Size-Based Differentiation of Cancer and Normal Cells by a Particle Size Analyzer Assisted by a Cell-Recognition PC Software

Abstract

Detection of anomalous cells such as cancer cells from normal blood cells has the potential to contribute greatly to cancer diagnosis and therapy. Conventional methods for the detection of cancer cells are usually tedious and cumbersome. Herein, we report on the use of a particle size analyzer for the convenient size-based differentiation of cancer cells from normal cells. Measurements made using a particle size analyzer revealed that size parameters for cancer cells are significantly greater (e.g., inner diameter and width) than the corresponding values for normal cells (white blood cells (WBC), lymphocytes and splenocytes), with no significant difference in shape parameters (e.g., circularity and convexity). The inner diameter of many cancer cell lines is greater than 10 µm, in contrast to normal cells. For the detection of WBC having similar size to that of cancer cells, we developed a PC software “Cancer Cell Finder” that differentiates them from cancer cells based on brightness stationary points on a cell surface. Furthermore, the aforementioned method was validated for cancer cell/clusters detection in spiked mouse blood samples (a B16 melanoma mouse xenograft model) and circulating tumor cell cluster-like particles in the cat and dog (diagnosed with cancer) blood samples. These results provide insights into the possible applicability of the use of a particle size analyzer in conjunction with PC software for the convenient detection of cancer cells in experimental and clinical samples for theranostics.

The occurrence of anomalous cells deviating from the original cells may be a symptom of a diseased state in a body, hence their detection and classification are very important. One typical example of diseases that are characterized by the uncontrolled division and growth of abnormal cells would be cancer. Some cancer cells, referred as circulating tumor cells (CTCs), spread and reach an arresting state beyond the original tumor sites to secondary organs (metastasis), which is a prime cause of cancer-associated mortality. Therefore, quick and convenient detection of cancer is highly desirable for diagnosis and treatment in personalized (precision) cancer medicine.^1–7)

Localized cancers can be detected by a histopathological examination of suspicious tissue, but this technique has constraints due to the fact that the analysis varies from person to person according to their visual acuity and experience.⁸⁾ On the other hand, it has been reported that non-adherent abnormal CTCs can be detected based on the expression of tumor antigens and biophysical deformities.^9–17) The positive expression of CTC protein markers such as cytokeratin-8, -18 and -19, CD44 and the epithelial cell adhesion molecule (EpCAM) is often found in a CTC.¹⁸⁾ To date, the CellSearch system (Veridex, LLC, U.S.A.), which is based on immunostaining for CKs⁺, CD45⁻and nucleus⁺, is the only device that is currently approved by the Food and Drug Administration for CTC detection,.^19–21) Since the expression of CTC markers has been reported to vary with cancer heterogeneity, dynamic protein turnover, and during epithelial-to-mesenchymal transition (EMT), immunological based detection may be a less reliable approach.^22–24)

The capture, isolation and enrichment of non-adherent abnormal CTC are based on their difference from hematologic cells in size and biophysical deformities such as shape, stiffness and electrical polarizability.^{12–17,25–29)} Some studies have reported that tumor cells are larger in size as compared to hematologic cells and this difference accounts for the basis for size-based isolation using filtration systems and Isolation by Size of Epithelial Tumor cells (ISET, Rarecells) etc.^30,31) In a recent study, CellSearch confirmed CTC’s (EpCAM⁺, CK⁺, DNA⁺, CD45⁻) median size was 10.7 µm (prostate), 11 µm (colorectal cancer), 13.1 µm (metastatic breast cancer).³⁰⁾ Contrary to this, it has been reported that the size of leukocytes may also fall in the same range as that of CTCs, hence size based enrichment may have limited efficacy in terms of discriminating target cells.³²⁾ In contrast to their normal cell counterparts/hematologic cells, cancer cells have been reported to differ cytomorphologically e.g., cancer cells exhibit a higher nuclear cytoplasmic ratio, variable size, abnormal nuclear DNA content etc.^33,34) Cytomorphological deformities in CTC has been correlated to a poor clinical outcome in colorectal, prostate and metastatic breast cancers.³⁵⁾ To overcome the limitation of size-based enrichment methods, additional approaches should be used to increase the probability/efficacy of CTC detection/capture.

In this study, we combined two approaches to discriminate cancer cells from hematologic/normal cells: 1) the detection of cancer cells based on their size and shape and 2) the recognition of cancer cells based on cytomorphological deformity (surface contour). Namely, a Jasco-Occhio particle size analyzer IF nano-200 was employed for the comprehensive measurement of various size and shape parameters of cells. For the physical characterization of cells, a suspension of cells was introduced into a flow cell equipped with a high definition CCD camera that captures images by using stroboscopic illumination. The captured images are automatically analyzed for shape and size parameters by an in-house PIA-Pro Particle Image Analysis Software. Using the aforementioned Injection Flow Particle Image Analyzer, we first determined the appropriate combination of size and shape parameters to discriminate various cancer cells (established cancer cell lines) from normal cells (red blood cells, white blood cells, splenocytes and lymph node cells). With circularity (shape parameter) and inner diameter (size parameter; maximum inscribed disc within a particle) as chosen parameters for analysis, we confirmed that a majority of cancer cell lines have greater inner diameters (≥10 µm) than those of normal cells. Furthermore, a PC software program “Cancer Cell Finder” was designed to differentiate cancer cells from white blood cells (inner diameter ≥10 µm) based on surface luminance (brightness/darkness). This software was validated on various established cancer cell lines and mouse derived lymph node cells and splenocytes. The detection of CTC- and cluster-like particles in blood samples from a cat and dog diagnosed with cancer was also carried out. The effect of the measurement using the particle size analyzer on cell viability, growth, and CTC marker expression was also assessed.

RESULTS AND DISCUSSION

Detection of Cultured Cancer Cells on Particle Size Analyzer (Jasco-Occhio Particle Size Analyzer, IF-200nano)

A JASCO-Occhio particle size analyzer IF-200nano from JASCO INTERNATIONAL Co., Ltd., Japan, was employed for the comprehensive size and shape characterization of cells. The Jasco-Occhio particle size analyzer, IF nano-200, consists of 1) a sample inlet jet nozzle 2) an injection pump, 3) a flow cell equipped with a high definition CCD camera, and 4) in-built PIA-Pro particle image analysis software for the analysis of the sample images (Fig. S1 in Supplementary Materials). The IF-200nano is a wet-dispersion particle image analyzer that can analyze cell suspensions that are spatially separated. Cells dispersed in a liquid (the recommended concentration of cells is ca. 2000 cells/mL) are introduced into a flow cell via the jet nozzle and a CCD camera captures sample images passing through the flow cell by using stroboscopic illumination. A threshold needs to be set for image analysis for background correction. The captured images are then extracted and morphological parameters of each particle are calculated using the PIA-Pro Particle Image Analysis Software which are presented in various user-selectable formats. Information related to characteristic shape parameters such as ISO solidity (ISO: International Organization for Stadardization), convexity, luminance, luminance RSD (RSD: relative standard deviation), ISO circularity, elongation, ISO aspect ratio, and ISO roundness, and size parameters such as ISO inner diameter, ISO area diameter, mean diameter, width, length and ISO max distance (Table S1 in Supplementary Materials) of a particle is displayed with the image of the individual particles for visual verification.

In this work, SP2/O myeloma cells were chosen as the first example, because the results of our previous experiments suggested that the average inner diameter of SP2/O cells is 11.5 µm, which is greater than that of mouse red blood cells (4.4 µm on average).¹²⁾ We first observed SP2/O cells spiked in mouse blood at different dilution ratios on the particle size analyzer (Figs. 1a–f). The images obtained from the particle size analyzer clearly showed an increase in the number of larger cells (SP2/O cells) at increasing ratios of SP2/O cells added to blood, confirming the possible application of the particle size analyzer for the size-based detection of cancer cells in blood. Furthermore, the size distribution of blood cells and SP2/O cells were visualized by number (horizontal axis as area diameter in µm and vertical axis as the number (%)) in Fig. 1g and by volume (horizontal axis as area diameter in µm and vertical axis as volume (%)) in Fig. 1h, indicated that the size distribution by volume exhibits a better sensitivity in detecting cancer cells than by number due to the fact that the radius of SP2/O cancer cells is larger than that of blood cells.

Fig. 1. Detection and Analysis of SP2/O Cancer Cells

SP2/O cells (indicated by circles in Fig. b–f) spiked in diluted C57BL/6 mouse blood (200×), at various ratios (by volume) of mouse blood to the SP2/O cell suspension (1×10⁶ cells/mL) (a) 1 : 0, (b) 1 : 0.125, (c) 1 : 0.25, (d) 1 : 0.5, (e) 1 : 1 and (f) 1 : 2 (v/v) were analyzed on the particle size analyzer. The size distribution of blood cells and SP2/O cells were analyzed by number (g) and volume (h). The bold curves in (g) and (h) are integration curves corresponding to the number (g) and volume (h) plotted against area diameter (µm) from 0.1 µm (left) to 1000 µm (right). Plot by volume (h) showed better sensitivity than plots by number (g).

Various combinations of sizes and shapes were tested in an attempt to detect a mixture of SP2/O cells in blood. Figure 2 displays a representative two-dimensional distribution of SP2/O cells (1×10⁶/mL) in mouse blood (the cell suspension in blood was diluted 51× with phosphate buffered saline (PBS) for the analysis to include 5.1×10⁹/mL red blood cells (RBC) and ca. 2×10⁴ SP2/O cells/mL) as scatterplots based on size and shape parameters (each small dot in the scatterplot corresponds to a cell (particle) as depicted in Fig. S1b in Supplementary Materials). For example, Figs. 2a–d are scatterplots of SP2/O cells spiked in mouse blood (by volume) with respect to circularity (%) (vertical axis) versus mean diameter (µm) (horizontal axis) (Fig. 2a), convexity (%) (Fig. 2b), luminance (%) (Fig. 2c), and max distance (µm) (Fig. 2d). Figures 2e–g displays distribution of SP2/O cells in blood by volume as scatterplots with respect to inner diameter (µm) (horizontal axis) versus area diameter (µm) (Fig. 2e), roundness (%) (Fig. 2f) and circularity (%) (vertical axis) (Fig. 2g).

Fig. 2. Representative Two-Dimensional Distribution of SP2/O Cells Spiked in Mouse Blood

SP2/O cells spiked in mouse blood (final cell concentration: 2×10⁴/mL) were analyzed on the particle size analyzer by different combinations of size and shape parameters to standardize suitable parameters for the detection of SP2/O cells in the presence of blood cells. The data are presented as scatterplots with variable parameters by volume. The scatterplot of SP2/O cells spiked in mouse blood, circularity (%) versus inner diameter (µm), confirms that SP2/O cells are larger (inner diameter >10 µm). In Figs. 2h and i, scatterplots of blood cells and SP2/O cells (2×10³ cells/mL of PBS) are presented (plotted by volume), respectively, confirming that the inner diameter of most of the blood cells are less than 10 µm, while that of cancer (SP2/O) cells is 11.6 µm (average).

In addition, Figs. 2h and i display the distribution of mouse blood (1000× dilution in PBS) and SP2/O cells (2×10³/mL of PBS) by volume with respect to inner diameter (µm) (x axis) and circularity (%) (y axis), respectively. Scatterplot of the blood cells confirmed that almost all blood cells are located in the left quadrant of the scatterplot with inner diameters of less than 10 µm, whereas SP2/O cells are mainly located in the right quadrant with inner diameters of greater than 10 µm. These results suggest that inner diameter and circularity are the optimal combination for detecting cancerous SP2/O cells spiked in mouse blood.

Based on the aforementioned results, 11 adherent and non-adherent cancer cell lines, LLC, B16, K562, HeLa, A549, Colon-26, Jurkat, Molt-4, MDA-MB-468, MDA-MB-157 and MC38 were analyzed on the particle size analyzer using a combination of inner diameter and circularity, as presented in the form of a scatterplot in Figs. 3a–k (displayed by volume), indicating that all of these cancer cell lines have majorly inner diameters of greater than 10 µm.^30,33) The values of these inner diameters are summarized in Table S2 in Supplementary Materials. In addition, images of the top 5 largest individual cells of these cancer cell lines obtained by the particle size analyzer are shown in Fig. S2 in Supplementary Materials with circularity (circ) and maximum diameter (Dimax) values. Images of the cells captured by particle size analyzer offer an advantage over other size-based detection system to identify a cell/clusters/debris.

Fig. 3. Scatterplots of Cancer and Normal Cells with Respect to Inner Diameter (µm) and Circularity (%) by Volume

Scatterplot of (a) LLC, (b) B16, (c) K562, (d) HeLa, (e) A549, (f) Colon-26, (g) Jurkat, (h) Molt-4, (i) MDA-MB-468, (j) MDA-MB-157, (k) MC38, (l) red blood cells, (m) platelets, (n) white blood cells, (o) lymph node cells, and (p) splenocytes (cell aggregates are shown in the inset).

For comparison of normal blood cells and cancer cell lines, the size and shape parameters of normal cells from blood, spleen and lymph nodes of mouse (C57BL/6) were examined. Total blood cells were segregated into platelets, peripheral blood mononuclear cells (PBMCs) and RBC by density gradient centrifugation and analyzed by the particle size analyzer.^36,37) Splenocytes deprived of RBC and lymph node cells were also analyzed to confirm the size of white blood cells. Scatterplots (circularity versus inner diameter) of the normal cells such as RBCs, platelets, white blood cells (WBCs), lymph node cells and splenocytes are shown in Figs. 3l–p, respectively. In addition, images of the largest 5 individual cells of the normal cells are shown in Fig. S3 in Supplementary Materials. Size distribution curves for cancer cells and normal cells are also shown in Figs. S4 and S5 in Supplementary Materials, respectively. The scatterplot and size distribution graph obtained from the particle size analyzer clearly show the size difference between normal cells and cancer cell lines and also confirm that a combination of “inner diameter” and “circularity” are appropriate parameters for further analyses to differentiate blood cells (inner diameter <10 µm) and cancer cells. Images of cancer cell lines and normal cells taken by the particle size analyzer are presented in Fig. S6 in Supplementary Materials.

Shape and size parameters for normal cells were compared with those of cancer cell lines such as adherent (LLC and B16) and non-adherent (Jurkat and Molt-4) type cells, as summarized in Fig. 4. The shape and size parameter values of all normal cells and cancer cell lines are listed in Table S2 of Supplementary Materials. Statistical analyses indicated that the size parameters, inner diameter, area diameter, mean diameter and width were all significantly higher in cancer cell lines compared to those of normal cells (Fig. 4a), while no significant difference was found in shape parameters between cancer cells and normal cells (Fig. 4b). For further analyses, we used the inner diameter as a typical parameter, because a combination of inner diameter and circularity was found to be appropriate for discriminating cancer cells from normal cells.

Fig. 4. Size and Shape Parameters of Normal and Cancer Cells Analyzed by the Particle Size Analyzer

Analysis of (a) size parameters of normal and cancer cells such as inner diameter, area diameter, mean diameter, width, length and max distance and b) shape parameters such as solidity, convexity, luminance, luminance RSD, circularity, elongation and roundness, respectively in arbitrary units (a.u.). The data are presented as the mean±S.E.M. with * p<0.05 considered as significant and “NS” as non-significant.

Development of a New Software Program “Cancer Cell Finder (CCF)” to Distinguish Cancer Cells from Normal Cells

It has been reported that the size of some WBCs is similar to that of cancer cells.^32,38) In our study, we observed one WBC having an inner diameter of 10.3 µm among 777 particles (inner diameters of 4–50 µm) (Fig. 3n). This information prompted us to develop a new computer software program to distinguish WBCs (inner diameter of >10.3 µm) from cancer cells using images obtained from the particle size analyzer. Deformity in cancer cells in contrast to normal cell is widely known and some reports have linked the deformity in cancer cells with their metastatic propensity.³⁹⁾ As shown in Fig. 5, the difference in surface deformity (roughness) between cancer cell line (SP2/O cells and LLC cells) and normal cells (WBC and lymph node cell) formed the basis of our new method for distinguishing cancer cells from normal cells. In Figs. 5a and b, WBC and lymph node cells exhibit bright area in the center and darker area on the edge and smoother surface, while darkness curves of SP2/O and LLC are more complicated (Figs. 5c, d). This difference could be attributed to the abnormally increased DNA content in cancer cells which might have affected cell luminance during image capture by the particle size analyzer.

Fig. 5. Software (“CCF ver. 1.0”) Designed to Discriminate Cancer Cells and WBCs Based on Brightness Diagram of Cell’s Interior

Particle size analyzer images of WBC (a), lymph node cell (b), SP2/O cancer cell (c) and LLC cell (d), were sectioned horizontally and vertically. The brightness functional curves of sections, which pass through the center of gravity of each cell (bold plain and bold dashed lines) analyzed by counting stationary point method is shown for WBC (e and i), lymph node cell (f and j), SP2/O cancer cell (g and k) and LLC cell (h and l) (1 pixel corresponds to 0.185 µm). Arrows indicate stationary points in the graphs. Counts (stationary points) are presented for WBCs (m), lymph node cells (n), SP2/O cancer cells (o) and LLC cells (p) (dashed lines indicate the appropriate threshold (11 SP) between cancer cells and normal cells).

Based on this observation, we decided to count the stationary point (SP) (SP means a point, at which the gradient of a darkness/brightness curve is zero.) for discriminating cancer cells from normal blood cells based on the following algorithm^40–42) (Fig. 5): 1) Prepared binarization (black and white) images of the microscopic images (or images from the particle size analyzer) of the cells with a given threshold. 2) Based on the obtained binarization images, cell areas were extracted and their center of gravity determined (Figs. 5a–d). 3) The defined two cross sections passing through the center of gravity of each cell (bold plain and bold dashed lines in Figs. 5a–d) and respective luminance (darkness) functional curves of sections were obtained (Figs. 5e–h and 5i–l). 4) The total number of SP were counted on the aforementioned two luminance curves (bold plain and bold dashed lines in Figs. 5a–d), based on which the cells were classified into two categories: i) a cell having SP number of less than or equal to 10 (≤10) is considered as a non-cancer cell, and ii) a cell having SP number greater than 10 (>10) is considered as a cancer cell. This PC software is denoted as “Cancer Cell Finder (CCF) ver. 1.0.”

The resolution (pixel density) of particle size analyzer’s camera is 0.185 µm/pixel, based on the maximum outer diameter that was evaluated using the following formula: outer diameter (µm)=maximum horizontal pixel×0.185. The outer diameters calculated for WBC, lymph node cell, SP2/O and LLC, are 13.1 µm (10.3 µm inner diameter), 9.4 µm (6.6 µm inner diameter), 19.8 µm (14.3 µm inner diameter) and 26.1 µm (23.0 µm inner diameter), respectively, are shown in Figs. 5a–d.

In Figs. 5m and n, normal cells, WBC and lymph node cells have total stationary points of lower than 10, whereas LLC and SP2/O cancer cells have stationary points of greater than 10 (Figs. 5o, p). The probabilities of “CCF ver. 1.0” for detecting normal cells such as WBC and lymph node cells was 83 and 100%, respectively, and for cancer cells, such as SP2/O cells and LLC cells, was 95 and 100%, respectively, as summarized in Table 1. Similarly, we analyzed various other cancer cell lines and evaluated their stationary points, and the results are shown in Table 1. Based on these results, it was concluded that this “CCF ver. 1.0” may be more useful in differentiating localized cancer-originated CTCs rather than lymphoma/leukemia cells from WBCs.

Table 1. Stationary Points Observed for Cancer and Non-cancer Cells as Analyzed by the CLE Value Point Method of “CCF ver. 1.0” (See Fig. 5)

		Total cells	The number of cells that have >10 CLE value points (SP)	Probability of detection (%)
Normal cells	WBC	42	7	83 (as normal cells)
	Lymph node cells	42	0	100 (as normal cells)
Cancer cells	SP2/O	42	40	95
	LLC	42	42	100
	A549	41	41	100
	B16	42	41	98
	Colon-26	36	36	100
	HeLa	35	30	86
	Jurkat	42	21	50
	K562	42	41	98
	MC38	40	40	100
	MDA-MB-157	42	42	100
	MDA-MB-468	40	27	68
	Molt-4	42	10	24

Reculture of Particle Size Analyzer Processed Cells

To check the effect of particle analysis on the viability of cancer cells, LLC, MDA-MB-157 and MDA-MB-468 cells were recultured (24 h) after analysis using the particle size analyzer. As shown in Fig. S7 in Supplementary Materials, after being processed through the particle analyzer, the cancer cells retained their adherent characteristics. The results of a dye exclusion assay of these cancer cell lines before and after particle analyzer measurements shows that LLC and MDA-MB-157 cells are sensitive to the analysis on the particle analyzer, while other cell lines Jurkat, Molt-4, and MDA-MB-468 cells are not (Fig. S8 in Supplementary Materials).

Further, the effect of this method on the proliferation of B16, LLC, MDA-MB-468, MDA-MB-157 and Jurkat cells was evaluated by means of a WST assay. After the particle size analyzer measurement, the cells were subjected to a proliferation rate evaluation for 24–45 h. The data suggest that there was a significant difference in proliferation between LLC, Jurkat and MDA-MB-468 cells (Figs. S9b–d in Supplementary Materials) with that of control cells but no significant difference was observed in the case of B16 and MDA-MB-157 cells (Figs. S9a and e in Supplementary Materials).

The effect of the particle size analyzer processing on the expression of CTC markers (CD44 and EpCAM) in MDA-MB-468 cell lines was also examined. After processing though the particle size analyzer, MDA-MB-468 cells were cultured in vitro for 4 d and the expression of CTC markers in the particle size analyzer processed and control cell samples was evaluated by immuno-blotting with CD44 and EpCAM (Fig. S10 in Supplementary Materials). Significant difference in the expression of CTC markers (CD44, p=0.975 and EpCAM, p=0.871) was not observed in MDA-MB-468 cells before and after the particle size analyzer measurements (Fig. S10b in Supplementary Materials), suggesting that the processing of cancer cells on the particle size analyzer negligibly affects the CTC characteristics of cancer cells, at least, at the expression level. The effect of particle size analyzer measurement on cells is summarized in Table S3 in Supplementary Materials.

Detection of Cancer Cells and Cancer Clusters Spiked in Blood

The detection of cancer cells spiked in blood was carried out using LLC cells expressing the enhanced green florescent protein (LLC-EGFP) for easy visualization on the microscope. Mixtures of LLC cells (1×10⁷/mL) in 0.9% saline (Fig. 6a) and in blood (estimated number of RBC: 0.7–1.3×10¹⁰/mL) (Fig. 6b) were observed on the particle size analyzer (final LLC cell concentration: 8×10⁴/mL after dilution (126×) in 0.9% saline). A typical example of a scatterplot of LLC-EGFP cells spiked in blood is presented in Fig. 6c (LLC cells are shown in the inset). Individual cells observed with inner diameters greater than 10 µm were subjected to a “CCF ver. 1.0” analysis to classify them to cancer or non-cancer cells based on surface roughness/brightness (Figs. 6d, 6e), as explained in Fig. 5. The “CCF ver. 1.0” analysis data for the cells confirmed that 8 out of 10 cells have over 11 IFP values and hence can be assumed to be cancer cells (indicated by “*” in Fig. 6e). In 166 pictures taken on particle size analyzer (LLC cells spiked in blood sample), ca. 1.1×10¹⁰ RBC/mL and ca. 1.6×10⁷ LLC cells/mL in blood were detected which were approximately similar to their initial numbers in the spiked samples prepared for these experiments.

Fig. 6. Analysis of Single Cells (LLC) and Clusters (LLC and B16 Cells) Spiked in Saline and Mouse Blood on Particle Size Analyzer and “CCF ver. 1.0”

Particle size analyzer images of (a) LLC cells spiked in 0.9% saline and (b) a mixture of LLC cells and mouse blood (LLC cells encircled). (c) Scatterplot of LLC cells spiked in mouse blood (LLC cells are shown in the inset). (d) Top 10 single cells (inner diameter) observed in figure (c). (e) Confirmation of cells (Fig. 6d) as cancer or normal cells via software analysis (cancer cell detection threshold >10, dashed line), “*” indicates cancer cell. Scatterplots of LLC (f) and B16 (g) clusters (insets depicting largest clusters), respectively. (h) and (i) LLC and B16 clusters spiked in mouse blood (insets depicting largest clusters), respectively.

Aceto et al. reported that CTC clusters possess a greater metastatic potential and a higher resistance to apoptosis than single CTCs, indicating the clinical relevance of CTC clusters in metastasis.^3,43) To examine the efficiency of detection of CTC clusters by our method, we prepared cancer cell line tumor spheres (clusters) in vitro. B16 and LLC-EGFP cells were cultured in a mixture of 0.6% methylcellulose and Dulbecco’s modified Eagle’s medium (DMEM), for 5–6 d and then harvested carefully for analysis by the particle size analyzer (images of clusters are shown in Fig. S11 of Supplementary Materials).^44,45) Scatterplots of LLC and B16 cell clusters (in saline) confirms that the clusters can also be detected by the particle size analyzer and their inner diameters are greater than 10 µm (Figs. 6f, h). Moreover, the detection of cancer cell clusters in blood was carried out. LLC-EGFP and B16 clusters were carefully spiked in C57BL/6 mouse blood and then diluted with PBS for analysis on the particle size analyzer. As shown in Figs. 6h and i, the analyzed clusters were located in the right quadrant of the scatterplot with inner diameter of greater than 10 µm. The data implies that cancer cell clusters can also be detected on the particle size analyzer.

We next carried out the filtration of LLC-EGFP and B16 clusters through 10 µm nylon mesh and the changes in numbers and size of these clusters were assessed with particle size analyzer before and after filtration to confirm the damage to clusters during particle size analyzer measurement. Figures S12a-1 and S12b-1 in Supplementary Materials show the scatterplot of LLC and B16 clusters (x-axis: inner diameter and y-axis: circularity) before filtration and Figs. S12a-2 and S12b-2 show those after filtration. Tables S12a-3 and S12b-3 in Supplementary Materials indicate that total number of particles was reduced and negligible clusters (>10 µm) were observed after filtration through 10 µm nylon mesh, implying that clusters and larger particles were filtered off and not broken down into small particles during measurement on the particle size analyzer. The largest 10 distribution particles before and after filtration are shown in Fig. S13 of Supplementary Materials, confirming the results mentioned above.

Detection of Tumor Cells in B16 Tumor Mouse Model

We next verified our approach for differentiating cancer cells in a mouse model, to which B16 cells had been injected, because B16 cancer cells possess high metastatic nature and its detection probability by “CCF ver. 1.0” is 98% (Table 1). Blood and splenocytes harvested from the tumor bearing mouse (13 d post inoculation of B16 cancer cells) were first analyzed by flow cytometry to check the presence of B16-EGFP cancer cells (Fig. S14 in Supplementary Materials). Because a greater number of tumor cells was detected by flow cytometry in the spleen sample than in the blood of the tumor bearing mouse, we measured splenocytes on the particle size analyzer (Fig. 7). Scatterplots of splenocytes derived from a normal mouse and a B16 tumor bearing mouse are shown in Figs. 7a and b, respectively. Cells 1–8 in Fig. 7c (splenocytes from normal cells) with inner diameters of >10 µm were subjected to “CCF ver. 1.0,” which confirmed that all of them had stationary point of <10 (Fig. 7e). In the tumor bearing mouse samples, 5 cells (No. 9–13) among 12 cells (No. 9–20) in Fig. 7d were judged as cancer cells and the rest as normal cells (No. 14–20 in Fig. 7d), which possibly are splenocytes or activated white blood cells (Fig. 7f). The features of the cells No. 9–13 are similar to those of in vitro cultured B16 cells, 21C–24C in Figs. 7d and f.

Fig. 7. Cancer Cell Detection in in Vivo B16 Tumor Xenograft Model

Scatterplot of splenocytes of (a) normal mouse, and (b) B16 tumor bearing mouse. Top single cells observed in scatterplot of (c) normal mouse (Nos. 1–8), and (d) B16 tumor bearing mouse (Nos. 9–20) and in vitro cultured B16 cancer cell lines (Nos. 21C–24C). The results of “CCF ver. 1.0” analysis for verification of cells presented in (c) and (d) as normal/cancer cells are summarized in (e) normal mouse (Nos. 1–8), and (f) B16 tumor bearing mouse (Nos. 9–20) and in vitro cultured B16 cancer cell lines (Nos. 21C–24C) (cancer cell detection threshold >10 is indicated with a dashed line), “*” indicates cancer cell.

We do not deny the possibility that the lymphocytes will be enlarged upon stimulation by cytokines, but based on our aforementioned data we believe that it is possible to verify larger lymphocytes as normal cells using our software “CCF.”

Detection of CTC Cluster-Like Particles in the Blood of Cat and Dog Diagnosed with Cancer

As a clinical application of the particle size analyzer for detecting CTC clusters in the whole blood of cancer patients, blood samples obtained from normal and cancer diagnosed cats and dogs were analyzed in a similar manner to the aforementioned method. As shown in Figs. 8a and e, scatterplots (y axis-circularity (%) and x axis-inner diameter (µm)) of normal dog (sample ID: TU-0001-D) and cat blood (sample ID: TU-0003-C) showed negligible particles (cell/clusters) greater than 10 µm, respectively. On the other hand, particle size analyzer detected 2 larger cancer cluster-like particles in the blood sample of the dog with osteosarcoma (sample ID: TU-0002-D) (Fig. 8b) and 1 cluster in the blood sample of a cat diagnosed with a mammary adenocarcinoma (sample ID: TU-0004-C), Fig. 8d (debris are marked as 2 in sample TU-0004-C). Top 10 individual cells based on inner diameter were subjected to “CCF ver. 1.0” analysis to categorize them into cancer/non-cancer cells (Figs. 8c, d). These data confirmed that one cell (>10 stationary points) out of 2 cells having inner diameter of >10 µm (No. 1, 2) in the dog blood sample (sample ID: TU-0002-D) appeared to be a cancer cell and other cells having inner diameter of <10 µm were judged to be normal cells (Figs. 8c, d). For cat blood samples (Figs. 8e, f), one CTC-like cluster particle was similarly judged as a cancer cell, whereas cells with inner diameter of <10 µm were judged as normal cells (Figs. 8g, h). More clinical information is provided in Table S4 of Supplementary Materials. These data have allowed us to conclude that a combination of particle size analyzer and “CCF ver. 1.0” is also capable of detecting CTC cluster-like particles in blood samples of cancer-bearing animals.

Fig. 8. Detection of CTC Cluster-Like Particles in the Blood of Cat and Dog Diagnosed with Cancer

Scatterplots of (a) normal dog (TU-0001-D), (b) dog suffering from cancer (TU-0002-D) Images of the clusters detected by the particle size analyzer are shown as insets in the scatterplots. (c) Top single cells/particles observed in scatterplot of cancer bearing dog (b) and “CCF ver. 1.0” analysis of cells/cells/particles are shown in (d). Scatterplots of (e) normal cat (TU-0003-C), and (f) cat suffering from cancer (TU-0004-C). Images of the clusters detected by the particle size analyzer are shown as insets in the scatterplots. (g) Top single cells/particles observed in scatterplot of cancer bearing cat (f) and “CCF ver. 1.0” analysis of cells/particles are shown in (h) (cancer cell detection threshold >10, dashed line), “*” indicates cancer cell/cluster.

Detection of Cancer Cells in Normal Human Blood Cells

We conducted measurements of human lung adenocarcinoma A549 cells spiked in human blood. The size of A549 cells in human whole blood (diluted in saline) was measured by the particle size analyzer (Fig. 9a), in which only a few large normal cells (inner diameter >10 µm) were observed. A typical scatterplot of a blood sample with large cells (inner diameter=10.2 µm) is shown in Fig. 9a and a “CCF ver. 1.0” analysis of the top 7 largest cells (Fig. 9b) indicated that their SPs are <10 (Fig. 9c), implying that these are normal cells. Next, white blood cells isolated from human blood were measured on the particle size analyzer (n=6) and “CCF ver. 1.0.” As shown in Figs. 9e and f, all of the 8 cells picked up from Fig. 9d were judged as normal cells (SPs are <10), although some of them had an inner diameter of >10 µm. Additional information is provided in Figs. S15a and S15b in Supplementary Materials.

Fig. 9. Detection of Human Lung Adenocarcinoma A549 Cells Spiked in Human Blood by the Particle Size Analyzer and “CCF ver. 1.0”

Analysis of human whole blood (a) typical scatterplot of human whole blood, (b) top single cells observed in the scatterplot (1–7), and (c) “CCF ver. 1.0” analysis of cells. Analysis of human white blood cells, (d) typical scatterplot of human white blood cells, (e) single cells observed in the scatterplot (1–8), and (f) “CCF ver. 1.0” analysis of cells. Analysis of A549 cells (g) typical scatterplot of A549 cells, (h) top single cells observed in the scatterplot (1–8), and (i) “CCF ver. 1.0” analysis of cells. Analysis of A549 cells and human white blood cells (j) typical scatterplot of a spiked sample, (k) single cells observed in the scatterplot (1–10), and (l) “CCF ver. 1.0” analysis of cells. Analysis of A549 cells and human whole blood (m) typical scatterplot of spiked sample, (n) single cells observed in the scatterplot (1–8), and (o) “CCF ver. 1.0” analysis of cells. Cancer cell detection threshold is >10, indicated as dashed line, “*” indicates cancer cells.

In Figs. 9g–i, human A549 cancer cells were analyzed and the largest 8 cells were confirmed as cancer cells. We next analyzed a mixture of human white blood cells and A549 cells at the same concentration (Figs. 9j–l). In this case, cell No. 1–5 were judged as cancer cells among 10 cells having inner diameter of 6.8–25 µm (Figs. 9k, l). Pictures of normal white blood cells and A549 cells taken on particle size analyzer in shown in Figs. S16 in Supplementary Materials.

These results prompted us to attempt to detect A549 cells (1×10⁵) spiked in human whole blood (6.3×10⁶ normal cells) (diluted) on the particle size analyzer. As shown in Figs. 9m and n, cell No. 1–5 among 8 cells were judged as cancer cells (Fig. 9o). Furthermore, the ability of the particle size analyzer to detect cancer cells in spiked sample was verified by an analysis of larger cells (A549 cells) in the absence and presence of blood cells. Figure 9m presents a scatterplot of A549 cells (in blood) at the same concentration as those analyzed in spiked saline (Fig. S15c in Supplementary Materials). The total number of A549 cells (inner diameter >10 µm) spiked in blood sample (n=11) was 6.5±1.0×10⁴ large cells/3.2×10⁶ blood cells. For comparison, the number of A549 cells spiked in saline was determined to be 6.0±2.0×10⁴ A549 cells/mL, which is almost same number as that of the aforementioned numbers in blood (Fig. S15d in Supplementary Materials), suggesting that human cancer cells can be detected in human blood without substantial loss.

Finally, we checked the morphology of human umbilical vein endothelial cell (HUVEC), which are a type of normal cells. As shown in Fig. S17a in Supplementary Materials, the inner diameter of HUVEC are 4–50 µm and some large cells possessing peculiar shapes and different circularity values were unexpectedly judged to be cancer cells by “CCF ver. 1.0” (Figs. S17b, c in Supplementary Materials). The reason for this finding is unclear but may be attributed to the harvesting methods (trypsinization) employed to detach the adherent HUVEC. More detailed investigations need to be done to analyze adherent normal cells.

CONCLUSION

In this study, we employed a Jasco-Occhio particle size analyzer IF nano-200 for the size- and size-based detection of cancer cells and clusters. Sample measurement using the particle size analyzer is more convenient than previously reported methods. It permits particles/cells to be comprehensively characterized in a short time (ca. 5 min for 1 mL of sample). The automated analysis data can be displayed in various user-selectable formats such as size and shape distribution curves and numerical values. Debris, single cell or clusters in a scatterplot can be verified visually by images of individual particles, thereby ruling out the possibility of generating any negative data. The test samples do not require any major preprocessing (e.g. cell fixation) and can be operated with any diluent, thereby permitting the viability of the target cells to be maintained for in vitro establishment and for further characterization, contrary to immunological based detection which requires cells to be fixed.

It should be noted that the definition of the detection limit (sensitivity) of CTC in our method is somewhat different from that of previous methods. Namely, the detection limit of our method is dependent on the volume of blood sample analyzed and the number of images acquired by the particle size analyzer, and hence, it is dependent on the amount of memory and the processing speed of the computer (the greater is the memory and processing speed, the higher the sensitivity is). In our current experiments (particle size analyzer used in this work), the detection limit was ca. 2000 CTCs/mL of blood (possibly, containing 1.1×10⁶ RBCs). Because the cancer cell numbers used in these cell spiking experiments are greater than those in actual clinical samples, the limit of detection of the CTC will need to be improved.

A convenient software program “CCF ver. 1.0” was developed to distinguish cancer cells from normal cells based on differences in the luminance (darkness) extent of curve of the cell surface. It is suitable for differentiating adherent cancer cells from WBCs and splenocytes, thus making it attractive for use in hospitals, medical facilities and laboratories, thereby overcoming the need for expensive CTC detecting devices. We do not deny the possibility that the size and shape of CTC and normal cells are altered by anticancer agents, cytokines such as transforming growth factor-beta (TGF-β) or other effects. Further studies will need to be done to examine these issues and will be reported elsewhere. Over all, an analysis of blood samples (diagnosed with cancer) with a particle size analyzer or more simple and cheaper instruments assisted by a cell-recognition software such as “CCF” provides insights into possible clinical applications of our combination approach for detection CTCs and CTC clusters for evaluating the progression and theranostics of cancer and related diseases.

MATERIALS AND METHODS

Reagents

DMEM, RPMI 1640 medium, fetal calf serum (FCS), streptomycin sulphate, benzypenicillin potassium, sodium chloride, potassium chloride, tris base, glycine, sodium phosphate dibasic, sodium dihydrogen phosphate were purchased from Wako Pure Chemical Industries, Ltd., Osaka, Japan. Trypsin-ethylenediaminetetraacetic acid (EDTA) was procured from Gibco, Japan.

Cell Culture

The cell lines, LLC-EGFP (mouse; lewis lung carcinoma expressing green fluorescent protein), B16 (mouse; melanoma), Jurkat (human; T cell lymphoma), Molt-4 (human; T cell lymphoma), MC38 (mouse; colon cancer), K562 (human; chronic myelogenous leukemia), HeLa (human; uterine cancer), A549 (human; lung adenocarcinoma), Colon-26 (mouse; colorectal cancer), and SP2/O (mouse; melanoma) were provided by Prof. Ryo Abe (Tokyo University of Science), MDA-MB-468 (human; breast cancer) and MDA-MB-157 (human; breast cancer) were purchased from ATC C and HUVEC were purchased from RIKEN cell bank (Japan). Carcinoma and normal cells were maintained at 37°C in a humidified 5% CO₂/95% O₂ in their respective media supplemented with 10% FCS, L-glutamine and antibiotic mixture (penicillin; 69.9 ng/mL and streptomycin; 139.3 ng/mL). LLC-EGFP, B16, MDA-MB-468, MDA-MB-157, A549 and MC38 were maintained in DMEM, Jurkat, Molt-4 and SP2/O cell lines were maintained in RPMI 1640 medium supplemented with HEPES (2-[4-(2-hydroxyethyl)-1-piperazinyl]ethanesulfonic acid) and mono-thioglycerol (0.5 mM/L). Colon-26 cells and HeLa cells were maintained in RPMI 1640 and MEM medium, respectively. HUVEC were maintained in medium 200 supplemented with low serum growth supplement (Gibco).

Analysis in Particle Size Analyzer

Adherent and non-adherent cell lines were cultured to 80–90% confluency for analysis. Cells were harvested and pelleted down by centrifugation at 1400 rpm for 7 min at 4°C. Obtained pellets were suspended in a sterile PBS at pH 7.4 for measurement on particle size analyzer. Prior to cell measurement, flow chamber of the particle size analyzer was washed with 5–10 mL of double-distilled water (dd H₂O), followed by background and light calibration, and subsequently analysis of cells (1–2 mL aliquot of the cell suspension solution) with the following parameters: light intensity (9), volume sampling (0.5%), volume analyzed (1–2 mL), magnification (4.5×), camera pixel (0.185) and syringe diameter (6515 µm).

Standardization of Shape and Size Parameters

For analysis of only blood, the blood was diluted 1000× with PBS and analyzed. To standardize suitable parameters for the detection of cancer cells in blood cells, SP2/O cells (1×10⁶/mL) were spiked in C57BL/6 mouse blood. A 100 µL aliquot of this sample was diluted with PBS (50×) (final cell concentration: ca. 2×10⁴ SP2/O cells/mL) and analyzed on the particle size analyzer. Various shape and size parameters combination were standardized to identify suitable parameters for the detection of cancer cells in the presence of normal cells.

Cell Cluster Preparation

LLC-EGFP and B16 cells were seeded (1.5×10³ cells/well) in 0.6% sterile methylcellulose/DMEM (500 µL/well) in a 24-well dish (Ultra low attachment plate, Costar) and incubated for 5–6 d. Post 3 d of seeding, 200–300 µL of media was added. After incubation, the clusters were imaged using a microscope (Biorevo, BZ-9000, Keyence) at 40× and 200× magnifications. For particle size analyzer measurements, the clusters were collected (from 3 wells, ca. 250–300 clusters) and pelleted down by centrifugation at 1400 rpm for 6 min at 4°C and suspended in 4 mL sterile PBS (pH 7.4). The clusters were kept on ice until used.

Cancer Cell and Cluster Detection in Particle Size Analyzer

LLC-EGFP cells (1×10⁷) were spiked in 1 mL of mouse blood (C57BL/6) and in saline, and then further diluted (126×) with 0.9% saline for measurement on the particle size analyzer (final cell concentration: ca. 8×10⁴ cells/mL). For the analysis of individual cluster analysis, 1 mL of cluster suspension in PBS was analyzed (see cluster preparation section for details). After the analysis, the suspension was filtered through 10 µm nylon mesh and the filtrate was analyzed again on particle size analyzer. The filtered clusters on the mesh were visualized by a microscope (Biorevo BZ-9000, Keyence) at a 200× magnification. For the detection of clusters in blood, clusters were harvested (from 3 wells of a 24-well plate), centrifuged at 1400 rpm at 4°C for 6 min and the resulting pellet was carefully suspended in 50 µL of C57BL/6 healthy mouse blood. The spiked mixture was diluted with PBS (126× for LLC-EGFP clusters and 251× for B16-EGFP clusters) and analyzed on the particle size analyzer (2 mL).

Normal Cells Isolation and Detection

Prior to the analysis of cancer cell lines, we examined the shape and size parameters of normal blood cells (RBC, WBC, and platelets), splenocytes and lymph node cells that were harvested from 4–5 months old C57BL/6 mice (SANKYO LABO SERVICE Co. Inc., Shizuoka, Japan). To harvest the RBCs, WBCs and platelets from blood, blood was obtained through cardiac puncture (500–1000 µL) from anesthetized C57BL/6 mice (isoflurane) and collected in a vial containing heparin (1000 U/mL). The blood sample was diluted with 2 mL of wash buffer (PBS supplemented with 0.5% calf serum (CS) and azide) followed by the addition of 1 mL of lymphocyte separate media (Immuno-Biological Laboratories Co., Ltd., Gunma, Japan) to the base of the vial and the resulting suspension was centrifuged at 1400 rpm at room temperature (r.t.) for 30 min. After centrifugation, the sample solution in the vial was separated into 3 different layers. 1) The top transparent layer (platelets) was diluted in 2–5 mL RPMI (0.5% CS) and analyzed using the particle size analyzer. 2) The middle buff-colored white layer corresponding to peripheral blood mononuclear cells (PBMCs) was collected in new vial with the addition of 2 mL of wash buffer and centrifuged at 1400 rpm at 4°C for 8 min. 3) The pellet (RBC) was suspended in 400–500 µL of RPMI (10% CS) and filtered through a 70 µm mesh before use. All 3 separated cell suspensions were maintained on ice until used. For analysis using the particle size analyzer, the cell suspension was diluted with PBS and around 1 mL was used in the analysis. The lowest RBC layer was diluted with PBS (100×) and 1 mL was analyzed on particle size analyzer.

For the detection of lymph node cells, axillary and inguinal lymph nodes were harvested from C57BL/6 mice and suspended in 10 mL of cold RPMI media (0.5% CS). The cells were squeezed out through the nodes by sterile forceps and filtered through a 70 µm mesh followed by centrifugation at 1400 rpm for 8 min at 4°C. The pellet was suspended in RPMI (10% CS) and filtered again through 70 µm mesh to remove cell aggregates. The resulting filtrate (1 mL) was analyzed on particle size analyzer.

To isolate splenocytes, spleens from C57BL/6 mice were harvested, minced and squeezed with sterile glass slides. The isolated cells were suspended in 25 mL of cold RPMI media (0.5% CS) and centrifuged at 1400 rpm for 8 min at 4°C. The resulting pellet was treated with 1 mL of ACK (Ammonium-Chloride-Potassium) lysing buffer for 1 min at r.t. to lyse the RBCs. The reaction was terminated by the addition of 40 mL of RPMI media (0.5% CS) followed by filtration through 70 µm mesh. The suspension was then centrifuged at 1400 rpm at 4°C for 8 min and the resulting pellet was suspended in 10 mL RPMI media (0.5% CS). The resulting mixture was filtered through 70 µm mesh, 1–2 mL of which was analyzed using the particle size analyzer. All animal experiments were conducted in accordance with the guidelines of Tokyo University of Science (approved animal experiment number S15005).

B16-EGFP Tumor Mouse Model

All animal experiments were conducted in accordance with the guidelines of Tokyo University of Science (approved animal experiment number: S15005). To test the applicability of our method for cancer cell detection, a B16-EGFP tumor mouse model was used. Briefly, 4×10⁶ B16-EGFP cells/200 µL of PBS was injected by tail vein in 4–5 months old C57BL/6 mice. Mice were sacrificed 13 d post cell inoculation and spleen and peripheral blood (by cardiac puncture) were harvested. Individual cells were collected from the samples as mentioned in the above section and analyzed in particle size analyzer and flow cytometry.

Detection of CTC-Like Clusters in Cat and Dog Blood Samples

Cat and dog blood samples were obtained from Tottori University, Japan, with clinical information. For particle size analysis, the blood samples were diluted (500×) in PBS (pH 7.4), 5 mL of which was used for the analysis on the particle size analyzer. All animal experiments were conducted in accordance with the guidelines of Tottori University.

Detection of Cancer Cells in Human Blood Samples

Human whole blood serotype A from a single donor was purchased from BioPrendic International (France). For the analysis of human samples, the detection of A549 cells spiked in human blood was carried out. Whole blood was diluted 100× with saline before measurement on the particle size analyzer. White blood cells (lymphocytes, monocytes) from human blood were isolated using lymphocyte separation medium (MP Biomedicals, France) following the manufacturer’s instructions measured on particle size analyzer suspended in saline. A549 cells at a concentration of ca. 8×10⁴/mL in saline was measured on particle size analyzer. A mixture of white blood cells (1×10⁶/mL) and A549 cells (1×10⁶/mL) in saline was measured on the particle size analyzer. Human blood sample was diluted 50× in saline before spiking with A549 cells. HUVEC, harvested by trypsinization, were suspended in saline and analyzed on the particle size analyzer (7×10⁴/mL).

Statistical Analysis

The quantitative data are representative of three or more independent experiments and expressed as the mean±standard error of the mean (S.E.M.). Statistical analysis was performed using ANOVA (one way ANOVA) followed by Bonferroni’s test to determine significant differences in mean and p<0.05 was considered as statistically significant.

Acknowledgments

This work was supported by Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan (Nos. 19659026, 22390005, and 24659085 for S.A.) (https://www.jsps.go.jp/english/e-grants/), High-Tech Research Center Project for Private Universities (matching fund subsidy from MEXT), and Tokyo University of Science (TUS) fund for strategic research areas for S.A.

Conflict of Interest

The authors declare no conflict of interest.

Supplementary Materials

The online version of this article contains supplementary materials

REFERENCES

• 1) Spano D, Heck C, De Antonellis P, Christofori G, Zollo M. Molecular networks that regulate cancer metastasis. Semin. Cancer Biol., 22, 234–249 (2012).
• 2) Mehlen P, Puisieux A. Metastasis: a question of life or death. Nat. Rev. Cancer, 6, 449–458 (2006).
• 3) Aceto N, Bardia A, Miyamoto DT, Donaldson MC, Wittner BS, Spencer JA, Min Y, Pely A, Engstrom A, Zhu H, Brannigan BW, Kapur R, Stott SL, Shioda T, Ramaswamy S, Ting DT, Lin CP, Toner M, Haber DA, Maheswaran S. Circulating tumor cell clusters are oligoclonal precursors of breast cancer metastasis. Cell, 158, 1110–1122 (2014).
• 4) Paoletti C, Li Y, Muñiz MC, Kidwell KM, Aung K, Thomas DG, Brown ME, Abramson VG, Irvin WJ Jr, Lin NU, Liu MC, Nanda P, Nangia JR, Storniolo AM, Traina AT, Vaklavas C, Poznak CHV, Wolff AC, Forero-Torres A, Hayes DF. Significance of circulating tumor cells in metastatic triple negative breast cancer patients within a randomized, phase II trial: TBCRC 019. Clin. Cancer Res., 21, 2771–2779 (2015).
• 5) Fidler IJ. Tumor heterogeneity and the biology of cancer invasion and metastasis. Cancer Res., 38, 2651–2660 (1978).
• 6) Cohen SJ, Punt CJA, Iannotti N, Saidman BH, Sabbath KD, Gabrail NY, Picus J, Morse M, Mitchell E, Miller MC, Doyle GV, Tissing H, Terstappen LW, Meropol NJ. Relationship of circulating tumor cells to tumor response, progression-free survival, and overall survival in patients with metastatic colorectal cancer. J. Clin. Oncol., 26, 3213–3221 (2008).
• 7) Bono De JS, Scher HI, Montgomery RB, Parker C, Miller MC, Tissing H, Doyle GV, Terstappen LW, Pienta KJ, Raghavan D. Circulating tumor cells predict survival benefit from treatment in metastatic castration-resistant prostate cancer. Clin. Cancer Res., 14, 6302–6309 (2008).
• 8) Lan C, Heindl A, Huang X, Xi S, Banerjee S, Liu J, Yuan Y. Quantitative histology analysis of the ovarian tumour microenvironment. Sci. Rep., 5, 16317 (2015).
• 9) Yu M, Stott S, Toner M, Maheswaran S, Haber DA. Circulating tumor cells: approaches to isolation and characterization. J. Cell Biol., 192, 373–382 (2011).
• 10) Punnoose EA, Atwal S, Liu W, Raja R, Fine BM, Hughes BG, Hicks RJ, Hampton GM, Amler LC, Pirzkall A, Lackner MR. Evaluation of circulating tumor cells and circulating tumor DNA in non-small cell lung cancer: association with clinical endpoints in a phase II clinical trial of pertuzumab and erlotinib. Clin. Cancer Res., 18, 2391–2401 (2012).
• 11) Hou JM, Greystoke A, Lancashire L, Cummings J, Ward T, Board R, Amir E, Hughes S, Krebs M, Hughes A, Ranson A, Lorigan P, Dive C, Blackhall FH. Evaluation of circulating tumor cells and serological cell death biomarkers in small cell lung cancer patients undergoing chemotherapy. Am. J. Pathol., 175, 808–816 (2009).
• 12) Ariyasu S, Hanaya K, Watanabe E, Suzuki T, Horie K, Hayase M, Abe R, Aoki S. Selective capture and collection of live target cells using a photoreactive silicon wafer device modified with antibodies via a photocleavable linker. Langmuir, 28, 13118–13126 (2012).
• 13) Lv S, Wang J, Xie M, Lu N, Li Z, Cai X, Cai S, Zhang P, Dong P, Huang W. Photoresponsive immunomagnetic nanocarrier for capture and release of rare circulating tumor cells. Chem. Sci., 6, 6432–6438 (2015).
• 14) Vona G, Sabile A, Louha M, Sitruk V, Romana S, Schütze K, Capron F, Franco D, Pazzagli M, Vekemans M, Lacour B, Bréchot C, Bréchot PP. Isolation by size of epithelial tumor cells: a new method for the immunomorphological and molecular characterization of circulatingtumor cells. Am. J. Pathol., 156, 57–63 (2010).
• 15) Riahi R, Gogoi P, Sepehri S, Zhou Y, Handique K, Godsey J, Wang Y. A novel microchannel-based device to capture and analyze circulating tumor cells (CTCs) of breast cancer. Int. J. Oncol., 44, 1870–1878 (2014).
• 16) Stott SL, Hsu CH, Tsukrov DI, Yu M, Miyamoto DT, Waltman BA, Rothenberg SM, Shah AM, Smas ME, Korir GK, Floyd FP, Gilmand AJ Jr, Lord JB, Winokur D, Springer S, Irimia D, Nagrath S, Sequist LV, Lee RJ, Isselbacher KJ, Maheswaran S, Haberd DA, Toner M. Isolation of circulating tumor cells using a microvortex-generating herringbone-chip. Proc. Natl. Acad. Sci. U.S.A., 107, 18392–18397 (2010).
• 17) Okano H, Konishi T, Suzuki T, Suzuki T, Ariyasu S, Aoki A, Abe R, Hayase M. Enrichment of circulating tumor cells in tumor-bearing mouse blood by a deterministic lateral displacement microfluidic device. Biomed. Microdevices, 17, 9964 (2015).
• 18) Krebs MG, Metcalf RL, Carter L, Brady G, Blackhall FH, Dive C. Molecular analysis of circulating tumour cells-biology and biomarkers. Nat. Rev. Clin. Oncol., 11, 129–144 (2014).
• 19) Beije N, Jager A, Sleijfer S. Circulating tumor cell enumeration by the cell search system: the clinician’s guide to breast cancer treatment? Cancer Treat. Rev., 41, 144–150 (2015).
• 20) Rossi E, Basso U, Celadin R, Zilio F, Pucciarelli S, Aieta M, Barile C, Sava T, Bonciarelli G, Tumolo S, Ghiotto C, Magro C, Jirillo A, Indraccolo S, Amadori A, Zamarchi R. M30 neoepitope expression in epithelial cancer: quantification of apoptosis in circulating tumor cells by CellSearch analysis. Clin. Cancer Res., 16, 5233–5243 (2010).
• 21) Riethdorf S, Fritsche H, Muller V, Rau T, Schindlbeck C, Rack B, Janni W, Coith C, Beck K, Janicke F, Jackson S, Gornet T, Cristofanilli M, Pantel K. Detection of circulating tumor cells in peripheral blood of patients with metastatic breast cancer: a validation study of the CellSearch system. Clin. Cancer Res., 13, 928 (2007).
• 22) Gorges T, Tinhofer I, Drosch M, Röse L, Zollner T, Krahn T, Ahsen OV. Circulating tumour cells escape from EpCAM-based detection due to epithelial-to-mesenchymal transition. BMC Cancer, 12, 178 (2012).
• 23) Lustberg MB, Balasubramanian P, Miller B, Garcia-Villa A, Deighan C, Wu Y, Carothers S, Berger M, Ramaswamy B, Macrae ER, Wesolowski R, Layman RM, Mrozek E, Pan X, Summers TA, Shapiro CL, Chalmers JJ. Heterogeneous atypical cell populations are present in blood of metastatic breast cancer patients. Breast Cancer Res., 16, R23 (2014).
• 24) Yu M, Bardia A, Wittner BS, Stott SL, Smas ME, Ting DT, Isakoff SJ, Ciciliano JC, Wells MN, Shah AM, Concannon KF, Donaldson MC, Sequist LV, Brachte E, Sgroi D, Baselga J, Ramaswamy S, Toner M, Haber DA, Maheswaran S. Circulating breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition. Science, 339, 580–584 (2013).
• 25) Jin C, McFaul SM, Duffy SP, Deng X, Tavassoli P, Black PC, Ma H. Technologies for label-free separation of circulating tumor cells: from historical foundations to recent developments. Lab Chip, 14, 32–44 (2014).
• 26) Hong B, Zu Y. Detecting circulating tumor cells: current challenges and new trends. Theranostics, 3, 377–394 (2013).
• 27) Lin HK, Zheng S, Williams AJ, Balic M, Groshen S, Scher HI, Fleisher M, Stadler W, Datar RH, Tai YC, Cote RJ. Portable filter-based microdevice for detection and characterization of circulating tumor cells. Clin. Cancer Res., 16, 5011–5018 (2010).
• 28) Mohamed H, Murray M, Turner JN, Caggana M. Isolation of tumor cells using size and deformation. J. Chromatogr. A, 1216, 8289–8295 (2009).
• 29) Bhagat AA, Hou HW, Li LD, Lim CT, Han J. Pinched flow coupled shear-modulated inertial microfluidics for high-throughput rare blood cell separation. Lab Chip, 11, 1870–1878 (2011).
• 30) Hosokawa M, Kenmotsu H, Koh Y, Yoshino T, Yoshikawa T, Naito T, Takahashi T, Murakami H, Nakamura Y, Tsuya A, Shukuya T, Ono A, Akamatsu H, Watanabe R, Ono S, Mori K, Kanbara H, Yamaguchi K, Tanaka T, Matsunaga T, Yamamoto N. Size-based isolation of circulating tumor cells in lung cancer patients using a microcavity array system. PLOS ONE, 8, e67466 (2013).
• 31) Farace F, Massard C, Vimond N, Drusch F, Jacques N, Billiot F, Laplanche A, Chauchereau A, Lacroix L, Planchard D, Le Moulec S, André F, Fizazi K, Soria JC, Vielh P. A direct comparison of CellSearch and ISET for circulating tumour-cell detection in patients with metastatic carcinomas. Br. J. Cancer, 105, 847–853 (2011).
• 32) Bagnall SJ, Byun S, Begum S, Miyamoto DT, Hecht VC, Maheswaran S, Stott SL, Toner M, Hynes RO, Manalis SR. Deformability of tumor cells versus blood cells. Sci. Rep., 5, 18542 (2015).
• 33) Park S, Ang RR, Duffy SP, Bazov J, Chi KN, Black PC, Ma H. Morphological differences between circulating tumor cells from prostate cancer patients and cultured prostate cancer cells. PLOS ONE, 9, e85264 (2014).
• 34) Zink D, Fischer AH, Nickerson JA. Nuclear structure in cancer cells. Nat. Rev. Cancer, 4, 677–687 (2004).
• 35) Ligthart ST, Coumans FAW, Bidard FC, Simkens LHJ, Punt CJA, Groot MRD, Attard G, Bono JSD, Pierga JY, Terstappen LWMM. Circulating tumor cells count and morphological features in breast, colorectal and prostate cancer. PLOS ONE, 8, e67148 (2013).
• 36) Noble PB, Cutts JH. Separation of blood leukocytes by Ficoll gradient. Can. Vet. J., 8, 110–111 (1967).
• 37) Böyum A, Scand J. Isolation of mononuclear cells and granulocytes from human blood. Scand. J. Clin. Lab. Invest. Suppl., 97, 77–89 (1968).
• 38) Pødenphant M, Ashley N, Koprowska K, Mir KU, Zalkovskij M, Bilenberg B, Bodmer W, Kristensen A, Marie R. Separation of cancer cells from white blood cells by pinched flow fractionation. Lab Chip, 15, 4598–4606 (2015).
• 39) Zhang W, Kai Z, Ueno NT, Qin L. A brief review of the biophysical hallmarks of metastatic cancer cells. Cancer Hallm., 1, 59–66 (2013).
• 40) Ejiri M, Okutomi M, Okabe T, Ozawa S, Goto M, Saito H, Sato Y, Shimizu M, Tanaka M, Ninomiya T, Fujiyoshi H, Hori O, Mita T, Yamauchi Y. Digital Image Processing. (Okutomi M ed.) CG-ARTS Society, Tokyo, pp. 190–191 (2015).
• 41) Gonzalez RC, Woods RE. Digital Image Processing, 2nd ed. (Horton MJ ed.) Prentice Hall, Inc., New Jersey, pp. 672–675 (2002).
• 42) Yoshida T, Sugiura Y, Aikawa N. A closed-form design of linear phase FIR band-pass maximally flat digital differentiators with an arbitrary center frequency. IEICE Trans., 12, 2611–2617 (2014).
• 43) Au SH, Storey BD, Moore JC, Tang Q, Cheng YL, Javaid S, Sarioglua AF, Sullivand R, Maddend MW, Keefed RO, Haberd DA, Maheswaran S, Langenaud D, Stotta SL, Toner M. Clusters of circulating tumor cells traverse capillary-sized vessels. Proc. Natl. Acad. Sci. U.S.A., 113, 4947–4952 (2016).
• 44) Dontu G, Abdallah WM, Foley JM, Jackson KW, Clarke MK, Kawamura MJ, Wicha SM. In vitro propagation and transcriptional profiling of human mammary stem/progenitor cells. Genes Dev., 17, 1253–1270 (2003).
• 45) Fillmore CM, Kuperwasser C. Human breast cancer cell lines contain stem-like cells that self-renew, give rise to phenotypically diverse progeny and survive chemotherapy. Breast Cancer Res., 10, R25 (2008).