It has been estimated that 1,596,670 new cases of cancer will be diagnosed in the United States and 571,950 individuals will die from this disease in 2011 (1). The majority of these deaths are as a result of the development of metastases (2). These deaths are mainly due to the ineffectiveness of current therapies in treating metastatic disease and a general lack of understanding of the metastatic cascade. Metastatic disease has been correlated with the presence of circulating tumor cells (CTCs) in the blood (3). Detection of very small numbers of these rare cells has been shown to be predictive of overall survival in metastatic breast (4), prostate (5), and colorectal (6) cancer, where patients with ≥5 (breast and prostate) or ≥3 (colorectal) CTCs in 7.5 ml of blood have a poorer prognosis then those with fewer or no detectable CTCs.
Several methods have been utilized to enrich and detect CTCs, including density-gradient centrifugation (7, 8), immunomagnetic selection (9, 10), polymerase chain reaction (PCR)-based assays (11, 12), and flow cytometry (FCM) techniques (13, 14). All of these approaches have unique advantages and disadvantages; however, one commonality they all share is a lack of standardization; a necessity for use in the clinical setting. The development of the U.S. Food and Drug Administration (FDA) cleared CellSearch® system by Veridex provides a standardized method for the sensitive detection and quantification of these rare CTCs in human blood using fluorescence microscopy and immunology based techniques (4–6). This system is currently considered the gold standard in CTC enumeration and is the only CTC platform approved for in vitro diagnostic (IVD) use in the clinic at the present time.
The CellSearch system consists of two components, (i) the CellTracks® AutoPrep® system, which automates the blood sample preparation, and (ii) the CellTracks Analyzer II®, which scans the prepared samples. The CellTracks AutoPrep system uses an antibody mediated, ferrofluid-based magnetic separation technique and differential staining with fluorescent particles to distinguish CTCs from contaminating leukocytes in blood samples. Initially, the system performs a positive epithelial cell adhesion molecule (EpCAM) selection for CTCs using antiEpCAM antibodies conjugated to iron nanoparticles incubated in a magnetic field. The remainder of the fluid is then aspirated, selected tumor cells are resuspended, and fluorescently labeled antibodies and nuclear stain are added. This sample is then incubated in a magnetic cartridge, called a Magnest™, and scanned using the CellTracks Analyzer II.
The CellTracks Analyzer II utilizes a 10× objective lens to scan samples using different filters, each with the exposure time optimized to the appropriate fluorescent particle. CTCs are identified by binding anti-EpCAM, anti-pan-cytokeratin (CK)-phycoerythrin (PE) (CK8, 18, and 19), and the DNA stain 4′,6-diamidino-2-phenylindole (DAPI). Leukocytes are identified by staining with anti-CD45-allophycocyanin (APC) and DAPI. After the scan is complete, a gallery of computer-defined potential tumor cells is presented, from which the user must select, via qualitative analysis, which cells are CTCs based on the differential staining discussed earlier. The CellSearch system has been primarily utilized for the detection and enumeration of these rare cells. However, this platform does allow for single-cell characterization of CTCs for user-defined markers of interest, using an additional fluorescein isothiocyanate (FITC) fluorescence channel not required for CTC identification and enumeration (15). However, the detailed process for user-defined protein marker assay development and optimization using this platform is not well-defined.
Tumor profiling of metastatic lesions is not routine practice in the clinic. In fact, this profiling is often impractical or even impossible depending on the location and size of the metastatic tumors. Therefore, CTCs could act as a real-time, minimally invasive liquid biopsy, and the characterization of these rare cells could inform clinical decision-making. For example, human epidermal growth factor receptor 2 (HER2) is over-expressed in a subset of breast cancer patients, and has been exploited as a marker for targeted therapy using the HER2 receptor interfering monoclonal-antibody Herceptin® (16). However, this therapy has only been shown to be effective in patients whose primary tumor expresses sufficient levels of HER2. Fehm et al., demonstrated that approximately one-third of breast cancer patients with metastases whose primary tumors were HER2− had HER2+ CTCs (17, 18). Whether or not these patients with HER2+ CTCs would benefit from treatment with Herceptin still requires investigation; however, CTCs hold great promise for improving personalized cancer treatment.
Two general categories of protein markers are available for exploration on CTCs; markers that reflect tumor biology (tumor phenotyping) and may act as target molecules for therapy, and those that reflect cellular response to therapy. Currently, Veridex has developed and optimized two tumor phenotyping reagents for assessing well characterized therapeutic targets, and these are commercially available for research use only (RUO) applications on the CellSearch system. Using these reagents, CTCs can be analyzed for expression of either HER2/neu or epidermal growth factor receptor (EGFR). The development and optimization of CellSearch system assays for other user-defined markers of interest on tumor cells could identify new targets for novel therapies and enable a better understanding of the mechanisms that allow these cells to escape into the circulation, extravasate into distant tissue and form clinically relevant macrometastases.
The aim of this study was therefore to develop and optimize protocols for characterization of CTCs on the CellSearch platform for two proteins of interest, CD44 and M-30. CD44 has been associated with metastasis and has shown to be expressed by “cancer stem cells” (CSCs), a subpopulation of tumor cells that are believed to be the cells responsible for tumor initiation and metastasis (19). The ability to characterize and track CD44+ cells would therefore be an important tool for understanding the metastatic cascade. The M-30 CytoDeath antibody recognizes a neoepitope of CK18 that is exposed following caspase cleavage at residue 396 during the early events of apoptosis (20). This marker could be utilized as a measure of therapy effectiveness, and potentially indicate the necessity for a change in treatment much earlier than standard clinical evaluation techniques such as imaging. For the first time in the literature, the current study provides a detailed description of the process of user-defined protein marker development and optimization using the CellSearch system, and will be an important resource for the future development of protein marker assays by users of this platform.
MATERIALS and METHODS
Cell Culture and Reagents
The MDA-MB-468 human breast cancer cell line was obtained from Dr. Janet Price [M.D. Anderson Cancer Center, Houston, TX (21)]. The 21NT human breast cancer cell line was obtained from Dr. Vimla Band [Dana Farber Cancer Institute, Boston, MA (22)]. The LNCaP human prostate cancer cell line was obtained from Dr. John Lewis [London Regional Cancer Program, London, ON, Canada (23)]. MDA-MB-468 cells were maintained in αMEM + 10% FBS. 21NT cells were maintained in αMEM + 10% FBS, 3.32 μg/ml of HEPES, 1% nonessential amino acids (10 mM), 96.02 μg/ml of sodium pyruvate, 0.25 mg/ml of L-glutamine, 43.63 μg/ml of gentamicin, and 0.87 μg/ml of insulin (Sigma-Aldrich, St. Louis, MO). LNCaP cells were maintained in RPMI-1640 + 10% FBS. All growth media and supplements were obtained from Invitrogen (Carlsbad, CA). FBS was obtained from Sigma-Aldrich (St. Louis, MO). For induction of apoptosis, MDA-MB-468 human breast cancer cells were grown to ∼65% confluency at which time cells were treated with fresh growth media (αMEM + 10% FBS) or 0.1 μg/ml of paclitaxel (Biolyse Pharma, St. Catherines, ON, Canada) in αMEM + 10% FBS. Both the treated and untreated cells were then grown for an additional 48 h before they were harvested and analyzed as described below.
Flow Cytometry Sample Preparation
For comparison of FITC mean fluorescence intensity (MFI), measured on a four decade log scale, MDA-MB-468 cells were spiked into whole blood at a concentration of ∼0.01% (equivalent to ∼200 MDA-MB-468 cells) of the total white blood cell (WBC) count determined by analysis of whole blood using an LH 780 hematology analyzer (Beckman Coulter, Hialeah, FL). The 0.01% spiked solution was then incubated with fluorescently conjugated antibodies for 20 min at room temperature, including 5 μl of anti-CD45-PC5 (clone J33; Beckman Coulter/Immunotech, Marseille, France) and either 150 μl of anti-EGFR-FITC (Veridex, LLC, Raritan, NJ) or various amounts (0.05–0.5 μg) of anti-CD44-FITC (clone G44-26; BD Biosciences). Following incubation, red blood cells were lysed using 1× ammonium chloride lysing solution (Beckman Coulter/Immunotech) for 10 min at room temperature. Samples were then analyzed using a Beckman Coulter EPICS XL-MCL flow cytometer.
For analysis of differential CD44, EpCAM, and CK expression in the three cell lines, MDA-MB-468, 21NT, and LNCaP tumor cells (5 × 105) were resuspended in flow buffer (PBS + 2% FBS), fixed and permeabilized, as necessary (CK), using the IntraPrep™ Fix/Perm kit (Beckman Coulter, Fullerton, CA), and incubated with either 150 μl of anti-EGFR-FITC (Veridex; 20 min), 0.5 μg of anti-CD44-FITC (clone G44-26; BD Biosciences; 20 min), 0.0075 μg of anti-EpCAM-PE (clone EBA-1; BD Biosciences; 20 min), or 25 μl of anti-CK-8/18/19-PE (Veridex; 30 min). Following two washes with an excess volume (≥2 ml) of flow buffer, samples were analyzed using a Beckman Coulter EPICS XL-MCL flow cytometer. PBS and 5.0 μg of FITC-conjugated mouse IgG2bK (clone MPC-11; BD Biosciences) or 0.0625 μg of PE conjugated mouse IgG2bK (clone 27–35; BD Biosciences) were used as negative controls.
For analysis of M-30, paclitaxel-treated and untreated MDA-MB-468 cells (5 × 105) diluted in flow buffer were fixed and permeabilized using the IntraPrep Fix/Perm kit (Beckman Coulter) and subsequently incubated with 0.1 μg of anti-M-30-FITC (Alexis Biochemicals, Lausen, Switzerland; 30 min) and either 0.0075 μg of anti-EpCAM-PE (clone EBA-1; BD Biosciences; 20 min), 25 μl of anti-CK-8/18/19-FITC (Veridex; 30 min), or 0.07 μg of DAPI (Sigma-Aldrich; 15 min). Following two washes with an excess volume (≥2 ml) of flow buffer, samples were analyzed using either a Beckman Coulter EPICS XL-MCL flow cytometer (EPCAM/CK) or a Beckman Coulter Navios flow cytometer (DAPI). All FCM analysis was performed between October 2008 and August 2011.
It is important to note that FCM was used throughout this study as a method to determine the percentage of cells within any given cell population that were positive for our marker of interest (i.e., EGFR, CD44, and M-30) and not as a method with which to compare to the CellSearch system. Although FCM would seem like an ideal method for CTC characterization based on the results presented in this study, its lower sensitivity for rare-event detection make it less than ideal for use in a clinical setting, which further highlights the need to optimize appropriate protocols for CTC characterization on the gold standard CellSearch system.
Flow Cytometry Analysis and Gating Strategy
The flow procedures were performed on either a single laser (488 nm Argon Laser) four-color Beckman Coulter XL-MCL flow cytometer or a three laser (405 Violet, 488 Argon, 633 Helium-Neon Laser) 10 color Beckman Coulter Navios flow cytometer. Alignment and calibration checks were performed daily on both flow cytometers using FlowCheck/FlowSet (for XL-MCL flow cytometer) and FlowCheckPro/FlowSetPro (Beckman Coulter Navios flow cytometer). Fluorochrome emission on the Beckman Coulter XL-MCL was captured for FITC conjugates using a 525/40 nm band pass (BP) filter and for PE conjugates using 575/25 nm BP filter. Emissions on the Beckman Coulter Navios cytometer for FITC, PE, and DAPI were collected using 525/40 nm, 575/30 nm, and 450/40 nm BP filters, respectively. Antibody combinations are detailed elsewhere in the Materials and Methods section.
For all assays, forward and side scatter was used for gating to eliminate cellular debris and cell doublets from analysis. A minimum of 10,000 events were acquired for all analyses. For single-color assays, gates were set based on either FMO or appropriate IgG controls. For two color assays, an FMO strategy was used to determine compensation coefficients. Gating for blood samples spiked with tumor cells was set up using 10% tumor cell spiked samples (compared to WBC count). Tumor cells were defined as CK+CD45− and/or EGFR+/CD44+, whereas WBCs were defined as CK−CD45+ and/or CD44−/EGFR−. Relative MFI was determined based on IgG isotype or FMO controls (as appropriate) in order to demonstrate changes in expression of the specific marker in question. Populations of interest were observed to be single populations (i.e., similar to what is seen in Fig. 1A), and means were observed to be comparable to medians. Therefore, the mean was taken as the value for determining MFI. All data were analyzed offline using the Kaluza™ Beckman Coulter analysis software.
Blood from healthy volunteer donors was drawn into 10 ml CellSave preservative tubes (Veridex) containing EDTA and a proprietary cellular preservative. Blood in excess of 7.5 ml (the standard blood volume for use with the CellSearch system) was removed from each CellSave preservative tube and discarded. One thousand (EGFR/CD44) or 4000 (M-30) cultured human tumor cells from the appropriate cell line, diluted in 100 μl of flow buffer, were spiked into each CellSave preservative tube. To avoid sample degradation, spiked blood was processed on the CellSearch system within 96 h of blood collection as per the manufacturer's guidelines. Following thorough mixture by inversion, 7.5 ml of spiked blood was collected from each CellSave preservative tube and mixed with 6.5 ml of dilution buffer (Veridex). Spiked blood samples were mixed by inversion five times and then centrifuged at 800g with the brake off for 10 min at room temperature. Prepared blood samples were loaded into the CellTracks AutoPrep system as per the manufacturer's guidelines and processed using the CellSearch CTC kit as described in the Introduction section.
Undiluted anti-EGFR-FITC (Veridex) was added to Veridex-supplied reagent cups in the CellSearch kit cartridge for processing on the CellTracks AutoPrep system. Anti-CD44 antibodies conjugated to either FITC (clone G44-26; BD Biosciences) or PE (clone G44-26; BD Biosciences) and FITC-conjugated M-30 CytoDeath antibodies (Alexis Biochemicals) were diluted according to Veridex recommendations (24) in 1× ultrapure PBS (Invitrogen) to obtain the desired working concentration. Ultrapure PBS and PE conjugated mouse IgG2bK (clone 27–35; BD Biosciences) diluted in ultrapure PBS at the corresponding working concentration of anti-CD44-PE were used as negative controls. For optimization, two variables were altered: the antibody concentration utilized on the CellTracks AutoPrep, and the length of time the FITC/PE fluorophore was exposed to the laser (altered using the “Research Protocol” option under the “Set Up” tab, on the CellTracks Analyzer II). All CellSearch analysis was performed between October 2008 and August 2011.
Statistical analysis was performed using GraphPad Prism® 5.0 (La Jolla, CA). When analyzing two groups of data the differences between the means was determined using a Student's t test. M-30 samples run in parallel were analyzed using a paired Student's t test. When analyzing three or more groups of data, two-way ANOVA followed by a Bonferroni post-test was used. In all cases P < 0.05 was considered to be statistically significant.
Initial CellSearch System Validation with a Commercially Available Optimized Marker
To first determine the sensitivity and specificity of the CellSearch system to detect an optimized marker in the FITC channel, a commercially available marker from Veridex, anti-EGFR-FITC, was processed with blood spiked with MDA-MB-468 human breast cancer cells on the CellSearch system. MDA-MB-468 cells alone were first analyzed by FCM to determine the percentage of the cell population that was EGFR+. These cells were found to be highly EGFR+, with >99.9 ± 0.03% of cells expressing EGFR (Figs. 1A and 1C). Blood from a healthy volunteer donor was then spiked with MDA-MB-468 cells and processed on the CellSearch system with the commercially available anti-EGFR-FITC antibody. The CellSearch system was able to characterize 95.1 ± 2.34% of spiked cells as EGFR+ (Figs. 1B and 1C). Comparison of the ability of these two techniques for analyzing the percentage of the cell population that was EGFR+ demonstrated that markers that have been properly optimized for use with the CellSearch system are not significantly different from results obtained by FCM (Fig. 1C).
CD44 Protocol Development for Use With the CellSearch System
FCM experiments were next performed to determine the optimal concentration of anti-CD44-FITC required for initial testing on the CellSearch system. Blood samples spiked with MDA-MB-468 cells at a concentration equivalent to ∼0.01% of the WBC count (equivalent to ∼200 MDA-MB-468 cells) were incubated with anti-EGFR-FITC and analyzed by FCM. To ensure appropriate gating of tumor cells, 1% (compared with WBC count) spiked samples were utilized. The sample showed a signal-to-noise ratio of 268.5. Several concentrations of anti-CD44-FITC were then analyzed, and based on the FITC MFI values, the anti-CD44-FITC concentration that gave the closest signal-to-noise ratio to that of the anti-EGFR-FITC was 3.5 μg/ml, with a signal-to-noise ratio of 257.5 (Table 1 and Supporting Information Fig. S1).
Table 1. Optimization of CD44 antibody concentration by flow cytometry
Signal to noise ratio
On the basis of these results, anti-CD44-FITC diluted in 1× ultrapure PBS at a working concentration of 3.5 μg/ml was chosen as the initial concentration to use with the CellSearch system. As in Figure 1, MDA-MB-468 cells were first analyzed by FCM to determine the percentage of the cell population that was CD44+. These cells were found to be highly CD44+ with 98.4 ± 0.90% of cells expressing CD44 (Figs. 2A and 2C). Blood samples spiked with MDA-MB-468 cells were then analyzed using various exposure times and increasing amounts of anti-CD44-FITC on the CellSearch system using the CTC kit. After extensive optimization, the highest percentage of CD44+ cells that could be obtained on the CellSearch system was 69.3 ± 2.67% at a concentration of 4.0 μg/ml and an exposure time of 0.5 s (Figs. 2B and 2C), and this was still significantly different (P = 0.0005) than the CD44 positivity results obtained by FCM (Fig. 2C), and thus were not considered to be suitably optimized. The discrepancy observed between CD44+ cells identified by FCM and the CellSearch system was thought to be a result of either (i) the FCM assay not taking into account the EpCAM and CK8/18/19 expression of the cells; and/or (ii) the CellSearch system not detecting cells expressing lower levels of CD44. Upon analysis by FCM, lack of EpCAM, and/or CK 8/18/19 expression did not appear to contribute to the observed discrepancy, since 99.5 ± 0.47% and 99.0 ± 0.70% of cells were EpCAM+ and CK8/18/19+, respectively (Supporting Information Figs. S2A,C and S3A,C). Therefore, the CellSearch CXC kit, optimized for use with lower antigen density markers, was investigated.
CD44 Protocol Optimization Using the CellSearch CXC Kit
The CellSearch CXC kit differs from the FDA-cleared CTC kit in that the fluorescence detection of CK8/18/19, normally represented in the PE channel, and the user's marker of interest, normally represented in the FITC channel, are reversed. Therefore the PE channel represents the user's marker of interest and the FITC channel represents CK8/18/19. This change allows for better visualization of lower antigen density markers, ∼50,000 antigens/cell on the CXC kit versus ∼100,000 antigens/cell using the CTC kit (24). MDA-MB-468 cells were analyzed by FCM for CD44 positivity using an anti-CD44-PE antibody and were again found to be highly CD44+, with 99.5 ± 0.39% positive for CD44 (Figs. 2D and 2F). Blood samples spiked with MDA-MB-468 cells were then analyzed using 1.0 μg/ml of anti-CD44-PE (equivalent to the volume used for 4.0 μg/ml of anti-CD44-FITC) and an exposure time of 0.6 s, the maximum recommended by Veridex for the PE channel, as a starting point. Using these parameters, 98.8 ± 0.51% of cells were observed to be CD44+, a value not significantly different from that obtained by FCM (Figs. 2E and 2F).
After demonstrating that our developed marker was capable of identifying CD44+ cells to an acceptable degree using the CellSearch CXC kit, additional optimization was required before the assay could be considered for use in a clinical setting. Three cell lines with varying CD44 expression levels were chosen for this purpose: MDA-MB-468 human breast cancer cells (high expression), 21NT human breast cancer cells (low expression), and LNCaP human prostate cancer cells (no expression) (Fig. 3A). As with the MDA-MB-468 cells, the 21NT cells and LNCaP cells were analyzed for EpCAM and CK8/18/19 positivity to ensure suitable identification by the CellSearch system (Supporting Information Figs. S2A,C and S3A,C). All three lines were highly EpCAM+ and CK8/18/19+, 99.5 ± 0.47%/99.0 ± 0.70% (MDA-MB-468), 87.2 ± 7.35%/98.6 ± 0.67% (21NT), and 98.4 ± 0.97%/99.2 ± 0.20% (LNCaP), respectively, with varying antigen densities represented as EpCAM and CK8/18/19 MFI, measured on a four decade log scale, 24.7 ± 4.49/69.0 ± 19.7 (MDA-MB-468), 11.5 ± 3.05/70.7 ± 15.8 (21NT), and 16.4 ± 1.97/37.5 ± 3.22 (LNCaP), respectively (Supporting Information Figs. S2B and S3B). All three cell lines spiked into healthy volunteer blood samples were initially processed using the CellSearch CXC kit along with 1× ultrapure PBS to determine the level of background noise at various exposure times (0.05–0.6 s) and compared to their respective FCM results. At all exposure times tested, no significant differences were observed between the positivity levels obtained on the CellSearch system and those obtained by FCM (Figs. 3B–3D). However, background noise does appear to increase, although not significantly, with increasing exposure time, and therefore exposure time should be kept to a minimum.
As demonstrated, MDA-MB-468 cells spiked into blood and processed with 1.0 μg/ml of anti-CD44-PE on the CellSearch system (CXC kit) at an exposure time of 0.6 s resulted in a similar percentage of CD44+ cell detection to that observed by FCM (Fig. 2F). However, with the goal of keeping exposure times to a minimum, various additional exposure times were investigated. Results demonstrate that 1.0 μg/ml of anti-CD44-PE was adequate for identifying CD44 positivity in MDA-MB-468 cells (high CD44-expressing cell line) at all exposure times except 0.05 s (Fig. 3B). However, using 1.0 μg/ml of anti-CD44-PE in 21NT cells (low CD44-expressing cell line), we observed that only exposure times of 0.4 and 0.6 s were adequate for identifying CD44+ cells (Fig. 3C). Therefore, the concentration of anti-CD44-PE was increased to 1.5 μg/ml. Upon analysis of both the MDA-MB-468 and 21NT cells on the CellSearch system (CXC kit) at this increased antibody concentration, CD44 positivity of MDA-MB-468 cells was adequate at all exposure times except 0.05 s, and CD44 positivity of 21NT cells was now deemed to be adequate at 0.2 and 0.4 s (Figs. 3B and 3C). A higher percentage of cells were detected at 0.6 s; however, this was accompanied by an increase in false positive cells in the CD44− LNCaP cell line control sample using 1.5 μg/ml (Fig. 3D). On the basis of these results, an antibody concentration of 1.5 μg/ml was determined to be optimal for CD44 visualization as it allowed the use of lower exposure times, thereby minimizing background noise and false positive results. Finally, spiked blood samples for all three cell lines were processed with appropriate IgG controls on the CellSearch system and analyzed at various exposure times to ensure antibody specificity (Figs. 3B–3D). On the basis of these results, an antibody concentration of 1.5 μg/ml and an exposure time of 0.2 s were chosen as the optimized parameters for this marker. Additional CellSearch gallery images for each cell line under optimized conditions can be found in Supporting Information Fig. S4.
M-30 Protocol Development for Use with the CellSearch System
Next, we chose to investigate M-30, a marker of cellular death. FCM experiments were initially performed to determine the optimal concentration of paclitaxel (0.1 μg/ml) to use for apoptosis induction of the MDA-MB-468 cells (data not shown). Paclitaxel-treated and untreated cells were then processed in parallel on the flow cytometer and the CellSearch system to control for various confounding factors (i.e., confluency of the cells at time of treatment, slight variations in drug dilution, etc.) that would make direct comparison of the M-30+ cells difficult. It should be noted that, during the development and optimization of the M-30 protocol, images in the FITC channel were observed to be clear, bright, and easily identifiable as M-30+ cells. This was not the case, however, with CD44, where many images showed a faint signal in the FITC channel, but appeared grainy in nature and were therefore not clear and bright enough to be classified as CD44+. On the basis of these results, the CXC kit was not utilized in the development of the M-30 protocol as the images generated using the CellSearch CTC kit were of the same quality as those shown using the CellSearch CXC kit for CD44 and therefore use of this kit would likely not have significantly increased the number of M-30+ CTC. (Supporting Information Fig. S5).
By FCM, untreated cells showed very low levels of M-30 positivity, with 0.8 ± 0.13% being M-30+, and a small proportion of the paclitaxel treated cells (17.6 ± 1.18%) were classified as apoptotic (M-30+), based on gates set at the outermost limits of an untreated control incubated with equivalent volumes of M-30 antibody (Figs. 4A and 4C). After extensive optimization on the CellSearch system (CTC kit) 3.5 μg/ml of anti-M-30-FITC and an exposure time of 0.8 s produced the highest percent of M-30+ treated cells, with 10.9 ± 2.42% being M-30+. In addition, this concentration and exposure time when processing blood spiked with untreated cells showed M-30 positivity that was not significantly different from the FCM results (Figs. 4B and 4C).
In an attempt to explain this observed discrepancy, it was necessary to determine if paclitaxel treatment would affect the ability of the CellSearch system to detect the treated CTCs by decreasing the MFI of EpCAM, CK8/18/19, and/or DAPI. Following 48 h of treatment with 0.1 μg/ml of paclitaxel, treated and untreated cells were incubated with anti-M-30-FITC and either anti-EpCAM-PE, anti-CK8/18/19-PE, or DAPI and were analyzed by FCM. Following paclitaxel treatment, there was no significant loss of EpCAM or CK positivity, with 97.5 ± 1.51% and 97.7 ± 0.49% of cells observed to be EpCAM+ and CK+, respectively in untreated samples versus 89.6 ± 5.89% and 91.5 ± 3.88%, respectively in paclitaxel-treated samples. In addition, following treatment the vast majority of cells that were apoptotic (M-30+) also appeared to be EpCAM+ (22.7 ± 3.41% [M-30+] versus 19.9 ± 4.62% [M-30+EpCAM+]) and CK+ (24.6 ± 3.13% [M-30+] versus 24.9 ± 2.26% [M-30+CK+]) and therefore a significant population of M-30+/EpCAM− or M-30+/CK− cells (that would not be captured by the CellSearch system) was not found (Figs. 5A and 5B). However, following paclitaxel treatment there was a significant decrease in DAPI MFI relative to untreated cells, 60.4 ± 1.10% versus 96.9 ± 0.46%, respectively, and there was a significant difference between the apoptotic (M-30+) cells and those that were both apoptotic (M-30+) and DAPI+, 17.6 ± 1.18% versus 8.7 ± 0.79%, respectively (Fig. 5C). It is important to note that this decrease in DAPI MFI did not necessarily constitute cells becoming DAPI−. Instead cells that expressed high levels of DAPI (DAPIhi phenotype) prior to treatment appeared to exhibit decreased DAPI staining following treatment (DAPIlo phenotype), which could therefore affect adequate visualization of DAPI positivity on the CellSearch system. This decrease in DAPI MFI lead us to further investigate the percentage of cells that were M-30+DAPIhi by FCM compared with those that were M-30+DAPI+ by the CellSearch system (since DAPI positivity is a requirement to classify an event as a CTC on the CellSearch system). This comparison appeared to rectify the observed difference. However, to validate this observation, reanalysis of the CellSearch data was necessary, to determine if inclusion of those cells that were not originally classified as a CTC due to poor DAPI staining would increase the number of M-30+ cells to that originally observed by FCM (17.6 ± 1.18%). However, upon reanalysis the percentage of M-30+ cells increased only marginally and this increased value was still significantly different (paired t test) from the M-30+ data obtained by FCM (data not shown).
The majority of cancer-related deaths are because of ineffective treatment of metastatic disease and an incomplete understanding of the biology of metastasis. Advances in the area of CTC detection and enumeration allows for investigation of the early stages of metastasis that until recently was limited by technological challenges. The characterization of CTCs could be a powerful clinical tool, acting as a real-time, minimally invasive liquid biopsy that would inform clinical decision-making and help direct tailored, individualized therapy. In the present study, our aim was therefore to develop protocols that would allow for the characterization of CTCs using the U.S. FDA-cleared CellSearch system. To the best of our knowledge, this is the first study in the literature to describe the detailed process of protocol development and optimization using this platform. We have demonstrated the appropriate steps that must be taken for proper optimization of user-defined protein marker assays on this system, including comparison of results with a well validated protein expression technology (FCM); appropriate troubleshooting; and detailed optimization techniques using cell lines with various target marker antigen densities. In addition, we have demonstrated that not all markers are ideal candidates for use with the CellSearch system.
Although previous studies have examined CTCs for the expression of CD44 (25–27), none of these previous studies have utilized the CellSearch system to do so. As this system is still the gold-standard and the only FDA-cleared instrument for CTC enumeration and clinical decision-making, the ability to characterize CTCs in combination with enumeration using this particular platform is more clinically applicable then the ability to do so using other techniques. Additionally, Rossi et al., 2010, is the only published study (to our knowledge) that has utilized user-defined, non-Veridex optimized protein markers assays (specifically M-30), on the CellSearch system (28). However, this manuscript fails to provide details of proper optimization of the protocol for the use of this protein marker on the CellSearch. This highlights the necessity for the current study, which provides a detailed description of the process of protein marker optimization in order for users of this instrument to develop properly optimized protein marker protocols that could be utilized in a clinical setting.
The advantages of utilizing the CellSearch platform include system standardization; clearance by the U.S. FDA for assessment of prognosis in metastatic breast, prostate, and colorectal cancer; and the ability to examine CTC heterogeneity at the single cell level. However, the CellSearch system does have a number of disadvantages, including the use of the epithelial marker EpCAM for CTC enrichment. Others have shown that many of the tumor cells found in the circulation are actually mesenchymal in phenotype and are therefore potentially undetectable by this system due to a lack of EpCAM expression (29–32). In addition, it has been demonstrated that the epithelial-to-mesenchymal transition (EMT), which may produce these undetectable CTCs, is associated with enhanced cell aggressiveness (33, 34). Therefore, cells that may potentially form metastatic lesions and are of great interest for characterization may be missed by the CellSearch system. New CTC platforms are currently under development, and hold great promise for enhanced CTC detection and characterization (35–36). However, the CellSearch system, although not a perfect detection and characterization platform, is currently the clinical gold standard for CTC analysis, and for this reason we explored the development of additional CTC characterization assays using this system.
We initially began protocol development for the CD44 marker using the CellSearch CTC kit, with limited success. We hypothesized that the low CD44 positivity results might be due to contamination with leukocytes expressing CD44, thereby simulating a situation in which cells appear to have a lower than expected antigen density. We tested this hypothesis by utilizing the CellSearch CXC kit, which is optimized for the visualization of lower antigen density markers. Utilization of this kit resolved this observed discrepancy, showing levels of CD44 expression that were not significantly different from those observed by FCM. Three cell lines with various CD44 expression levels were then chosen to optimize this protocol. This was a necessary step in the optimization process as the CD44 antigen density in patient CTCs is unknown, and likely to be quite variable across patient samples. As demonstrated, the 21NT cell line (low CD44-expressing) was unable to be adequately visualized at a low exposure time using 1.0 μg/ml of anti-CD44-PE, and therefore the concentration had to be increased to 1.5 μg/ml to ensure adequate sensitivity. In addition, the LNCaP cell line (CD44−) was utilized to ensure specificity of the assay protocol.
As with all assays, limitations do exist when utilizing the CellSearch system for the visualization of CD44. CD44 is a marker of cancer stem cells (19), a phenotype that has been associated with EMT (37). There is always the possibility that CTCs from patient samples may express CD44, but largely by those cells that are undetectable by the CellSearch system, due to a lack of or low level of EpCAM expression. In addition, we have demonstrated that leukocytes can affect adequate CD44 visualization. Therefore, this assay could be compromised in patients with exceptionally high levels of contaminating leukocytes. However, we are confident that this assay is appropriately optimized for use in future clinical studies of metastatic cancer patients.
Next we investigated a different type of marker, one that measures cellular death in response to therapy. We attempted to optimize the integration of the early apoptosis marker M-30 with the CellSearch system. However, after extensive experimentation, we have demonstrated that this marker is unlikely to ever capture all early apoptotic cells in patient samples, as these results were unachievable even under highly controlled conditions. The lack of optimal M-30 positivity on the CellSearch system did not appear to be as a result of a decrease in either EpCAM or CK MFI in our paclitaxel-treated cells. However, when nuclear staining using DAPI was investigated by FCM, there did appear to be a significant decrease in the percentage of overall DAPIhi and M-30+DAPIhi cells, which could affect adequate visualization of DAPI positivity, and therefore CTC classification, on the CellSearch system. When only the M-30+DAPIhi FCM results were compared to our M-30+DAPI+ CellSearch data this appeared to rectify the observed difference. To validate this observation, reanalysis of the CellSearch data was performed to determine if inclusion of those cells that were not originally classified as a CTC due to poor DAPI staining would increase the number of M-30+ cells to that originally observed by FCM. However, reanalysis only marginally increased the percentage of M-30+ cells on the CellSearch system. This increased value was still significantly different from the M-30+ data obtained by FCM and therefore could not be a plausible explanation for why M-30 positivity using the CellSearch system was lower than that observed by FCM. It is possible that spiking cells that are in the process of cell death could reduce CTC recovery as some of these cells may be damaged during the prespiking preparation and others during CellSearch sample processing. Therefore, experiments would need to be performed to determine M-30+ CTC recovery, using spiked samples with high, medium, and low numbers of M-30+ CTCs, and to demonstrate that these recovery results are reproducible across laboratories before this protocol could be considered for clinical use. However, based on the results obtained in our study and our primary aim of demonstrating proper protocol development and troubleshooting, we will not be moving forward with this marker as we believe that there are other markers that are likely better suited to identifying therapy response.
In previous work by Rossi et al. (2010), attempts were made to utilize the M-30 assay on the CellSearch system (28). However, direct comparison between M-30 positivity by FCM and the CellSearch system was never performed; instead CellSearch results were compared to Annexin V positivity, with somewhat discordant results. The authors attempt to explain this difference as a result of nonclassification of some events as CTCs on the CellSearch system; however, experimentation was not undertaken to confirm this hypothesis. In addition, optimal M-30 antibody concentration (2 μg/ml) and exposure time (0.4 s) were chosen based on results from three patients with probable M-30+ CTCs, not in treated control spiked blood samples, as shown in this study.
The M-30 assay has potential for utilization as a measure of therapy effectiveness and as an early indicator of the necessity for a change in treatment. However, this assay is only able to identify apoptosis, one of many known mechanisms of cellular death (necrosis, autophagy, and mitotic catastrophe) (20). This presents a potential problem when examining CTC death as a marker of therapy effectiveness, as it is likely that not all therapy-induced cellular death will be apoptotic (38–40). Therefore, the discrepancies observed in the visualization of this marker in the present study could be a result of the overall complexity and lack of a complete understanding of cellular death in response to therapy.
Research in this area has led many in the field to believe that there is much overlap in the mechanisms that underlie these cellular death processes (38–40) and that cell death proceeds along a pathway in which many possible outcomes can occur. Therefore, many cells undergoing cellular death may be missed using this assay. An ideal apoptotic marker would measure all types of cell death; however, such a marker does not yet exist. The next ideal candidate would measure the most prevalent form of cellular death and be present for the longest detectable period of time. However, even this may be difficult to achieve. Others have demonstrated that the relative proportion of cells that undergo apoptotic cell death can change quite dramatically based on the stressor applied, stressor intensity (concentration and/or length of application), and the cells induced to undergo apoptosis (38). Problems therefore arise when examining patient samples as proper timing may be necessary to capture cell death in the appropriate state. Even if ideal conditions were satisfied, highly vascularized versus poorly vascularized /hypoxic tumors may respond differently (i.e., different cellular death pathways) to antitumor agents due to differences in drug concentrations received. The question then becomes whether the identification of all early apoptotic CTCs is necessary for prediction of therapeutic efficacy and clinical decision making. Instead, could the identification of any apoptotic cells represent a favorable prognosis for patients? In the study by Rossi et al. (2010) (28), CTC analysis was performed in blood from eight breast cancer patients using the integrated M-30 assay on the CellSearch system (2 μg/ml and exposure time 0.4 s). The change in the number of live verses dead (apoptotic) CTCs was determined and found to correlate with radiologic findings of disease status (progressive versus stable disease/partial response), as determined by radiology, with 100% concordance. Obviously, larger follow-up studies will have to be performed before any meaningful conclusions can be drawn from these data, but the results do appear promising.
In summary, we have demonstrated the detailed process of optimization that is required for the development of a user-defined marker on the CellSearch system. In addition, we have shown that not all markers are suitable candidates for use with this platform and the necessary troubleshooting that must be performed when dealing with markers that might alter CTC identification characteristics. This study will act as an important troubleshooting guide for the future development of protein marker assays by users of this platform.