Greatly Enhanced CTC Culture Enabled by Capturing CTC Heterogeneity Using a PEGylated PDMS-Titanium-Gold Electro-Microfluidic Device with Glutathione-Controlled Gentle Cell Release

Abstract

The circulating tumor cells (CTCs, the root cause of cancer metastasis and poor cancer prognosis) are very difficult to culture for scaleup in vitro, which has hampered their use in cancer research/prognosis and patient-specific therapeutic development. Herein, we report a robust electro-microfluidic chip for not only efficient capture of heterogeneous (EpCAM+ and CD44+) CTCs with high purity, but also glutathione-controlled gentle release of the CTCs with high efficiency and viability. This is enabled by coating the polydimethylsiloxane (PDMS) surface in the device with a 10-nm gold layer through a 4-nm titanium coupling layer, for convenient PEGylation and linkage of capture antibodies via the thiol-gold chemistry. Surprisingly, the percentage of EpCAM+ mammary CTCs can be as low as ~35% (~70% on average), showing the commonly used approach of capturing CTCs with EpCAM alone may miss many EpCAM- CTCs. Furthermore, the CD44+ CTCs can be cultured to form 3D spheroids efficiently for scaleup. In contrast, the CTCs captured with EpCAM alone are poor in proliferation in vitro, consistent with the literature. By capturing the CTC heterogeneity, the percentage of stage IV patients whose CTCs can be successfully cultured/scaled up is improved from ~6.3% to 68.8%. These findings demonstrate the common practice of CTC capture with EpCAM alone misses the CTC heterogeneity including the critical CD44+ CTCs. This study may be valuable to the procurement and scaleup of heterogenous CTCs, to facilitate the understanding of cancer metastasis and the development of cancer metastasis-targeted personalized cancer therapies conveniently via the minimally invasive liquid/blood biopsy.

Graphical Abstract

INTRODUCTION

Cancer is the second leading cause of death worldwide and cancer metastasis/relapse is the major cause of cancer-related mortality.^1,2 For cancer metastasis in a distant organ to occur, tumor cells must enter the blood circulation and these circulating tumor cells (CTCs) must survive and extravasate into the distant organ to regenerate metastatic tumor.^3–8 Many malignancies including the vast majority of breast cancers are of epithelial origin (known as carcinoma) and their primary tumor cells display epithelial surface antigen markers like the epithelial cell adhesion molecule (EpCAM).^9–12 Therefore, EpCAM has been widely used as the targeting marker to capture CTCs from cancer patients’ blood for cancer detection, prognosis, and research.^13–18

However, it is now well established that cancer cells in primary tumors are highly heterogenous due to the various distinct genetic alterations of cancer.^19–21 Similarly, it has been reported that CTCs derived from primary tumors are made of a very heterogenous subpopulation of cells with different morphology, phenotype, and genetic characteristics.^8,22–25 Due to this heterogeneity, each cell subset may contribute differently to metastasis and/or to drug resistance, which complicates clinical intervention.²⁶ Moreover, missing this CTC heterogeneity may lead to diagnostic errors and misinterpretation of testing results.

Furthermore, recent studies show that CTCs may go through the epithelial-to-mesenchymal transition (EMT) process to lose their epithelial properties and acquire mesenchymal features instead.^3,27–31 Consequently, some CTCs may not express EpCAM or other epithelial markers. Therefore, the conventional use of EpCAM alone for capturing CTCs may lead to little or no capture of this subpopulation of CTCs. Moreover, since CTCs are rare in the blood (from a few to a few thousand of CTCs per milliliter of blood), the capability of in vitro culture/scale-up of CTCs is crucial to obtain a large number of the cells for further use in cancer research/prognosis and development of patient-specific treatment. Hence, a method for efficient capture of both the epithelial and non-epithelial subtypes of CTCs with high viability from patient blood to enable effective in vitro culture/scale-up of the cells, is of pivotal significance.

Even though a few studies have reported in vitro culture of CTCs to scale up the cells for further use and analysis, it has been very difficult to culture the cells in vitro and CTCs from only a very small percentage (~3–16%) of patients could grow for scale-up in vitro according to the literature.^32–38 Moreover, the effect of CTC heterogeneity on its long-term in vitro culture/scale-up has not been studied, although the CTC heterogeneity may be critically important to provide patient-specific tumor models for developing cancer heterogeneity-targeted personalized cancer therapies.³⁹ CD44, a multifunctional transmembrane glycoprotein, is one of the most common cancer markers and a promising prognostic biomarker for many malignancies including breast cancer.^40–44 Furthermore, CD44+ cancer cells have been shown to possess increased malignant potential of tumor initiation as well as self-renewal and multilineage differentiation, compared to CD44- cancer cells.^45,46 Moreover, the expression of CD44+ on cancer cells has been correlated with EMT.^47–50 Therefore, a combination of EpCAM and CD44 antibodies should enhance the capture of the CTC heterogeneity including both the epithelial and mesenchymal subpopulations of CTCs.

In this study, we report a robust microfluidic chip containing virtually implemented zones surface-coated with gold (Au) (VIZA chip) that uses a combination of electro-microfluidic technology and thiol-gold chemistry, to achieve highly efficient capture, glutathione-controlled gentle cell release, and greatly improved 3D culture of viable CTCs from breast cancer patients’ blood samples. We show that the CTCs isolated from patient blood using either the CD44 antibody alone or a combination of CD44 and EpCAM antibodies may grow and form tumor spheroids efficiently in 3D suspension culture in vitro. In stark contrast, CTCs isolated with EpCAM alone are poor in forming tumor spheroids under the same 3D suspension culture in vitro, due to a lack of the CD44+ CTC subpopulation. This work may be invaluable for capturing and in vitro expanding the rare heterogenous subpopulations of CTCs, to facilitate the understanding of their role in cancer metastasis, for improving cancer prognosis and developing cancer metastasis-targeted personalized therapies conveniently via the minimally invasive liquid/blood biopsy.

RESULTS AND DISCUSSION

Mechanisms of cell capture and controlled release with the VIZA chip.

As shown in Figures 1a and S1a, the PDMS, titanium (Ti), and Au-based VIZA chip has two inlets: Inlet 1 for introducing capture medium without cells to remove air from the device and inlet 2 for flowing capture medium with cells after air is removed from the device, followed by five sequential bifurcations to evenly distribute the sample into the main channel for capturing CTCs. The main channel contains parallelly patterned microposts (100 rows, 32 columns, and 100 μm in both diameter and height), surface-coated (except microposts in the four side columns with two on each side next to the copper electrode) with a nanoscale (10 nm) Au film via a coupling Ti film of 4 nm thick by physical vapor deposition. The gold coating gives the golden appearance in the real picture of the device shown on the top right of Figure 1a, although the golden appearance is absent in the scanning electron microscopy (SEM) image of the main channel. The gold surface is then decorated with capturing antibodies (Abs including EpCAM, CD44, or both; monoclonal) by using thiol-polyethylene glycol-N-hydroxysuccinimide (HS-PEG-NHS) via the thiol-Au conjugation chemistry and NHS-Ab displacement chemistry, to form a thiolated PEG bridge that links the capturing antibodies (CD44 and/or EpCAM) to the Au surface for capturing the target cells (Figure 1b, left). Lastly, the Au surface is modified with methoxy-PEG thiol (mPEG-SH) via the thiol-Au conjugation chemistry, to give surface PEGylation for blocking non-specific binding and improving capture purity.

Figure 1. The electro-microfluidic VIZA chip for efficient capture and controlled release of cancer cells with high viability.

a) Schematic illustrations together with a real picture and a scanning electron microscopy (SEM) image, showing the design of the VIZA chip. The illustrations show the configuration of the two inlets, entry channels, main channel, microposts, electrodes, and virtual capture versus flow zones. Inlet 1 is for flowing capture medium to remove air from the device while samples are introduced into the device via inlet 2. The device is made of two sections: 1) the flow-distribution section (14 mm in length) which contains the two inlets and bifurcated microchannels, and 2) the main channel section (16 mm in length) which contains 100 rows and 32 columns of microposts. Each micropost is 100 μm in height and 100 μm in diameter. The space between two adjacent microposts is 50 μm. Two copper electrodes are inserted into the space between the main channel and the sidewalls on both sides of the VIZA chip for applying the electrical voltage across the main channel, to generate the virtual capture and flow zones. All the microposts except those in four side columns (two on each side) next to the copper electrodes are coated with gold (Au). Scale bar: 100 μm. b) Schematic illustration of the decoration of multiple capturing antibodies (CD44 and EpCAM: CD44&EpCAM) and methoxy-PEG (mPEG) on the gold surface for capturing heterogeneous populations of cells (CD44+ or EpCAM+: CD44+/EpCAM+ which includes cells that are positive for both CD44 and EpCAM) and blocking non-specific binding, respectively, and the glutathione (GSH)-mediated ligand exchange reaction that enables controlled release of cells captured on the microposts in the main channel. The gold surface was made by physical vapor deposition (i.e., sputter coating) of the polydimethyl siloxane (PDMS) substrate with a gold (Au) film of 10 nm in thickness via a coupling titanium (Ti) film of 4 nm thick. c) Quantitative data on the efficiency for the GSH-mediated controlled release versus the conventional high flow washing-mediated release of three different types of cancer cells out of the VIZA chip, and the viability of the three different types of cancer cells after releasing them out of the VIZA chip using the two different methods. d) Bright field and fluorescence micrographs showing that MCF-7 cells from the GSH-mediated release can be cultured in a tissue culture plate with high viability. Scale bar: 50 μm. Error bar represents the mean ± s.d. of n = 3 independent runs, and *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001 indicates statistical significance.

The thiolated PEG bridge on the Au surface renders an important feature of the VIZA chip, viz, controlled release of the target cells captured via the capturing antibodies. This can be conveniently done by flowing glutathione (GSH) solution into the device for a ligand exchange reaction between GSH and the -S-PEG-Abs (Figure 1b). The reaction starts by adsorption of the GSH sulfur on the Au surface while the sulfur in the leaving -S-PEG-Abs breaks its bond with Au. This leads to the intermediate state where the incoming GSH thiol and the leaving -S-PEG-Abs are covalently bonded to the Au surface. This is subsequently followed by the GSH-derived hydrogen donation to the leaving -S-PEG-Abs, resulting in the binding of GSH to the Au surface and the departure of the cell bound HS-PEG-Abs. GSH was chosen in this study for its biocompatibility as the most abundant nonprotein thiol in humans.⁵¹ Different methods such as enzymatic digestion and aptamer-mediated release have been previously used for controlled release of captured cells.⁵² However, enzymatic digestion is known to degrade cellular membrane proteins and negatively affect overall cell viability.^53,54 For the aptamer-mediated release, the pH or temperature change needed can alter the conformation of cell surface proteins, which may modulate the molecular characteristics and/or viability of the released cells.⁵² These issues can be addressed by our GSH-based method to rapidly (~30 min) release the captured cells with high viability (more than 90%) for downstream analysis.

Another important feature of the VIZA chip is that the capture of the target cells via the capturing antibodies inside the main channel is facilitated by applying a non-uniformly distributed alternating current (AC) electric field through the two copper electrodes on both sides of the main channel. In the absence of an electric field, the cells move with the capture medium at high speed in the flow zones (FZs, Movie S1 (30 frames per second (fps)), the bottom-right sketch in Figure 1a, and Figure S1b), leading to low capture efficiency. When a patterned (3 s on and 3 s off) electric peak-to-peak voltage (V_pp) of 200 V with a frequency of 200 kHz is applied via the two copper electrodes, a dielectrophoresis (DEP) force is generated on the cells to move them from the FZs (spaces with high flow speed and low electric field) into the virtual (i.e., without a physical boundary) capture zones (CZs, pocketed spaces with low flow speed and high electric field, Movie S2 (30 fps), the bottom-right sketch in Figure 1a, and Figure S1b), which was also theoretically shown with force analysis in the literature.^17,55–57 It is worth noting that DEP has been investigated as an important technique for microfluidic manipulation of various biological particles including living cells.^58–61 In addition, the electric field strength due to the voltage (200 V) applied on the capture medium (made of non-conductive water and sugars) across the main channel (~5 mm in width) is ~4×10⁴ V m⁻¹, which is more than 1000 times lower than the dielectric field strength of water (65–70×10⁶ V m⁻¹),^62,63 the maximum electric field strength to break down water and make it electrically conductive. Therefore, there is no electric current (that may cause damage to cells) in the capture medium in the device.

The experimental setup is demonstrated in Figure S2. As can be seen in Movie S2, cells are moved into the CZs in the main channel by the DEP force when the electric field is on, and they are quickly released into the FZs when the electric field is turned off because no antibody is used in the device for taking the movie. Therefore, the combination of the antibody-functionalized microposts and an electric field gradient is needed for efficient capture of target cells in the CZs (Movie S3, 30 fps). The patterned electrical voltage with 3 s on/off intervals is used to avoid cell blockage in the CZs and improve capture purity, because many cells (both target and non-target ones) could be moved into the CZs in the upstream flow of the main channel. During the 3 s off period, cells that are not bound with the antibodies on the microposts in an upstream CZ (e.g., due to either cell crowding with no chance to touch the antibodies on the upstream microposts or little time for a complete bonding before the DEP is turned off), move back into the FZ by the flow-induced force in the surrounding fluid (Movie S3). This allows the target cells to be captured later via binding with the antibody in the downstream CZs when the electric field is on again (Movie S3). It is worth noting that this flow-induced force in the CZs is not strong enough to break the bonding between the cell and the capturing antibody on the micropost because of the low flow speeds in the CZs. Therefore, it can only remove unbound cells out of the CZs. In the presence of antibodies on the microposts, the capture of cells on the microposts in the CZs instead of the FZs can be seen in Figure S3a-b.

To determine the effect of antibody concentration on the capture efficiency, the EpCAM capturing antibody of 2 to 100 μg ml⁻¹ was used to modify the main channel surface of the VIZA chip before 500 MCF-7 cells were flowed into the device at a fixed flow rate of 1 ml h⁻¹. As shown in Figure S3c, the maximum cell capture efficiency is reached at 55 μg ml⁻¹, at which concentration the micropost surface is likely saturated with antibody. Therefore, this antibody concentration was used for all further studies.

Cell capture performance of the device is determined by flowing 500 MCF-7 cells spiked in 1 ml of capture medium through the VIZA chip at a flow rate of 1 ml h⁻¹. Figure S3d shows that a bare microfluidic device with neither the surface antibody modification nor the electric field in the main channel fails to capture any cancer cells on the microposts. Similarly, few cells can be captured in the device with antibody modification without an electric field in the main channel. This can be attributed to the lack of an electric field gradient to generate the DEP force for moving the cells from the FZs into the CZs and the flow-induced force is high enough to break the binding between the antibody and cells in the FZs with high speed (Movie S1). Also, almost no cells can be captured in the presence of an electric field alone without antibody modification, because the flow-induced force is high enough to move non-bound cells from the CZs into the FZs when the electric field is turned off (Movie S2). In stark contrast, a high capture efficiency of 97.0 ± 0.8% (n = 3) is obtained in the combined presence of the target-capturing antibody (EpCAM) and electric field in the main channel (p ≤ 10⁻³, Figure S3d). Under this condition, the cancer cells are captured on the microposts in the CZs, as shown in Figure S3a-b and Movie S3.

Furthermore, different flow rates are tested to study the effects of flow speed on the MCF-7 cell capture efficiency for determining an optimal flow rate. As shown in Figure S3e, capture efficiencies between 97.0 ± 0.8% (n = 3) and 82.0 ± 4.3% (n = 3) are achieved when the flow rate increases from 0.5 ml h⁻¹ to 2 ml h⁻¹, corresponding to the flow speeds from 0.290 mm s⁻¹ to 1.116 mm s⁻¹. Of note, a flow speed of ~0.15–0.3 mm s⁻¹ has been typically used for microfluidic CTC capture with antibodies and irregular structures.^16,64 However, the capture efficiency decreases further to 50.3 ± 6.8% (n = 3) and 33.0 ± 8.3% (n = 3) when the flow rate inside the main channel is increased to 3 ml h^{− 1} and 4 ml h⁻¹, corresponding to the flow speeds of 1.740 mm s⁻¹ and 2.320 mm s⁻¹, respectively. This indicates that under the ultra-high flow speed (> 1.116 mm s⁻¹) condition, the resultant shear force may compete with the binding forces between cancer cells and capture antibodies. Since the highest capture efficiency of 97.0 ± 0.8% is achieved at 1 ml h⁻¹, this flow rate is used for all subsequent experiments. It is worth noting that this flow rate corresponds to a flow speed of 0.580 mm s⁻¹, which is not only 2–4 times of the flow speed typically used in the literature as aforementioned for improved throughput, but also within the range of the average blood flow speed in the human capillaries (0.5–1.5 mm s⁻¹).⁶⁵

We further compared the capture efficiency for different human breast cancer cell lines (MCF-7, MDA-MB-231, MDA-MB4–68, and HCC-1937) and a human prostate cancer cell line (PC-3) with different levels of EpCAM and CD44 expression, for which the cells were stained with calcein-acetoxymethyl ester (calcein-AM that becomes green fluorescent after interacting with esterase in cells) and the red fluorescence probe propidium iodide (PI to stain any cells with compromised plasma membrane) and spiked into the capture medium at a final concentration of 500 cells ml⁻¹ before flowing them separately into respective EpCAM antibody-modified devices. Captured and non-captured cells are counted with fluorescence microscopy to determine the capture performance. The capture efficiency is 97.0 ± 0.8%, 90.0 ± 3.6%, 84.0 ± 3.0%, and 92.3 ± 4.2% (n = 3) for MCF-7, MDA-MB-468, HCC-1937, and PC-3 cells, respectively (Figure S3f). The high capture efficiency can be attributed to the cells’ high expression of EpCAM.^17,66,67 In contrast, a very low capture efficiency of 6.31 ± 4.9% is obtained for the MDA-MB-231 cells, which can be explained by the fact that the MDA-MB-231 cells are post EMT with low EpCAM expression.^68,69 Importantly, when CD44 is used as the capturing antibody, the resultant capture efficiency for MDA-MB-231 cells is 95.3 ± 1.3%, presumably due to their high CD44 expression.^45,70,71 However, the use of CD44 as the capturing antibody leads to a low capture efficiency for MCF-7, MDA-MB-468, and PC-3 cells (3.7 ± 0.5%, 6.3 ± 1.2%, and 6.7 ± 1.2%, respectively), presumably because they are epithelial in nature with negligible expression of CD44. The moderate capture efficiency of HCC-1937 cells with both EpCAM and CD44 (84.0 ± 2.4% and 46.7 ± 2.1%) aligns very well with its basal-like nature with moderate expression of both protein markers.^72,73 Moreover, the capture of white blood cells (WBCs) is negligible in the VIZA chip with either EpCAM or CD44 antibodies (Figure S3f). This confirms the specificity of the VIZA chip to selectively capture cancer cells of different types per the specific capturing antibody immobilized on the microposts in the main channel.

We subsequently investigate the sensitivity of the VIZA chip for cancer cell capture by varying the concentration (2–1200 cells ml⁻¹) of MCF-7 cells spiked in capture medium without or with WBCs (4.0 x10⁶ cells ml⁻¹, the typical concentration of WBCs in human blood).¹⁷ The analysis of the number of captured versus spiked cells yields a Pearson’s correlation coefficient of 0.998 and 0.988 for the cases without and with WBCs, respectively, with an average capture efficiency (i.e., the slope of the fitted regression line) of ~95% for both cases (Figure S4a-b). This capture efficiency is well above the previously published efficiency of ~60% using antibody or aptamer-modified microfluidic chip.⁷⁴ This high capture efficiency can be attributed to both the formation of the distinct virtual CZs enabled by the patterned electric field applied via the copper electrodes and the high affinity and specificity of the thiolated antibodies on the Au surface in the main channel. It is worth noting that most of the cells are captured within the first ~60 rows of microposts (~5–10 mm of the main channel surface), with fewer cells toward the exit of the main channel (Figure S4c where 500 cells per ml is used). This indicates the VIZA chip is not saturated and has room to capture more cells, to ensure the capture of all target cells in the sample with the device.

Highly efficient controlled release of cells captured in the VIZA chip.

Besides the efficient capture, it is crucial to release the captured cancer cells from the device conveniently with high viability for further use. Therefore, we investigate the release efficiency (number of cells released from the device over the total number of cells captured in the device) of captured cells. A total of 500 MDA-MB-231, MDA-MB-468 and 500 MCF-7 cancer cells, spiked into 1 ml of capture medium containing 4 x 10⁶ WBCs, were flowed into different EpCAM-based VIZA chips for capture at 1 ml h⁻¹. Afterward, GSH (1.5 mg ml⁻¹ in 1x PBS containing 1.5%wt bovine serum albumin (BSA)) is injected to fill the main channel of the device, and after waiting for 30 minutes, the device is flowed with cell culture medium at 1.0 ml h⁻¹ into a Petri dish to collect the released cells. The release efficiencies are 94.0 ± 2.2%, 94.0 ± 3.3% and, 96.1±1.2% (n= 4) for the MDA-MB-231, MDA-MB-468, and MCF-7 cells, respectively (Figure 1c). These release efficiencies are much higher than that (25.2 ± 2.3%, 21.5 ± 2.6%, and 20.7 ± 1.5% (n= 4) for MDA-MB-231, MDA-MB-468, and MCF-7, respectively) achieved via the commonly used high-flow washing at 8 ml h⁻¹, corresponding to the flow speed of 4.64 mm s⁻¹.

The viability of released cells is studied with the standard live/dead (green/red) cell viability assay consisting of calcein AM (green) and PI (red). The viability of the captured cells after being released by the GSH ligand exchange reaction is 96.0 ± 0.4%, 90.4 ± 3.0%, and 94.4 ± 4.1% for MDA-MB-231, MDA-MB-468, and MCF-7 cells, respectively, whereas the high-flow (8 ml h⁻¹) washing release results in reduced viabilities of 55.0 ± 3.1%, 51.6 ± 4.6%, and 63.5 ± 8.5% for the MDA-MB-231, MDA-MB-468, and MCF-7, respectively (Figure 1c). The large decrease in viability of the cells released via high-flow washing may be attributed to the large shear stress emanating from the high flow speed gradient in the main channel, which can be avoided by using our GSH ligand exchange reaction-based method. Furthermore, the cells released by the GSH-controlled method can attach to a tissue culture plate with high viability as shown in Figure 1d for MCF-7 cells. Overall, these results demonstrate the ability to efficiently, rapidly, and safely release cells captured in the VIZA chip in a controlled manner for further downstream analysis. Our GSH-mediated release performed at room temperature may also address some of the issues associated with other uncommonly used methods reported in the literature, as summarized in Table S1. Of note, the cells released by the GSH-controlled method have one PEG linker attached to their outer plasma membrane surface right after release. However, this does not significantly affect their viability, probably because the number of attached PEG linker is so small (1) and it may quickly get internalized via endocytosis when the cells are cultured at 37 °C. In addition, the antibody may be modified with Au to interact with the thiolated PEG linker (or vice versa) so that the cells could be released without a PEG linker using GSH, which warrants further investigation in the future.

It is worth noting that multiple of the unit device shown in Figure 1a may be used in parallel to further increase the flow rate. An integrated VIZA chip with eight of the unit devices being connected in parallel (Figure S5a) was fabricated to test spiked cancer cell capture at the flow rate of 8 mL h⁻¹, which would allow 1 mL of sample to be processed in only 7.5 min (1/8 h). EpCAM antibody was used to functionalize the surface. A total of 1000 MCF-7 and MDA-MB-468 breast cells were stained with calcein-AM and PI as aforementioned before spiking them into 1 ml of capture medium for capture at 8 ml h⁻¹. The capture efficiency is 92.7 ± 2.3% (n= 5) and 94.2 ± 1.1% (n= 5) for MDA-MB-468 and MCF-7 cells, respectively (Figure S5b). Afterward, GSH is used to release the captured cells as aforementioned, and the device is flowed with the cell culture medium at 8.0 ml h⁻¹ into a Petri dish to collect the released cells. The release efficiency is 93.0 ± 4.0 % (n= 5) and 91.7 ± 4.2% (n= 5) for MDA-MB-468 and MCF-7, respectively. (Figure S5b). In a parallel study, non-stained cells were captured and released in the same way to determine their viability in the same way as aforementioned. The viability is 96.5 ± 1.8% and 94.3 ± 1.3 % for the MDA-MB-468 and MCF-7 cells, respectively (Figure S5c). The capture efficiency, release efficiency, and cell viability of the integrated VIZA chip is high and similar to the unit VIZA chip despite using an 8-time higher flow rate (8 ml h⁻¹). The ability to isolate cancer cells in less than 8 min is of an utmost importance for fast sample processing, especially in clinical setting. However, the unit VIZA chip was used for further studies in this work done in the lab.

Highly efficient capture of a heterogeneous cancer cell population with the combination of EpCAM and CD44.

In addition to a single surface antibody modification for capture a single subpopulation of cancer cells, a combination of two antibodies (EpCAM and CD44 at 50% and 50%, respectively) is used to test the multi-antibody capturing ability in comparison to the use of EpCAM or CD44 alone of the VIZA chip. A mixture of 250 calcein AM stained MDA-MB-231 (EpCAM^lowCD44^high) and 250 1,1’-dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate (DiI)-labeled MCF-7 (EpCAM^highCD44^low) cells, are suspended in capture medium with 4.0 x 10⁶ WBCs ml⁻¹ and injected into the device at 1 ml h⁻¹ for capture. The captured cells are counted against the non-captured cells to determine the overall capture efficiency. Figure 2a shows representative fluorescence images used to quantify the cell capture efficiency. When either EpCAM or CD44 antibody is employed alone, approximately half of the cells can be captured (Figure 2a-b). This is due to the inability of EpCAM and CD44 to capture MDA-MB-231 and MCF-7 cells, respectively (Figure S3f). This demonstrates that a single capturing antibody is not efficient enough to capture cells like the CTCs with heterogenous marker expressions or subpopulations. In contrast, when the combination of EpCAM and CD44 is used to capture a mixture (1:1) of the MCF-7 and MDA-MB-231 cancer cells, the capture efficiency is improved to 95.0 ± 0.8% (n=3) (Figure 2a-b). This confirms the ability of the VIZA chip to accommodate multi-antibody modification and capture cancer cells of heterogeneous marker expressions, and suggests the importance of using a combination of different antibodies to ensure high efficiency for capturing all cells in a heterogenous cancer cell population with different stages of EMT in tumors and among the CTCs that detach from primary tumors and intravasated into the blood circulation.^75–77

Figure 2. Capture of the cancer cell heterogeneity with a combination of two antibodies.

a) Fluorescence micrographs showing captured cells from a man-made mixture (1:1) of MDA-MB-231 and MCF-7 cancer cells with the VIZA chip functionalized with EpCAM antibody alone, CD44 antibody alone, or the combination of the two antibodies. Scale bar: 100 μm. b) Quantitative data on the capture efficiency of the VIZA chip functionalized with EpCAM, CD44, and EpCAM & CD44 for capturing a heterogeneous cancer cell population of either a man-made mixture of half MCF-7 and half MDA-MB-231 cells or naturally formed MCF-7 mammospheres. The combination of the two antibodies is significantly much more efficient in capturing cancer cells of a heterogeneous population than either antibody alone. c) Scatter dot plots showing the number of CTCs captured from early (I-III) and late (IV)-stage breast cancer patients altogether, along with healthy donor samples using CD44, EpCAM, and EpCAM & CD44. The EpCAM & CD44 was used for studying blood samples from healthy donors. d) Scatter dot plots showing the number of CTCs captured from early (I-III) stage breast cancer patients by CD44, EpCAM, and EpCAM & CD44. e) Scatter dot plots showing the number of CTCs captured from late (IV) stage breast cancer patients by CD44, EpCAM, and EpCAM & CD44. f) Scatter dot plot showing a simple addition of the number of CTCs captured by CD44 and EpCAM in separate devices in comparison to the number of CTCs captured by CD44 & EpCAM in one device, from all patients altogether. g) Ratio of the number of CTCs captured by CD44 to the number of CTCs captured by EpCAM & CD44 from patients with stages I, II, III, and IV breast cancer. h) Box and whisker plots showing the purity of CTCs captured with the three different methods. Error bar in panel b represents the mean ± s.d. of n = 3 independent runs. The difference is considered statistically significant when p ≤ 0.05. ***p < 0.001, **p < 0.01, and *p < 0.05; n.s.: not significant; CD44: CD44 antibody alone, EpCAM: EpCAM antibody alone, and EpCAM & CD44: combination of EpCAM and CD44 antibodies.

To further confirm the significance of using this combination of CD44 and EpCAM antibodies, the efficiency of capturing cells in a naturally formed heterogeneous cell population is studied by flowing into the VIZA chip, 500 cells (in capture medium with 4.0 x 10⁶ WBCs ml⁻¹) detached from the human MCF-7 cells-derived 3D breast tumor spheroids (i.e., mammospheres) that are enriched with human mammary cancer stem cells (CSCs).^46,78 As shown in Figure 2b, CD44 or EpCAM antibody alone results in a capture efficiency of 35.0 ± 4.6% and 50.0 ± 3.2%, respectively. These are not significantly different from the flow cytometry analyses of the two subpopulations of cells in the MCF-7 mammospheres (41.8 ± 3.8% and 58.8 ± 4.5% for CD44 and EpCAM, respectively, Figure S6). Since the capture efficiency of a regular MCF-7 cell population with the CD44 antibody is only 3.7 ± 0.5%, this result confirms that MCF-7 derived tumor spheroids are enriched with CD44+ mammary CSCs. Importantly, when the combination of CD44 and EpCAM is used, the capture efficiency is increased to 90.0 ± 5.4%. These data show that the use of EpCAM alone (as commonly done in the literature⁷⁹) for capturing a heterogeneous cancer cell population may miss not only many cancer cells but also the highly tumorigenic CD44+ subpopulation, which may be addressed by combining EpCAM with CD44 as the capturing antibodies (at least for breast cancer). This is further confirmed by capturing CTCs from breast cancer patients’ blood samples, as detailed below.

Augmented capture of CTCs from breast cancer patients’ blood samples with the combination of EpCAM and CD44.

To evaluate the clinical potential of the VIZA chip with the combination of EpCAM and CD44 antibodies for capturing CTCs, blood samples from a cohort of 60 female breast cancer patients are used. According to the American Joint Committee on Cancer (AJCC) staging,^80,81 32 and 28 of the 60 patients have early-stage (i.e.., stage I, II, or III) and late-stage (stage IV) breast cancer, respectively. Blood samples from 7 healthy individuals are also studied. A summary of the cancer stage, subtype, and marker expression together with the race and age of all the 60 patients are given in Table S2. A total of 1 ml of each patient’s blood sample is pre-treated to remove RBCs, resuspended in 1 ml of capture medium, and then flowed into the VIZA chip at 1 ml h⁻¹ for capturing CTCs. To characterize the captured cells, multi-fluorescence staining (Figure S7a) including DAPI (blue) for DNA, carcinoembryonic antigen (CEA, an important molecule for intracellular recognition and attachment of cancer cells),⁸² cytokeratin 18 (CK18, an epithelial tumor marker)^18,83 for breast cancer cells/CTCs (green), and CD45 for WBCs (red); to distinguish CTCs from WBCs.^84,85 DAPI is used to identify all cells in the VIZA chip. Captured cells with (CEA+/CK18+)CD45- expression are identified as CTCs. Captured cells with CD45+ expression, are identified as WBCs. CD44 alone, EpCAM alone, and a combination of both antibodies are used to functionalize different VIZA chip devices to capture CTCs from the patients’ blood samples. For each patient, two types of experiments are conducted: one for CTCs enumeration via fluorescence microscopy and the other without staining for capture, controlled release, and in vitro culture.

First of all, CTCs are captured only in cancer patients’ blood samples and no cancer cells are captured from the healthy control samples with the combination of CD44 and EpCAM antibodies (Figure 2c). This is not surprising since tumor cells should be absent in the circulation of individuals who are free of cancer. As shown in Figure 2c for all (i.e., both early/I-III and late/IV stages in blue and red circles, respectively) cancer patients altogether, a significantly higher number of captured CTCs ranging from 4 to 187 (median = 50) CD44+/EpCAM+ (including CD44+, EpCAM+, and CD44+EpCAM+ if any, Figure 1b) CTCs ml⁻¹ is observed for the combination of EpCAM and CD44 capturing antibodies than either EpCAM alone (4 to 121 EpCAM+ CTCs. ml⁻¹, median = 36) or CD44 alone (1 to 64 CD44+ CTCs ml⁻¹, median = 13). This demonstrates the importance of using a combination of the two antibodies to capture CTCs than any one of them.

The capture of CTCs for the early (I-III) and late (IV) stages of patients is further analyzed separately. With CD44 or EpCAM antibody alone being used for capturing, CTCs captured in early-stage (I-III) breast cancer patients range from 1 to 28 CD44+ CTCs ml⁻¹ (median = 8) and 4 to 76 EpCAM+ CTCs ml⁻¹ (median = 21), respectively (Figure 2d). Again, a significantly higher number of CTCs ranging from 4 to 88 CD44+/EpCAM+ (including CD44+, EpCAM+, and CD44+EpCAM+ if any, Figure 1b) CTCs ml⁻¹ (median = 30), are captured when a combination of EpCAM and CD44 is used. Similarly, the use of EpCAM and CD44 antibody combination results in capturing significantly more CTCs from the late-stage (IV) cancer patients’ blood samples (17 to 187 CD44+/EpCAM+ CTCs. ml⁻¹, median = 92) than using either EpCAM alone (8 to 121 EpCAM+ CTCs.ml⁻¹, median = 62) or CD44 antibody alone (6 to 64 CD44+ CTCs, median = 36) (Figure 2e). These data demonstrate the heterogeneous nature of the CTCs in a cancer patient’s blood and highlight the importance of using more than one antibody for cancer detection and capture of the CTC heterogeneity.

It is worth noting that when the number of CTCs captured by EpCAM and that captured by CD44 antibodies are added together, they did not yield any statistical difference from the total number of CTCs captured by the combination of EpCAM and CD44 antibodies (Figure 2f), indicating the combination of the two antibodies in one device does not compromise their efficacy of capturing their respective target cells in this study. Interestingly, the ratio of CD44+ CTCs to all CTCs captured by a combination of EpCAM and CD44 antibodies shows an increasing trend from stage I to IV, with the difference between stage IV and stage I being statistically significant (Figure 2g). This might explain the highly aggressive and metastatic nature of stage IV cancer since CD44+ cancer cells have been shown to be highly tumorigenic and invasive.^43,86,87 Furthermore, the ratio or percentage of CD44+ mammary CTCs captured with both EpCAM and CD44 can be as high as ~65% for stage IV breast cancer (Figure 2g), indicating the percentage of EpCAM+ CTCs can be as low as ~35% in a CTC population (although they are ~70% on average). This further demonstrates the necessity of using CD44 in addition to EpCAM for capturing the CTC heterogeneity and in particular, the crucial subpopulation of CD44+ CTCs.

In addition, the combination of two antibodies does not affect the capture purity (percentage of CTCs out of all cells in the device) either, and the purity is all high (≥ ~80% on average) for capturing with EpCAM alone, CD44 alone, and a combination of both antibodies (Figure 2h), compared to the less than ~30% purity reported in the literature.^16,17 This is attributed to both the surface PEGylation that minimizes non-specific binding with non-CTCs and the use of the patterned electric voltage that prevents blockage of non-CTCs in the CZs. It is worth noting that the thiol-Au surface chemistry allows facile antibody modification, which makes the VIZA chip highly adaptable in capturing various cancer cells in a heterogeneous population via their specific antibodies.

The CD44+ CTCs are highly tumorigenic and capable of self-renewal.

As CD44 is a well-established marker of cancer stem cells (CSCs) isolated from primary breast tumors,^45,88,89 we hypothesize that CD44+ CTCs isolated with CD44 antibody from the patients’ blood samples are made of a stem cell-like population of cells with a high capacity for tumorigenesis and self-renewal compared with CTCs captured with EpCAM alone. To test this hypothesis, CTCs isolated with EpCAM alone, CD44 alone, and a combination of both EpCAM and CD44 are released from the VIZA chip using the aforementioned GSH-based method. They are then cultured in serum-free defined culture medium in ultra-low attachment plates (ULAP) to evaluate their proliferation and tumor spheroid formation ability.

Before culturing the released CTCs, we studied their viability first. CTCs from blood samples of four patients 16,19, 26 and 37 were used for this purpose. Briefly, the CTCs captured by the combination of EpCAM and CD44 antibodies were released into a 6-well cell culture plate containing 1mL of cell culture medium and placed at 37 °C in a humidified air with 5% CO₂ incubator for 24 h. A standard live/dead (green/red) cell viability assay consisting of calcein AM (green) and PI (red) was used to quantify the viability of the cells. The viability was 90.0%, 94.7%, 91.3%, and 92.8% for patient 16, 19, 26 and 37, respectively (Figure S7b). The high viability of the released CTCs confirms that the mild nature of the procedure of CTC capture and GSH-controlled release in the VIZA chip.

For further culture, CTCs (60 to 150) captured by the aforementioned three methods from 34 patients (16 late stage IV and 18 early stages I-III) are seeded in a serum-free defined medium in ULAP and monitored for over 35 days to study their spheroid formation efficiency (SFE) calculated as the percentage of the total number of tumor spheroids formed out of the total number of live CTCs seeded. As an example, evident (i.e., visible by eye) spheroid formation of CD44+ CTCs derived from patient #16 (ER-PR-HER2+, Table S2) can be observed on day 25 of culture, and the spheroids with high viability can reach ~100 μm in diameter on day 35 (Figure 3a). In contrast, no evident spheroids can be observed to form from CTCs captured by EpCAM antibody alone from the same patient’s blood sample. The SFE data on day 35 for CTCs captured using the three different methods (i.e., CD44 alone, EpCAM alone, and combination of EpCAM and CD44) from all the 34 patients are shown in Figure 3b. Up to ~90% (dependent on the cancer stage) of the CD44+ CTCs captured with CD44 alone form tumor spheroids. For the CTCs (a mixture EpCAM+ CTCs and CD44+ CTCs) captured by the combination of EpCAM and CD44, up to ~70% of them can form tumor spheroids, which is significantly lower than that for the CD44+ CTCs captured by CD44 alone. Interestingly, for the EpCAM+ CTCs captured by the EpCAM antibody alone, few tumor spheroids can be observed to form in only ~11.1% (2 out of 18, at stages II&III) early stage and ~6.3% (1 out of 16) stage IV patients. The latter is consistent with that (~3–16%) reported in the literature for stage IV patients.^32–38 With further culture, these spheroids gradually dissociate into single cells and disappear after reaching 66, 68, and 90 μm in diameter (on average) on day ~25 (Figure S8). In contrast, CD44+ CTCs derived spheroids keep growing in size and reach up to ~300 μm after day 60. As shown in Figure S9, CTCs captured by CD44 from patient 26 keep growing and fuse together to form larger aggregates. These data indicate the highly tumorigenic nature of the CD44+ CTCs in vitro compared to the EpCAM+ CTCs and capturing CTCs with EpCAM alone misses most if not all of the critical subpopulation of CD44+ CTCs that may be responsible for cancer metastasis and relapse. It is worth noting that for the 18 early-stage cancer patients-derived CD44+ CTCs studied, the CD44+ CTCs from only ~11.1% (2 out of 18, all from patients with stage III cancer: #19 and #37, Table S2) are able to form large tumor spheroids (100–300 μm in diameter on average), while the CD44+ CTCs from ~68.8% (11 out of 16) of the stage IV patients are able to form tumor spheroid. This percentage (68.8%) is much higher than that (~3–16%) reported in the literature for in vitro culture of CTCs from stage IV breast cancer patients.^32–38 This may explain why further cancer metastasis is more likely to occur in late than early-stage cancer patients.

Figure 3. Enhanced in vitro cultivability of CTCs captured with CD44 compared to that captured with EpCAM.

a) Typical phase-contrast and fluorescence micrographs showing the formation and growth of breast cancer patient tumor spheroids with high viability (green fluorescence), from single CTCs captured from the blood samples of patient #16 by CD44 alone and those captured by EpCAM alone, under 3D suspension culture in defined medium. The spheroids reach ~100 µm in diameter on day 35. b) Spheroid formation efficiency (SFE) from a total of 34 breast cancer patients (18 stage I-III, and 16 stage IV) derived CTCs that are captured by CD44, EpCAM, and EpCAM & CD44. c) Typical micrographs of tumor spheroids at passages 0 to 3 showing a spherical shape during in vitro culture by serial passaging under suspension culture in the serum-free defined medium. Scale bar: 100 μm. d) SFE from 0^th to the 3^rd passages of 3D suspension culture in serum-free defined medium (n= 5). e) The size of the spheroids at the four different passages over time from ten breast cancer patients’ CTCs captured by CD44. The spheroids at the 0^th, 1^st, and 2^nd passages at day 35 are enzymatically dissociated into single cells for 3D suspension culture in the defined medium to form spheroids at the 1^st, 2^nd, and 3^rd passages, respectively. f) Bar plot showing the number of breast cancer patients whose CTCs yielded tumor spheroids during the 3D suspension culture in the defined medium or 2D culture in regular cancer cell media including DMEM with 10% fetal bovine serum (FBS) and DMEM/F12 with 10% FBS. Out of all CTCs from the 60 patients studied, CTCs from only five patients are able to adhere/proliferate in the two regular cancer culture media, while CTCs from 36 patients formed tumor spheroids under the 3D suspension culture in the serum-free defined medium. CTCs from a total of 24 patients do not grow under any of the culture methods. g) Spheroids at 0^th passage grown from CTCs of patient #16 are dissociated into single cells and cultured under 2D in DMEM with 10% FBS can adhere to the culture surface after 3 days of culture. After detaching the resultant 2D cancer cells on day 5 for 3D suspension culture in the defined medium, tumor spheroids of ~100 µm can be seen after only 10 days of culture. Scale bars: 100 µm, except that it is 50 µm for the images on days 0–25 in panel a. ***p < 0.001, and **p < 0.01; n.s.: not significant; CD44: CD44 antibody alone; EpCAM: EpCAM antibody alone; and EpCAM & CD44: combination of EpCAM and CD44 antibodies.

To evaluate the self-renewal ability of the CD44+ CTCs, they are subjected to the serial passaging assay for 3 more passages to quantify their SFE, by dissociating the spheroids into single cells for culture in the same way in the ULAP. The EpCAM captured CTCs are excluded from this study due to their inability to form tumor spheroids. As shown in Figure 3c-d, similar SFE on day 35 is observed for all the passages of the cells captured using either CD44 antibody alone or the combination of CD44 and EpCAM antibodies (from patient #16, as an example), suggesting that the capability of self-renewal that is an important characteristic of all stem-like cells. It is worth noting that the spheroids of the 1^st, 2^nd, and 3^rd passages derived from CTCs captured using CD44 alone from 10 different patients (#16, #18, #19, #25, #26, #37, #39, #40, #42, and #45, Table S2), grow much faster and larger in size than that of the 0^th passage, reaching ~100 μm in diameter in only 15 days of culture (Figure 3e).

Of note, among the CTCs derived from the 60 patients that are captured by a combination of EpCAM and CD44 capture antibodies, the CTCs from only 4 of them can adhere/proliferate under 2D culture in the regular cancer cell culture medium: Dulbecco’s Modified Eagle Medium (DMEM)/F12 with 10% fetal bovine serum (FBS, Figures 3f and S10), and the CTCs from only 1 of them can adhere/proliferate under 2D culture in the regular cancer cell culture medium: DMEM with 10% fetal bovine serum (Figures 3f and S10). In stark contrast, the CTCs from 36 of them can form tumor spheroids under 3D suspension culture in the serum-free defined culture medium, although the CTCs from 24 of them do not grow under any of the three culture conditions (Figure 3f). This is consistent with previous studies reporting that it is difficult to grow patient derived CTCs in regular cancer cell culture media.^32–37 Interestingly, even though the 0^th passage CD44+ CTCs do not attach and/or proliferate under 2D culture in regular cancer cell culture media (i.e., DMEM or DMEM/F12 supplemented with 10% serum), cells from the 0^th passage spheroids can adhere to the 2D culture plate after 72 h (i.e., on day 3) of culture and continue to grow (Figure 3g). Moreover, when detached for suspension culture in the defined medium in ULAP, these cells can form tumor spheroids of ~100 μm in diameter efficiently in 10 days (Figure 3g). It is worth noting that other non-antibody-based methods (e.g., size-based ones) may be used for capturing heterogeneous CTCs.^7,8 However, no study explored the effect of the CTC heterogeneity on the culture of CTCs until this work, and further investigation on the culturability of the CTCs captured by the non-antibody-based methods is warranted.

Analysis of marker expression in spheroids grown from CD44+ CTCs.

We first examined patient #16 derived spheroids for CD44+ and CD24-, the two most commonly used markers for highly tumorigenic breast CSCs. Indeed, the expression of CD44 is high with 97.6 ± 0.9% of cells in the spheroids being CD44+, and the expression of CD24 is low with 91.2 ± 6.0% of cells being CD24- (Figure 4a-b), suggesting that the CD44+ CTC subpopulation bears the characteristics of CSCs. Further studies are performed to understand the expression of two other commonly used CSC markers: aldehyde dehydrogenase 1 family member A1 (ALDH1A1) and CD133 (Figure 4b), together with two pluripotent stem cell markers (NANOG and OCT4), and Ki-67 that is a marker for all proliferating cancer cells (Figure 4a-b). The percentage of ALDH1A1+, CD133+, NANOG+, OCT4+, and Ki-67+ subpopulations in the tumor spheroids derived from CD44+ CTCs are 94.2%, 76.7%, 72 ± 4.2%, 89.4%, and 67.6% respectively. These data further support the cells in the CD44+ CTCs-derived spheroids are indeed CSC-like cells, although they are highly heterogeneous in terms of not only the expression of the stemness and pluripotency markers but also their proliferative activity. Presumably, the CD44+ CTCs are also CSC-like, which may contribute to their improved capability of growing into spheroids under 3D suspension culture in the defined medium. In contrast, EpCAM+ CTCs are more differentiated cells that do not possess the CSC property, and they could not grow under 3D suspension culture in the defined medium. However, there must be other reasons that affect the in vitro culture of CTCs since CTCs (CD44+ and EpCAM+) from ~30% stage IV patients and nearly all stages I-III patients could not be grown still, which warrants further investigation.

Figure 4. Heterogeneous marker expression profile of cells in tumor spheroids grown from CTCs captured with CD44.

a) Representative fluorescence micrographs showing immunocytochemical analysis of commonly used cancer stem cell (CD44 and CD24-), pluripotent stem cell (NANOG and OCT4), and cancer proliferation (Ki-67) markers for breast cancer cells. b) Typical flow cytometry histograms and quantitative percentage data showing the expression of various cancer stem cell (CD44, CD24-, ALDH1A1, and CD133), pluripotent stem cell (NANOG and OCT4), cancer proliferation (Ki-67), epithelial (EpCAM), mesenchymal (VIMENTIN), and WBC (CD45) markers. The percentages of cells positive for selected markers were determined using FlowJo Software. Error bar represents the mean ± s.d. of n = 3 independent runs. c) Representative fluorescence micrographs showing immunocytochemical analysis of commonly used epithelial (EpCAM), mesenchymal (VIMENTIN), and WBC (CD45) markers. Scale bars: 100 μm. The cell nuclei in all images were stained with DAPI.

The tumor spheroids derived from the CD44+ CTCs are also examined for the epithelial cell marker: EpCAM, a mesenchymal cell marker: VIMENTIN, and a white blood cell marker, CD45 (Figure 4b-c). Surprisingly, 62.2% of the cells in the spheroids are VIMENTIN+ and only 60.3% of the cells express EpCAM, indicating most of the spheroid cells (or the CD44+ CTCs, assuming the in vitro 3D culture causes negligible changes to the cells) undergo EMT either fully or partially as breast cancers are of epithelial origin. This is consistent with our spiked cell capture studies using cells from the spheroid derived from the MCF-7 human breast cancer cell line (Figure 2b), which shows that only 51.0 ± 3.3% of the cells in the MCF-7 spheroids express epithelial marker EpCAM although more than 95% of the parent MCF-7 cells are EpCAM+. Lastly and importantly, expression of CD45 is negligible on the cells in the CD44+ CTCs-derived spheroids, confirming the CTCs captured using CD44 antibody and the spheroid cells grown from them are not contaminated by WBCs.

CONCLUSION

In conclusion, we have designed an electro-microfluidic device (the VIZA chip)-based assay for highly efficient and rigorous capture and controlled release of not only EpCAM+ but also CD44+ CTC subpopulations simultaneously from breast cancer patients’ blood samples. We demonstrate the VIZA chip’s ability to capture the CD44+ CTCs by a single antibody and a combination of antibodies. By using a combination of CD44 and EpCAM, the CD44+/EpCAM+ CTCs in all stages of breast cancer can be captured with high efficiency and purity. This is achieved by using the VIZA chip with a combination of: 1) a facile surface immobilization of multiple antibodies through the gold-thiol chemistry enabled by coating the PDMS surface with a nanoscale gold layer via a nanoscale titanium coupling layer, 2) a DEP method for moving cells from the virtual flow zone into the virtual capture zone to improve throughput, 3) a GSH exchange chemistry for highly efficient and gentle cell release. Our data indicate capturing CTCs with EpCAM alone misses the critical CD44+ CTC subpopulation that can form tumor spheroids much more efficiently under in vitro 3D culture than the EpCAM+ CTC subpopulation, suggesting the CD44+ CTC subpopulation may be a major driver of cancer tumorigenesis and metastasis. These findings together with the VIZA chip may be invaluable for understanding cancer metastasis, advancing cancer prognosis, and developing cancer metastasis-targeted personalized cancer therapies conveniently via the minimally invasive liquid/blood biopsy.

Methods

Materials and reagents:

Cancer cell lines including MCF-7 human breast adenocarcinoma cells, HCC-1937 human breast ductal carcinoma cells, MDA-MB-231 human breast adenocarcinoma cells, MDA-MB-468 human breast adenocarcinoma cells, and PC-3 human prostate adenocarcinoma cells, were purchased from American Type Culture Collection (ATCC, Manassas, VA, USA). Fresh human WBCs were purchased from Zen-Bio (Durham, NC, USA), kept at 4 °C, and used within 1–2 days. Purified anti-mouse/human CD326 (EpCAM, clone 9C4, Cat # 324201), purified anti-mouse/human CD44 (CD44, clone BJ18, Cat # 338802), purified anti-mouse ALDH1A1 antibody (clone M562, Cat # 861902), Purified mouse isotype control IgG1 (clone MG1–45, cat # 401402), Purified anti-mouse CD24 Antibody (clone 30-F1, Cat: 138502) monoclonal antibodies were purchased from BioLegend (San Diego, CA, USA). Rabbit monoclonal anti-CD24 antibody (Cat # ab202073), rabbit polyclonal to CD133 antibody (Cat # ab19898), fluorescein isothiocyanate (FITC)-conjugated anti-carcinoembryonic antigen (CEA) (Cat # ab106739), and FITC-conjugated anti-cytokeratin 18 (FITC-CK18, Cat # ab52459) were purchased from Abcam (Cambridge, MA, USA). Alexa Fluor 532-labelled anti-CD45 (Cat # 58–0459-42), Live/dead viability/cytotoxicity kit: Calcein acetoxymethylester (calcein AM) and propidium iodide (PI), 1,1’-Dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate (Dil, Sigma-Aldrich, St. Louis, MO), rabbit monoclonal anti-OCT4 antibody (clone 3H8L6, Cat # 701756), mouse monoclonal anti-NANOG antibody (clone NNG-811, Cat # N3038), human/mouse/rat monoclonal anti-Ki-67 antibody (clone SolA15, Cat # 14–5698-82), goat anti-mouse IgG (H+L) secondary antibody, Alexa Fluor 488 (Cat # A28175), goat anti-rabbit IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor 568 (Cat # A11011), Purified rabbit isotype control IgG (cat # 02–6102) and mouse monoclonal anti-vimentin antibody (clone J144, Cat # MA3–745) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Thiol-polyethylene glycol-N-hydroxysuccinimide ester (HS-PEG-NHS, 5 kDa), and methoxy-polyethylene glycol thiol (mPEG-SH, 5 kDa) were purchased from NANOCS (New York, NY, USA). The Nanosep centrifugal filters (molecular weight cutoff (MWCO): 10 kDa) were purchased from VWR (Radnor, PA, USA). SU-8 photoresist and developer were obtained from Kayaku Advanced Materials, Inc (Westborough, MA, USA). Polydimethylsiloxane (PDMS) and its curing agent were obtained from Dow Corning (Midland, MI, USA). Translucent transfer capillary tubes were purchased from Cole-Parmer (Vernon Hills, IL, USA). All other materials were purchased from Sigma (St. Louis, MO, USA) unless otherwise specified.

Electro-microfluidic device design and fabrication.

The device was fabricated via standard photolithography and PDMS-based soft lithography as detailed previously,¹⁷ with modification. Briefly, a silicon wafer was patterned using SU-8 2025 photoresist following the procedures recommended by the manufacturer of the photoresist. The surface of a 4-inch silicon wafer was spin-coated with the photoresist that was then soft-baked, exposed to UV light through a patterned mask, developed with a SU-8 developer, and hard-baked to obtain a photoresist-based master containing a micropost array. The device is made of two sections: 1) the flow-distribution section which contains the inlet and flow channels, and 2) the main channel section which contains the negatives (i.e., cylindrical holes) of 100 rows and 32 columns of microposts. Each micropost is 100 μm in height and 100 μm in diameter. The space between two adjacent microposts is 50 μm. The obtained master/mold was rinsed thoroughly with deionized water and dried with nitrogen gas. The mold was placed and secured with tape in the center of a 15 cm (diameter) Petri dish. A mixture of PDMS and its curing agent (ratio of PDMS to curing agent in weight: 10:1) was poured onto the mold and degassed in a vacuum chamber for ~1 h. The bubble-free mixture was then cured/crosslinked in an oven at 75 °C for at least 2 h to form a PDMS elastomer that is patterned on the mold. The cured PDMS imprinted with the microfluidic channels was cut at the perimeter of the mold, peeled off slowly from the master, and placed onto a 4-inch silicon wafer before inserting it into a five-pocket Angstrom Engineering NexDep Ebeam Evaporator (Angstrom Engineering Inc, Ontario, Canada). The pressure inside the chamber was reduced to 5 X 10⁻⁶ torr before a 40 Å layer of titanium was deposited from a pure (99.99%) titanium target, followed by a 100 Å-thick Au layer deposited from a pure (99.99%) Au target. The thickness was controlled by an SQC-310 INFICON deposition controller (INFICON Inc., East Syracuse, NY, USA). The Au-coated device was removed and treated along with a microscope glass slide by an air plasma cleaner (Harrick Plasma, Ithaca, NY, USA) at 200 mTorr for 3 min right before binding the two together. Lastly, two copper wire electrodes were inserted into the two sidewall channels patterned in the device.

Sterilization:

To prevent any microbial contamination, Au-coated devices, syringes, needles, and capillaries were exposed to ultraviolet (UV) for 15 min before use. The surface modification steps, capture, and release processes were all conducted in an enclosed space sprayed with 70% ethanol right before every experiment.

Surface modification.

The capture medium, an aqueous solution (pH 7.4) with 10% (w/v) sucrose and 0.3% (w/v) glucose (electric conductivity: 1 mS m⁻¹) with a low viscosity (1 mPa s),¹⁷ was flowed through the main channel of the Au-coated device using a syringe pump (Harvard Apparatus, Holliston, MA, USA) at a flow rate of 1 ml h⁻¹ for 3 min to remove any residual air bubbles or dust. Antibody-linked PEG-thiol was prepared beforehand in an ice bath by reacting 20 μl of 0.5 mg ml⁻¹ CD44 or/and EpCAM with 20 μl of 1 mM HS-PEG-NHS in PBS at 4 °C overnight via the reactive ester coupling method. The free HS-PEG-NHS was removed by membrane filtration with a Nanosep filter (VWR, MWCO: 10 kDa) and centrifugation at 9,500 x g for 5 min. A total of 250 μl of 55 μg ml⁻¹ target-specific, antibody-linked PEG-thiol (CD44 or EpCAM) was flowed into the device at 0.5 ml h⁻¹ and further incubated in the main channel for 3 h at room temperature. For surface modification with a mixture of both CD44 and EpCAM antibodies, a total of 250 μl of a mixture of 55 μg ml⁻¹ CD44 and 55 μg ml⁻¹ EpCAM antibodies (1:1) were added into a syringe and flowed into the device at 250 μl h⁻¹ and further incubated for 3 h at room temperature. The main channel was then washed with the capture medium at 1 ml h⁻¹ for 5 min to remove unbound antibodies, followed by 30 min of incubation with 0.5 ml of 1.0 mM mPEG-SH to PEGylate the Au surface for blocking non-specific binding of the surface. The mPEG-SH was chosen over BSA for passivation due to its high binding affinity to the Au surface via the thiol group and the well-established ability of PEGylation to minimize nonspecific protein adsorption.

Scanning electron microscopy.

Scanning electron microscopy (SEM) was used to examine the surface of both the microchannels and microposts in the VIZA chip and the contact of captured cells on the microposts. Captured cells on the microposts were fixed in situ with 4% (w/v) paraformaldehyde (PFA, ThermoFisher Scientific) for 15 min, followed by the standard procedure of dehydration in ethanol and overnight drying for preparing SEM samples. After detaching the PDMS part of the device from the glass slide substrate using a razor blade, the specimens were mounted on the sample holders with a double-sided glue carbon tape and sputter-coated with 10 nm of gold using an Anatech Hummer XP sputter coater (Anatech, Sparks, NV, USA) at 800 V, 10 mA for 30 s, before placing the sample in the vacuum chamber of a Hitachi (Tokyo, Japan) SU-70 FEG SEM for imaging.

Culture and staining of cancer cell lines for spiking.

Cell culture media were purchased from Thermo Fisher Scientific unless otherwise specified. All cell lines were cultured in 75 cm² culture flasks in DMEM supplemented with 10% heat-inactivated FBS, 100 U ml⁻¹ penicillin, and 100 μg ml⁻¹ streptomycin solution at 37 °C in a humidified air with 5% CO₂ incubator. After reaching ~70–80% confluence, the cells were detached using 0.25% Trypsin- ethylenediaminetetraacetic acid (EDTA) (3 min at 37 °C and 5% CO₂) and neutralized with complete medium. For spiked cancer cell studies, cells were counted using a hemocytometer and stained with 5 µM of standard live/dead kit assay consisting of calcein-AM/PI for 15 minutes at 37 °C and 5% CO₂ before spiking them into the capture medium without or with WBCs (4 × 10⁶ cells ml⁻¹).

Suspension culture to obtain 3D mammospheres of MCF-7 cells.

MCF-7 cells were cultured in 6-well ULAP (Corning, Lowell, MA, USA) at a density of 20,000 cells ml⁻¹ in a serum-free defined medium made of DMEM/F12 supplemented with 1x B27, 20 ng ml⁻¹ basic fibroblast growth factor (bFGF), 20 ng ml⁻¹ recombinant epidermal growth factor (rEGF), 0.4% (wv⁻¹) bovine serum albumin, 5 μg ml⁻¹ insulin, 100 U ml⁻¹ penicillin, and 100 μg ml⁻¹ streptomycin. A total of 240 μl of the defined medium (fresh and warm) was added twice a week into each ULAP well. Spheroids were collected after 7 days into a sterile 1.5 ml centrifuge tube (Eppendorf, Hamburg, Germany) for further experiments.

Capture, release, and viability of spiked cells.

A total of 500 calcein-AM/PI stained MCF-7 cells were spiked into 1 ml of capture medium containing 4 x 10⁶ WBCs and then transferred into a 1 ml sterile plastic syringe (Becton Dickinson, Franklin Lakes, NJ). Subsequently, the sample-loaded syringe was placed on a programmable syringe pump (Harvard Apparatus) and connected to the sample inlet 2 of the VIZA chip via a capillary tube. Another syringe containing the capture medium (without cells) for washing was introduced into the device via inlet 1. Two copper electrodes were inserted into the space between the main channel and the sidewalls on both sides of the VIZA chip and connected to the amplified function generator (Agilent Technologies, Santa Clara, CA, USA) to apply the patterned (i.e., repeating pattern of on for 3 s and off for 3 s) electric voltage across the main channel in the device. The function generator was connected to the electrodes through a microcontroller with a switch that was connected to the VIZA chip to provide an AC electric peak-to-peak voltage (V_pp) of 200 V at a frequency of 200 kHz. These parameters were selected after preliminary optimization studies. The on and off time period of 3 s is optimal for moving the cells from the flow zone into the capture zone for capture when DEP is on. This time is also adequate to allow the cells to move back into the flow zone when DEP is off without clogging in the device.

Therefore, during cell capture, we programmed the microcontroller to turn on for 3 s and off for 3 s repeatedly, via the switch, allowing the application of the patterned electrical voltage across the main channel. The patterned electrical voltage was activated before the sample was pumped into the main channel of the VIZA chip at a flow rate of 1 ml h⁻¹. After all samples in the syringe was flowed out, the device was further flowed with the capture medium without any cells at 1 ml h⁻¹ for 5 min to push the residual sample in the connecting capillary tubes into the device and wash out unbound cells in the main channel. Captured cells in the main channel were either imaged/counted with a Zeiss (Oberkochen, Germany) LSM710 fluorescence microscope or released for culture/analysis (for cells without staining). Gently releasing the captured cells were achieved via a ligand exchange reaction by injecting GSH (1.5 mg ml⁻¹ in 1x PBS solution containing 1.5%wt BSA) into and filling the main channel of the device at 0.5 ml h⁻¹, followed by incubating at room temperature for 30 min. Capture medium was then flowed through the device and the released cells were collected into a petri dish for downstream analysis. Releasing of the captured cells with the conventional high flow rate was done by flowing the culture medium (DMEM) into the device at a flow rate of 8 ml h⁻¹ to dislodge the bound cells.

The capture efficiency was calculated as: (N_c / N_t) × 100%, where N_c is the number of spiked cancer cells that were captured in the main channel and N_t is the total number of spiked cancer cells. The capture efficiency for clinical samples was not reported since the total CTC number was not known beforehand. The release efficiency was calculated as: (N_r / N_c) × 100%, where N_r is the total number of released cells. The cell purity was calculated as: [N_c / (N_c + N_b)] × 100%, where N_b is the total number of captured blood cells (i.e., WBCs).

Cell viability was determined via live/dead staining by incubating cells with 5 µM of calcein AM and PI dyes for 15 min at 37 °C. The stained cells were imaged with the Zeiss LSM 710 microscope to count and evaluate their viability as: (C_l / C_t) × 100%, where C_l is the count of live cells (cells that appear green with no red color after being stained with the standard live/dead assay) and C_t is the total number of both dead cells (cells that appear red after being stained with standard live/dead assay) and live cells.

Procurement and preparation of patients’ blood samples.

Blood samples were obtained from 60 breast cancer patients (35 to 76 years old) who were scheduled for breast cancer follow-up or treatment. Blood samples were also procured from 7 healthy donors. The protocol for recruiting all human subjects was approved by the institutional review board (IRB) at the University of Maryland Greenebaum Comprehensive Cancer Center (UMGCCC). The IRB # was (HCR-HP-00058461) All specimens were collected in EDTA anticoagulant vacutainer tubes (Becton Dickinson, Franklin Lakes, NJ, USA). Red blood cells (RBCs) were removed from the blood samples using 1x RBC lysis buffer (Abcam, Cambridge, MA, USA) following the manufacturer’s protocol. Briefly, 1 ml of whole blood was incubated with 19 ml of 1x RBC lysis buffer at room temperature for 10 min, and then centrifuged at 400 x g for 5 min. The supernatant containing RBC debris was gently removed by pipetting. The pellet was resuspended in 1 ml of capture medium before use. Blood samples from healthy donors were prepared in the same way for control.

Immunostaining.

Captured CTCs/WBCs were fixed in situ with 4% (w/v) PFA for 15 min, permeabilized with 0.2% (w/v) Triton X-100 for 10 min, followed by washing with capture medium. The cells were further incubated for 1 h with 0.5 ml of 1.5% (w/v) BSA in PBST (1x PBS with 0.05% Tween 20) at room temperature to minimize non-specific binding. For CTC immunostaining, 250 μl of FITC-CEA in PBST with 1.5 % (w/v) BSA (1:25 dilution) and 250 μl of FITC-CK18 in PBST with 1.5 % (w/v) BSA in (1:50 dilution) were flowed into the device at 0.5 ml h⁻¹ and incubated in the main channel for 1 h followed by washing with capture medium through the device. For WBC immunostaining, 250 μl of Alexa Fluor-532 labeled anti-CD45 in 1.5% (w/v) BSA in PBST (1:25 dilution) was then flowed at 0.5 ml h⁻¹ and incubated in the main channel for 1 h followed by washing with capture medium. To counterstain the nuclei, 250μl of 4,6-diamidino-2-phenylindole (DAPI, 1.0 μM in 1x PBS) was flowed into the device at 0.5 ml h⁻¹, incubated for 15 min in the main channel, and washed with capture medium. All the incubation procedures were conducted in the dark at room temperature.

Culture of patients’ cells released from the VIZA chip.

Patients’ cells released from the VIZA chip were cultured not only under 2D in either DMEM or DMEM/F12 supplemented with 10% heat-inactivated FBS and 1% penicillin-streptomycin solution, but also in 3D suspension in 24-well ULAPs (Corning, Lowell, MA, USA). For the latter, the aforementioned serum-free defined medium was used. A total of 120 μl of the defined medium (fresh and warm) was added twice a week into each ULAP, and tumor spheroid formation and growth were monitored with a Zeiss PrimoVert inverted microscope. The SFE was calculated as the percentage of the total number of tumor spheroids formed out of the total number of live CTCs seeded. After the 3D spheroids reached or exceeded 100 μm in diameter after 35 days of culture, some spheroids were removed, for further studies: these were enzymatically dissociated with trypsin/EDTA into single cells. Obtained single cells were counted and cultured under either the aforementioned 3D suspension culture in the serum-free defined medium for multiple passages, or under the aforementioned 2D in either DMEM or DMEM/F12 supplemented with 10% heat-inactivated FBS and 1% penicillin-streptomycin solution. The rest of the spheroids were maintained in suspension culture for further monitoring up to 60 days.

Cryosectioning and immunocytochemistry.

To determine the expression of cell surface markers, CD44+ CTCs-derived tumor spheroids were collected by gravity sedimentation, fixed in 4% (w/v) PFA for at least 30 min followed by washing with 1x PBS. For dehydration, the spheroids were incubated in 20% (w/v) solution of sucrose overnight. They were then plated in a Sakura Finetek (Torrance, CA, USA) plastic box, embedded in Tissue-Tek Optimal Cutting Temperature (OCT) compound (Sakura Finetek), and kept at -80 °C for at least 5 min before cryosectioning. Ten micrometer-thick slices of the spheroids were obtained by cutting the frozen sample at -20 °C on a Leica (Wetzlar, Germany) CM1950 cryostat and then immediately adhering the slices to high adhesive glass slides for further processing. Slices were permeabilized with 0.2% (w/v) Triton X-100, followed by incubating with 1.5% BSA in PBST for 1 h to block nonspecific binding. This was followed by overnight incubation at 4 °C with respective primary antibodies: anti-mouse/human CD44 (1:100 dilution), anti-rabbit CD24 (1:200 dilution), anti-mouse/human EpCAM (1:100 dilution), purified anti-mouse vimentin monoclonal antibody (1:100 dilution), purified anti-rabbit OCT4 monoclonal antibody (1:100 dilution), purified anti-mouse/human NANOG monoclonal antibody (1:200 dilution), purified anti-mouse/human Ki-67 (1:200 dilution), purified anti-rabbit CD133 (1:100 dilution) and purified anti-mouse ALDH1A1 monoclonal antibody (1:100 dilution. Next day, unbound antibody was removed by washing the slices 3 times with 1x PBS and incubated at room temperature for 1 h with secondary antibodies: goat anti-rabbit IgG tagged with Alexa Fluor 568 diluted (1:200 dilution) and goat anti-mouse IgG tagged with Alexa Fluor 488 (1:200 dilution). All the antibodies were diluted in 1x PBST with 1.5 % BSA. Afterward, the slices were washed with 1x PBS and further stained for nuclei using DAPI (1 µM) for 15 min at room temperature. The cells were then covered with tissue mounting medium (CC/mount, Sigma) and cover glass for examination, followed by fluorescence imaging.

Flow cytometry analysis.

Tumor spheroids were collected by gravity sedimentation and enzymatically dissociated with trypsin/EDTA, followed by gentle pipetting to break up any cell clusters. The resultant cells were filtered through a 40 µm falcon cell strainer (VWR, Radnor, PA, USA) to ensure single cellularity and resuspended in ice-cold 1x PBS. After washing with 1x PBS, different centrifuge vials containing 1 x10⁶ cells were then stained with respective primary antibodies including anti-mouse/human CD44 (1:200 dilution), anti-mouse/human CD24 (1:200 dilution), purified anti-mouse/human EpCAM (1:200 dilution), purified anti-mouse vimentin monoclonal antibody (1:200 dilution), purified anti-rabbit OCT4 monoclonal antibody (1:200 dilution), purified anti-mouse/human NANOG monoclonal antibody (1:200 dilution), purified anti-mouse/human Ki-67 (1:20 dilution), purified anti-rabbit CD133 (1:200 dilution), purified anti-mouse ALDH1A1 monoclonal antibody (1:200 dilution), and incubated for 1 h at 4 °C. All the antibodies were diluted in 1x PBST with 1.5 % BSA. The samples were then washed three times with 1x PBS by centrifugation at 400 g for 5 min and resuspended in 200 μl of ice-cold 1x PBS. Secondary fluorochrome-labeled antibody (goat anti-mouse IgG (H+L), Alexa Fluor 488, 5 μg ml⁻¹, 1:200 dilution) or goat anti-rabbit IgG (H+L), Alexa Fluor 568 was added to each vial and incubated for 2 h in the dark at room temperature. The samples were then washed three times with 1x PBS by centrifugation at 400 g for 5 min and resuspended in 200 μl of ice-cold 1x PBS. Flow cytometry analysis was performed on a BD (Franklin Lakes, NJ, USA) FACSCelesta flow cytometer, and FlowJo (Ashland, OR, USA) software was used for data analysis.

Image acquisition and CTC enumeration.

All immunofluorescence images were acquired with a Zeiss LSM 710 fluorescence/confocal microscope and images were captured after fluorescence staining. All other images and movies were taken using a Zeiss PrimoVert inverted microscope with a FLIR Grasshopper 3 color camera for real-time imaging using the Zeiss Zen Blue software. All images were collected with the same exposure time for consistency and image data analysis was performed using the NIH (Bethesda, MD, USA) ImageJ software. CTC enumeration was performed manually by counting all captured cells in the main channel. The (CEA+/CK18+)CD45- cells were identified as CTCs, while CD45+ cells were identified as WBCs.

Statistical analysis.

All statistical analyses were performed using Microsoft (Seattle, WA, USA) Excel 2010. Data are reported as mean ± standard deviation (s.d.) from at least 3 independent runs unless otherwise stated. The statistical significance in mean values between two different cohorts was analyzed using Microsoft Excel based on Student’s two-tailed t-test assuming equal variance with no samples being excluded from the analysis. Unpaired t-tests were performed, except when comparing different ways (i.e., EpCAM alone, CD44 alone, and combination of EpCAM and CD44) of capturing CTCs from the same patients, for which paired t-tests were done. The differences were considered statistically significant when the p value is less than 0.05 (***p < 0.001, **p < 0.01, and *p < 0.05).

Supplementary Material

Supporting information

Figure S1: Schematic illustrations of the VIZA Chip and the virtual capture versus flow zones together with scanning electron microscopy (SEM) images of the microfluidic channels. Figure S2: The experimental setup. Figure S3: Characterization of the VIZA chip with cancer cells spiked in capture medium. Figure S4: Regression analysis of capture efficiency for various concentrations of spiked MCF-7 cells. Figure S5: The integrated VIZA chip consisting of eight of the unit devices connected in parallel for efficient capture and controlled release of cancer cells. Figure S6: Characterization of two cell sub-populations in MCF-7 mammospheres with flow cytometry. Figure S7: Immunofluorescence micrographs of CTCs captured from a breast cancer patient 19 by CD44 capturing antibody. Figure S8: In vitro culture of CTCs captured with EpCAM alone. Figure S9: Photomicrographs showing the fusion of multiple spheroids grown from CTCs. Figure S10. Photomicrographs of CTCs under 2D culture for 30 days. Table S1: A summary of methods used for the release of captured cancer cells. Table S2: The code, stage, subtype, race, and age of breast cancer patients from whom blood samples were procured in this work.

Movie S1

Movie S1: A movie showing MCF-7 cancer cells moving with capture medium in the virtual flow zone in the main channel of a microfluidic chip surface-modified with EpCAM antibody when no electric field is applied.

Movie S2

Movie S2: A movie showing MCF-7 cancer cells being moved from the virtual flow zone into the virtual capture zone in the main channel of a microfluidic chip with no antibody modification on the gold surface when an electric field is applied (DEP on) and their return from the virtual capture zone back into the virtual flow zone when the electric field is removed (DEP off).

Movie S3

Movie S3: A movie showing MCF-7 cancer cells being moved from the virtual flow zone into the virtual capture zone and captured there in the main channel of a microfluidic chip surface-modified with capture antibody when an electric field is applied (DEP on).