Fig. 1C). The gate region was refined as described previously .
CTC Detection After CD45 Depletion: Detection and Specificity
The efficacy of the protocol was determined using serial dilutions of various tumor cells into 1 mL of blood. CTCs were defined according to the criteria CK+ and/or EpCAM+ and CD45− as shown in Figure 1. A typical example for the full gating strategy is shown in Supporting Information Figure S1. The expected number of KATO-III, A549, PC-14, or MCF-7 cells spiked into the healthy donor samples (i.e., 0, 10, 100, 500, or 1,000 cells) plotted against the actual number of cells observed in the samples is shown in Figure 2, and the results are summarized in Supporting Information Table S1. Regression analysis of the number of observed tumor cells versus the number of expected tumor cells produced slopes of 0.888, 0.836, 0.760, and 0.828, and determination coefficients (R2) of 0.997, 0.997, 0.998, and 0.994 in KATO-III, A549, PC-14, and MCF-7 cells, respectively. The mean percentages of detected cells were 85.5% ± 10.9%, 86.5% ± 8.1%, 72.6% ± 8.6%, and 78.4% ± 11.4% for KATO-III, A549, PC-14, and MCF-7 cells (n = 12), respectively. Analysis of non-spiked blood samples revealed only a few numbers of positive counts, confirming the specificity of the assay (Supporting Information Table S1).
Next, CTC enumeration and a head-to-head comparison with the CellSearch system were performed in a blinded manner at two different sites: the Shizuoka Cancer Center (SCC) and National Cancer Center (NCC). Unknown numbers of KATO-III cells were spiked into 4.5 mL of blood. The same samples were analyzed at two different sites and using the CellSearch system (7.5 mL) in parallel. The detection rate ranged from 89.0% to 120.0% at SCC and from 88.0% to 130.0% at NCC, suggesting the robustness of our detection system. In a head-to-head comparison study with the CellSearch system in a blinded manner, slopes of 0.74, 0.64, and 0.79 were observed at SCC and NCC and using the CellSearch system, respectively (Supporting Information Fig. S2). Although these results suggested the robustness of our system and comparability to CellSearch system, these needs to be further validated in future studies.
Detection of Tumor Cells Independent of EpCAM Expression
Next, we investigated the EpCAM independency of our platform to enumerate CTCs. As shown in Supporting Information Figure S3, KATO-III cells displayed high EpCAM expression, whereas A549 cells exhibited partial EpCAM expression. PC-14 cells lacked EpCAM expression. One hundred cells of these cell lines were spiked into 4.5 mL of blood obtained from a healthy donor, and then and enumeration was performed. Median detection rates of spiked tumor cells are described in Table 1. The EpCAM-based CellSearch enrichment method detected more EpCAM-high KATO-III cells (93.5%) than EpCAM-partial A549 (43.5%) and EpCAM-null PC-14 cells (1%) as expected. In contrast, no significant differences in detection rates were observed among EpCAM-high KATO-III (77.5%), EpCAM-partial A549 (75.5%), and EpCAM-null PC-14 (78.5%) cells (Table 1). These results suggest that our CTC enrichment method is advantageous for capturing EpCAM− tumor cells.
Table 1. Detection data from the spike-in experiment using tumor cell lines with different EpCAM expression levels in comparison with the CellSearch system
Percentage of cells detected (range)
Percentage of cells detected (range)
Data shown here are the representatives of two independent experiments for each assay.
77.5% ± 7.5% (70–85%)
93.5% ± 4.5% (89–98%)
75.5% ± 2.5% (73–78%)
43.5% ± 2.5% (41–46%)
78.5% ± 4.0% (75–82%)
1% ± 0.0% (1–1%)
Detection of EpCAM and CK Double-Negative Tumor Cells
A previous study revealed that some normal-like breast tumor cell lines lack EpCAM expression . We observed that the normal-like breast tumor cell line Hs578T did not express either EpCAM or CK (Fig. 3A) according to flow cytometry. However, high protein levels of the EMT marker vimentin were observed in Hs578T cells (Fig. 3A). Using PE-conjugated vimentin as a marker, EpCAM/CK double-negative Hs578T cells were spiked into blood obtained from healthy donors, and enumeration was performed. A population of Hs578T cells was identified using Alexa700 against PE density plots after staining CTCs with anti-CD45-Alexa700 and anti-vimentin-PE (Fig. 3B). The expected number of spiked Hs578T cells plotted against the actual number of cells observed in the samples is shown in Figure 3C, and the results are summarized in Supporting Information Table S2. An excellent overall detection rate (75.9% ± 3.9%, n = 12) and a linear performance (slope = 0.784, R2 = 0.999) were observed (Fig. 3C). The standard CTC detection marker (CK) yielded a detection rate of less than 5% (data not shown). Thus, vimentin may be an additional marker to facilitate the enumeration of tumor cell populations that lack epithelial properties.
Detection and Quantitation of Cells that Underwent EMT
One possible mechanism giving rise to the presence of EpCAM−/CK− CTCs is EMT, a biologic process reported to play a significant role in tumor progression and metastasis. Thus, there is growing interest in methods that enable researchers to detect and analyze CTCs that have undergone EMT. To determine whether our assay can detect cells that have undergone EMT, we examined a TGF-β1 (an inducer of EMT)-induced EMT model of A549 cells. After 48 h of treatment with TGF-β1, a morphologic change from an epithelial to a mesenchymal phenotype was observed in a majority of cells (Fig. 4A). A549 cells migrated faster to close the wound after treatment with TGF-β1 (Fig. 4B). The results of western blot analysis also supported the emergence of EMT in a time-dependent manner (Supporting Information Fig. S4). TGF-β1-treated and untreated cells were spiked into human blood obtained from healthy donors. Identification of the spiked tumor cells by the FISHMAN-R system using FITC against PE density plots was achieved after staining cells with anti-CK-FITC and anti-vimentin-PE (Fig. 4C). Gated spiked cells were analyzed on histogram (Fig. 4D). In total, 26% of the detected untreated cells were vimentin+; however, after TGF-β1 treatment, 77.2% of the detected cells were vimentin+ (Fig. 4D). This increasing of rate of vimentin+ cells was consistent with the results of western blot analysis (Supporting Information Fig. S4).
We developed a sensitive and reliable multicolor flow cytometry protocol for CTC detection. Spiking experiments revealed that our protocol with negative enrichment was validated with a mean detection rate of 79.8% (range 60.0–111.2%, n = 60). We were also able to detect and quantitate tumor cells that underwent EMT, a population of cells that has been hypothesized to be missed by current techniques .
The novel protocol was validated using various tumor cell lines spiked in blood obtained from healthy donors. Control tumor cells were used to assure proper performance of the protocol via linear regression analysis. Using the protocol, we detected more than 80% of KATO-III and MCF-7 cells, an acceptable finding considering the complexity of the assay procedure. We also used A549 and PC-14 cells as low and null EpCAM expression models, respectively. The detection rates for both cell lines were comparable to that of high EpCAM-expressing KATO-III cells, suggesting that our methods are capable of detecting and analyzing CTCs independently of EpCAM expression. Then, we proceeded to the spike-in experiment in 4 mL of blood, and this study was performed at two different sites in a blinded manner. The detection rate ranged from 89.0% to 120.0% at SCC and from 88.0% to 130.0% at NCC, confirming the robustness of our detection system across laboratories. In a head-to-head comparison study with the CellSearch system in a blinded manner, the average detection rates were 90.3% at SCC, 87.65% at NCC, and 62.0% for the CellSearch system.
Although analysis of the CD45− cell fraction confirmed the presence of spiked tumor cells in this fraction (Fig. 1C), there is a possibility that some of the spiked tumor cells were depleted together with the blood cells. This could mainly be explained by the nonspecific binding of the magnetic anti-CD45 antibodies. Further upgrading of processing capacity and improvement of the microfluidic chip should be conducted to eliminate the CD45 depletion step in the near future.
The use of EMT markers, for example, vimentin, facilitated the detection of EpCAM/CK double-negative tumor cells, a potentially important tumor cell population, in peripheral blood. Those populations do not appear to meet the current standard definition of CTC. The loss of EpCAM and/or CK in tumor cells has been reported previously [27-36]. Some tumors do not express EpCAM, as the downregulation of adhesion molecules is necessary for metastasis and migration [27, 28] and the programmed downregulation of EpCAM is a component of EMT . The number of CTCs may change as a result of chemotherapy because EpCAM can be suppressed by treatment . The loss of CK in tumor cells is a function of an independent oncogenic process [31-33] or EMT [19, 29]. CTCs with a hybrid phenotype (vimentin+/CK−) were previously detected in patients with ovarian, colorectal, prostate, breast, and NSCLC [34-36]. Full-spectrum identification of CTCs would permit a more efficient and sensitive analysis of patient samples containing heterogeneous CTC populations. Our multicolor flow cytometry system allows us to detect a CTC-positive marker and an EMT-related marker in parallel, suggesting that a population of CTCs that has been missed by current platforms might be detectable using our system. However, vimentin is also expressed on mesenchymal stromal cells which normally circulate at very low frequency . Criteria for vimentin+ CTCs must be carefully defined and evaluated in future clinical studies.
Recent reports suggested that isolated CTCs expressing markers of EMT were related to the metastatic status in patients with metastatic breast cancer receiving standard therapies . Similar reports identified greater numbers of CK+/vimentin+ CTCs in patients with metastatic breast cancer than in women with early-stage disease . The epithelial and mesenchymal composition of breast tumor cells and the number of clustered CTCs changed dramatically during treatment . Moreover, the extent of apoptosis in CTCs of patients with breast cancer was related to the stage of the disease . These studies suggested that CTCs have a heterogeneous phenotype in varied states. Although our system provides an efficient assay platform given its ability to detect and quantitate the numbers of CTCs that underwent EMT, detection of those CTCs with phonotypical diversity by FISHMAN-R needs to be evaluated in future clinical feasibility studies.
In conclusion, a system was developed and validated for the enumeration of CTCs and the detection of EpCAM−/CK− tumor cells in peripheral blood. We demonstrated that the method enables the detection of EMT-induced CTCs in sufficient quantities. Theoretically, the FISHMAN-R system is compatible with the detection of CTCs downstream of any enrichment protocol. These data highlight the potential of our detection system, and further evaluation with clinical samples should be conducted.
The authors thank Mrs. Junko Suzuki (Shizuoka Cancer Center) for technical assistance.