Label-free Counting of Affinity-Enriched Circulating Tumor Cells (CTCs) using a Thermoplastic Micro-Coulter Counter (μCC)

Abstract

Coulter counters are used for counting particles and biological cells. Most Coulter counters are designed to analyze a sample without the ability to pre-process the sample prior to counting. For the analysis of rare cells, such as circulating tumor cells (CTCs), it is not uncommon to require enrichment before counting due to the modest throughput of μCCs and the high abundance of interfering cells, such as blood cells. We report a microfluidic-based Coulter Counter (μCC) fabricated using simple, low-cost techniques for counting rare cells that can be interfaced to sample pre- and/or post-processing units. In the current work, a microfluidic device for the affinity-based enrichment of CTCs from whole blood into a relatively small volume of ~10 μL was interfaced to the μCC to allow for exhaustive counting of single CTCs following release of the CTCs from the enrichment chip. When integrated to the CTC affinity enrichment chip, the μCC could count the CTCs without loss and the cells could be collected for downstream molecular profiling or culturing if required. The μCC sensor counting efficiency was >93% and inter-chip variability was ~1%.

Graphical abstract

A microfluidic-based micro-Coulter Counter (μCC) is developed with simple, low cost fabrication and easy assembly process. The μCC was integrated with a circulating tumor cell selection chip, providing the ability for label-free cell counting, which can maintain the cells integrity and viability for subsequent culturing or molecular profiling.

Introduction

Enrichment and enumeration of rare cells are important in a variety of applications, such as disease diagnosis or prognosis,¹ cell culture, hematological analyses, food safety screening, and environmental monitoring.^{2, 3} Enriching rare cells, such as circulating tumor cells (CTCs), and their enumeration is important because these cells can carry molecular signatures that can enable precision medicine by supplying indicators of disease through a liquid biopsy.⁴

Following enrichment, the cells must be enumerated, which can be done using flow cytometry (FC), or hemocytometry.⁵ FC is frequently used for cell enumeration and can provide high throughput processing.^{6, 7} However, FC relies on cell labeling with reporters that can affect cell viability. Also, it is difficult for FC to detect rare cells in mixed populations (sensitivity = 1 in 10⁴).^{8, 9} In addition, flow cytometers are expensive and require well-trained operators.¹⁰ Hemocytometry, while simpler instrumentally compared to FC, is labor-intensive and prone to errors due to operator bias.^{11, 12} Even though automated hemocytometers exist, they can generate high error rates due to incorrect identification of target cells, improper thresholding, averaging errors, and the assumption of a homogeneous cell distribution.¹³ Thus, these methods are not well suited to count low abundance cells.

Electrical-based sensing is an alternative approach, which can use either impedance sensing (impedance cytometer) or resistive pulse sensors (Coulter counters). For impedance counting, a single cell is shuttled through a pair of electrodes characterized by size and spacing comparable to those of the cell to be probed (~25 μm). When an AC field is applied to the electrodes, the cell transitioning between the electrodes will experience a change in its inter-electrode resistance.^14–19 Despite their ability of performing label-free measurements of single cells, the main limitation of impedance-based measurement devices is the cost associated with the microfabrication of the electrodes. In addition, the position of the electrodes with respect to their gap is critical to supply sufficient signal-to-noise ratio for detecting a wide range of particle/cell sizes. Finally, the position of the target within the sensing region is critical as this can induce variability in the signal. The use of hydrodynamic focusing was shown to reduce such variability.^{20, 21}

To obviate the need for microscale electrodes, Coulter counters (resistive pulse sensors) can be used, which allow for label-free and non-destructive electrical sensing.^22–25 In this method, an electric field is applied across a single tens of micron-sized aperture that separates two chambers. When single cells suspended in a conductive medium travel through the aperture, they generate a change in resistance through the aperture. When single cells suspended in a conductive medium travel through the aperture, they lead to a change in the resistance through the aperture produced by occupancy of a single cell within the sensing aperture volume.^26–28 The resultant resistance change is measured by macroscale electrodes placed at both sides of the aperture; the resistance change is proportional to the volume ratio between the cell and aperture.²⁹ The advantage of using macroscale electrodes is that their fabrication becomes much less demanding in terms of fabrication, lowering the cost of the device, which is important for in vitro diagnostic platforms that require disposable devices to eliminate sample-carryover artifacts. The single-cell counting process can provide information on cell size as well.³⁰

Use of impedance sensors or Coulter counters become problematic for scenarios in which rare cells must be enumerated especially when they are a vast minority in a mixed population. For example, circulating tumor cells (CTCs) are a vast minority in whole blood with the majority of cells being red blood cells and white blood cells (WBCs). Hence, the challenge is that both formats (impedance sensor or Coulter counter) have modest throughput when all nucleated cells from 1 mL of blood must be analyzed. Considering ~10 × 10⁶/mL WBCs and granulocytes, 5 × 10⁹/mL red blood cells and assuming impedance cytometry has a throughput of 20,000 cells s⁻¹ it would require ~70 h to process all cells in 1 mL of blood.¹⁵ Coulter counters have even lower throughput with the throughput proportional to the square of the sensing aperture diameter.

μCCs, however, have been used to count rare cells in blood samples. For example, a μCC was used to count rare cells pre-labeled with magnetic particles coated with antibodies specific to the target cell.^{31, 32} In one version of this magnetic bead labeling method, human umbilical vein endothelial cells were associated with anti-CD31 antibodies attached to 50 nm magnetic beads.³¹ Two μCCs were placed in series with a holding channel situated between the μCCs to allow retention of the magnetically labeled cells before counting. This magnetic holding step is basically an enrichment step (i.e., positive selection).

In the second example of this bead-based μCC method, A549 cancer cells were seeded into whole blood in which the red blood cells were lysed and the remaining WBCs labeled with anti-CD45 antibodies attached to magnetic beads – negative selection.³² The μCC was then used to count the A549 cancer cells. In this example, significant pre-processing of the blood was required, which can create loss of the rare cells prior to μCC counting.

Microfluidic technologies provide a path for assay step(s) integration to allow for automation of the sample processing pipeline, and reduce sample loss or contamination due to handling.^{33, 34} For example, microfluidic-based Coulter counters (μCC) can be interfaced to sample pre-processing microfluidic devices to provide the ability to prepare the sample prior to electrical counting in an automated fashion. These advantages potentially allow for implementation of cell counting in resource-limited areas for point-of-care testing, POCT.^{35, 36} For diagnostic applications, the μCC can be used as a disposable if the microfluidic can be produced using a low cost process and at high production rates as required for diagnostic applications.³⁷

Microfluidic devices have been developed to enrich CTCs from blood samples with high recovery and purity,^38–40 and provide a volume reduction with the removal of blood cells. Therefore, a CTC enrichment device coupled to a μCC could provide a platform for counting CTCs with high efficiency and short processing times. This system would allow for the enriched CTCs to be counted using a label-free method that would be conducive for analysis of the molecular content of the CTCs, such as phenotyping, gene expression, or sequencing.⁴¹

We have developed a microfluidic device for the efficient affinity-enrichment of CTCs, which consists an array of high-aspect ratio sinusoidally-shaped microchannels,⁴² capable of processing whole blood and enriching CTCs to a small volume (~10 μL). The CTC enrichment device, which is made from a plastic via replication technology, provides high recovery (>80%), high purity (>85%), and high throughput (1.5 mL/h).

Previously we reported the coupling of this CTC microfluidic enrichment device to a single cell impedance sensor.^{43, 44} The impedance sensor consisted of a microfluidic channel with a pair of Pt wires (75 μm diameter) inserted into guide channels positioned orthogonal to the fluidic channel. The wires were inserted manually into guide channels and possessed a gap of ~40 μm. After enrichment, the cells could be released from the capture surface of the microfluidic device to allow for counting using the impedance sensor. Due to the manual placement of the wires, however, a high production rate of the device at low cost was not realized. In addition, the success rate of producing operational devices was poor.

To overcome these drawbacks, we report a μCC sensor for rare cell enumeration following enrichment. Unique to this paper is that the μCC was made via a molding process in a thermoplastic that allows for high scale production at low cost as required for in vitro diagnostic tests. The operation of the μCC was characterized in terms of its detection efficiency and particle size sensitivity using polystyrene beads and cancer cells. The counting efficiencies were determined by comparing μCC readouts with visual microscopic inspection of beads or cells. Furthermore, the application of the μCC was demonstrated through its integration to a CTC enrichment device. CTCs were spiked into a healthy donor blood sample, affinity enriched, and subsequently released from the capture surface followed by counting using the μCC. The μCC was first prototyped using PDMS. Then, a plastic-based device that could be molded was demonstrated, which is conducive to high scale production at high process yield rates.

Experimental Methods

Fabrication of macroscale electrodes for PDMS device.

Ag/AgCl electrodes were patterned onto 1 mm thick microscope glass slides using stencil printing. First, the stencils with electrode patterns were cut from a film (Grafix, low tack 0.002 matte) using a desktop cutter (Silhouette CAMEO® 3). Stencils were adhered onto the glass surface and Ag/AgCl paste was uniformly spread over the stencil with excess paste removed using the edge of a glass slide. Electrode patterns were pre-baked for 1 min at 75°C. The stencil was then pulled off the surface of the glass slide and the electrodes were further baked at 75°C for 1 h. Removal of the stencil after a short pre-bake was determined to be critical for achieving well-defined electrode patterns with minimal shape defects and minimizing delamination of the thin Ag/AgCl electrodes from the glass slide.

Preparation of casting relief for replication of PDMS μCC.

SU-8 photoresist microstructures developed on a Si wafer were used as the relief for PDMS casting. The design of the microfluidic network used for the μCC is presented in Figure 1. It was composed of two fluidic chambers to accommodate the electrodes and a small channel, which served as the sensing aperture of the μCC connecting the chambers. A geometry consisting of two larger chambers connected via a small aperture is prone to stress cracking during photoresist development. To overcome this issue, the Si wafer was first coated with a continuous under-layer of SU-8 onto which another SU-8 layer was placed that contained the relief microstructures. The under-layer served two purposes: (i) Reduce mechanical stress between large and small patterns; and (ii) improve stability of the resist microstructures during casting eliminating issues with resist microstructure detachment from Si.

Figure 1.

(a) Schematic illustration of the assembly of μCC chips. (b) Overall image of μCC and zoom in of the aperture, with the microfluidic chamber filled with red dye. (c) 3D laser scanning microscopy of the aperture. (d) Optical microscopy images of a top-view and cross-section of the aperture in an unbonded PDMS substrate (~50 μm deep). (e) Screen printed Ag/AgCl electrodes and their thickness (~15 μm) measured by profilometry.

Standard SU-8 photoresist processing was used during relief fabrication. Briefly, a layer of SU-8 2002 photoresist was deposited onto a Si wafer using spin coating at 3,000 rpm for 30 s. Photoresist was soft baked at 70°C, 100°C, and 70°C for 1 min, 3 min, and 1 min, respectively. The photoresist layer was then flood exposed at 200 mJ/cm² (ABM UV Flood Source & Mask Alignment System, ABM-USA Inc., San Jose, CA). The post exposure baking was conducted at 70°C, 100°C, and 70°C for 1 min, 4 min, and 1 min, respectively, followed by a hard bake at 150°C for 5 min.

After preparation of the SU-8 under-layer, SU-8 2010 was spin coated at 250 rpm for 30 s and soft baked at 70°C, 100°C, and 70°C for 15 min, 4 h, and 15 min, respectively. The top layer of the SU-8 film was fly cut to a final film thickness of 60 μm and the wafer was baked at 65°C for 1 h. UV exposure through a dark field chrome mask was performed at 250 mJ/cm² using a contact mode. The post exposure baking was carried out at 70°C, 100°C, and 70°C for 4 min, 15 min, and 4 min, respectively. After cooling, the film was developed in 2 sets of fresh SU-8 developer for 7 min, and 2 min followed by an IPA rinse for 30 s, and hard baked at 150°C for 10 min to obtain the final microstructures, which served as the relief for PDMS casting.

Replication of PDMS μCC and assembly.

Prior to casting, the SU-8 relief was silanized. Trichloro(1H,1H,2H,2H-perfluorooctyl) silane (97%, Sigma-Aldrich, St. Louis, MO) was coated over the relief and placed in a desiccator, which was then evacuated to obtain a relative negative pressure in the chamber after which the chip was kept under vacuum for 4 h. For casting, PDMS and curing agent were thoroughly mixed at a weight ratio of 10:1 and vacuum was used to remove air bubbles. The SU-8 relief was placed onto a custom-made Plexiglas ring with an inner diameter of 10 mm. The PDMS mixture was poured onto the relief and cured at 70°C for 1 h. Finally, the PDMS device was peeled from the relief, and inlet/outlet holes were punched out.

Prior to assembly, both the glass slide patterned with electrodes and the PDMS substrate were washed with water and IPA and dried using a stream of compressed air. Both were exposed to O₂ plasma (Femto low-pressure plasma system, Diener Electronic) using optimized conditions (generator power: 50%; time: 1.5 min, pressure: 0.2 mbar). Following plasma treatment, both exposed surfaces were visually aligned and firmly pressed together for 1 min with a flat plastic plate to form the μCC. The electrode side of PDMS was further sealed with epoxy glue to avoid leakage (Figure 1a).

Fabrication of poly(methyl methacrylate), PMMA, μCC.

The PMMA μCC chip was fabricated using a previously reported method.⁴⁵ Briefly, the mold master was fabricated in brass via high-precision micromilling, HPMM (KERN MMP 2522, KERN Micro- und Feinwerktechnik GmbH & Co. KG; Germany). Polymer replicas of the mold master were produced via hot embossing into PMMA using a Precision Press model P3H-15-PLX (Wabash MPI, IN). Polymer plaques were dried in an oven at 65°C overnight prior to hot embossing. The embossing was performed at a force of 5 kN for 5 min at 155°C. After embossing, the microfluidic chips were cut out from the patterned PMMA wafers and 2 mm diameter access holes were drilled at each end of the microchannels. Microfluidic chips were then ultrasonicated in ∼0.5% Alconox solution for 3 min, rinsed with DI water, and dried in an oven at 65°C overnight. The size of the microchannels were the same as for the PDMS devices. The size of the aperture was designed to be 30 μm wide, 40 μm long, and 40 μm deep. One-hundred μm deep troughs were milled with a Carbide 3D desktop milling machine using a 1/32” milling bit into a PMMA cover plate; same size as the macro-electrodes patterned onto glass that was used for the PDMS device. However, this pattern could be generated on the PMMA cover plate using hot embossing as well to increase production rate. Ag/AgCl paste was deposited into the troughs in the PMMA cover plate, excess paste removed and finally dried for 3 h. PMMA substrate and cover plate with macro-electrodes were aligned under a microscope and thermally bonded at 100°C for 30 min. PEEK tubing was attached to the inlet/outlet ports of the device with epoxy glue.

μCC measurements.

The experimental setup used for characterization of the μCC is shown in Figure S2 and additional details on the operation are provided in the supplementary information (SI). The main components of the setup included custom designed electronics for applied voltage generation and data acquisition (NI DAQ-6009 card; 14-bit, 48 kHz, National Instruments, Austin, TX); PHD ULTRA™ syringe pump (Harvard Apparatus, Holliston, MA) for reagent/wash buffer delivery, and a 6-port manual rotary injector (Model No: 7725, Rheodyne) for sample introduction, which was used to simulate release of enriched cells from the CTC chip. Data acquisition was accomplished using software written in LabVIEW (National Instruments, Austin, TX) and data analysis, automated cell/particle counting, and data graphing accomplished using MATLAB (MATLAB 2014a, The MathWorks, Natick, MA. 2014) home-written scripts (see SI). Statistical data analysis was performed using RStudio v1.2.1335 and R v3.6.0. Shapiro-Wilk test of normality was used to determine possible Gaussian distribution of data sets and Wilcoxon rank sum and signed rank used to calculate the p-values to compare two data sets.

For μCC characterization, the setup was operated in a flow-injection mode in which the outlet of the syringe pump was connected to the μCC chip through a manual injector (Figure S2). The continuous flow of PBS (pH 7.4, 100 mM) was maintained by the syringe pump and cells or microbeads were introduced into the flow stream as 5 μL aliquots, transported through the μCC, and collected into a flat bottom well of a microtiter plate for microscopic evaluation. Resistive pulse signals were collected from the time of sample injection until the time when stable background without any peaks was observed. For integration with the CTC enrichment chip, the CTC chip was connected between the syringe pump and the μCC chip; the rotary injector was removed from the setup in these experiments.

CTC enrichment from whole blood.

The CTC enrichment chip was used for enriching CTC surrogates spiked in whole blood. Details of cell culture, staining, and CTC enrichment chip fabrication can be found in the SI. Prior to infusion of blood spiked with CTC surrogates, the CTC chip was thoroughly flushed with 2 mL of 0.5% BSA/PBS buffer at 55 μL/min. Healthy human blood seeded with SKBR3 or RPMI 8226 cells (pre-stained with Calcein-AM; see SI for more information on cell culturing and staining) was pumped through the pre-washed CTC enrichment chip at a flow rate of 25 μL/min using a syringe pump followed by a wash with 2 mL of 0.5 % BSA/PBS at 55 μL/min, which was required to clear the remaining blood from the CTC chip and to remove non-specifically bound cells. During the enrichment step, the CTC chip was kept connected to the μCC as shown in Figure S2. After washing, all cells from the CTC enrichment chip were stained with DAPI nuclear staining dye.

CTC release and counting.

After CTC surrogates were enriched using the CTC chip, it was infused with USER enzyme (4 U/10 μL, 1× Cut smart buffer; New England BioLabs Inc, Ipswich, MA), and kept at 37°C for 45 min. USER enzyme cleaves the single-stranded DNA heterobifunctional linker at the uracil residue, thus releasing the captured cells from the antibody-coated surface of the CTC chip. The optimal temperature for USER enzyme activity was maintained throughout the release process using a copper heating pad with a thin film resistive thermostat on the bottom of the chip that was controlled by a closed-loop PID (Proportional-Integral-Derivative) controller. A flow rate of 55 μL/min was applied to allow the released cells to flow through the μCC for counting.

Results and Discussion

Fabrication and assembly of PDMS μCC.

The PDMS μCC was assembled using two components: (1) A glass cover plate with screen-printed macro-electrodes; and (2) a PDMS microfluidic component containing the sensing aperture and access microchannels. The glass slide with screen-printed electrodes and PDMS substrate were assembled following O₂ plasma treatment to make the functional device (Figure 1a). The edges of the PDMS-glass chip were sealed with epoxy glue to prevent solution leakage between electrodes and the PDMS chip (Figure S1). The fluidic component contained one set of inlet/outlet holes, two hexagonally-shaped microfluidic channels under which 4 electrodes were positioned, and the sensing aperture, which allowed for resistive pulse sensing of single cells. (Figures 1b, c, d). The electrodes were screen printed onto the glass slide using a stencil and Ag/AgCl ink (Figure 1e). The major advantage of this design was the simplicity of generating the electrodes required for the measurement and their placement with respect to the access microchannels.

The metrology of the μCC was characterized using optical microscopy, profilometry, and 3D laser scanning confocal microscopy. An optical image of the μCC and a zoomed image of the sensing aperture are shown in Figure 1b, with the microfluidic device filled with a red dye for ease of visualization. The aperture of the μCC was examined closely with a 3D laser scanning confocal microscope of a fully assembled device (see Figure 1c). The cross-section of the aperture of a non-bonded device and its dimensions are shown in Figure 1d. The PDMS aperture produced from the SU-8 relief possessed a cuboidal shape with a length of 20 μm, width of 25 μm, and height of 50 μm. The dimensions of the sensing aperture provided sufficient space for particles to pass through with adequate detection sensitivity for transducing particles of ~10 μm diameter (see Eqs. 1 and 2). The sensing aperture maintained its designed geometry following bonding of the glass slide to the PDMS substrate as determined by scanning confocal microscopy interrogation of an assembled device. The screen-printed Ag/AgCl electrodes possessed a thickness <15 μm as determined by profilometry (see Figure 1e). The electrode thickness provided an intra-electrode conductance of 0.01 mS and did not block particles from flowing through the microfluidic channel.

Sensor operation principles and analytical figures-of-merit.

The experimental setup for the μCC sensor characterization and schematic illustration of the instrumental arrangement for μCC testing is shown in Figure S2. Briefly, a syringe pump was used to maintain flow through the sensor, and an inline 6-port manual rotary injector was used to introduce samples into the μCC. The μCC chip was connected to the electronics package, which was used to transfer data to the data acquisition card in the computer.

The μCC consisted of 2 pairs of electrodes. The outer electrodes were used for signal generation providing a ±7 V with a 50 kHz sine wave. Voltage changes due to a resistance change as particles passed through the aperture were measured using the 2 inner sensing electrodes (Figure 1b). For resistive pulse sensing, the aperture size with respect to the cells/particles to be counted is critical. It sets the change in aperture resistance ($\Delta R_D$) when a cell/particle travels through the aperture as noted in Eq. 1:

$$\Delta R_D=\raisebox{1ex}{$4{\rho }_m{d}_p^3$}\!\left/ \!\raisebox{-1ex}{$\pi D_r^4$}\right.$$ (1)

where ${\rho }_m$ is the solution permittivity, $d_p$ is the particle diameter, and $D_r$ is the diameter of the sensing aperture. As can be seen from Eq. 1, a larger pore diameter with respect to the particle size will reduce the change in resistance (i.e., resistive pulse signal) that is measured. The measurement scheme (4-electrode configuration, see Figures 2a and b) is accomplished by measuring the voltage ($V_D$) across the sensing aperture that has a fixed voltage applied across it ($V_0$) according to Eq. 2:

$$V_D=\raisebox{1ex}{$R_D\times V_0$}\!\left/ \!\raisebox{-1ex}{$2R_E+R_D$}\right.$$ (2)

where $R_E$ is the electrolyte resistance and $R_D$ is the aperture resistance.

Figure 2.

(a) Equivalent circuit diagram of the μCC. (b) Schematic representation of the signal change when a particle is passing through the μCC aperture. (c) Representative raw resistive pulse signals for 10, 15, and 20 μm beads as measured by the μCC operating at 20 μL/min volumetric flow rate. (d) Linear fit of the peak amplitude to the particle volume. Results are reported as an average of all peak amplitudes obtained for each bead size from measurements ±standard deviation, and distributions of particle volumes based on bead manufacturer information.

The μCC was calibrated using polystyrene beads with three different diameters (10 μm, 15 μm, and 20 μm) to evaluate the resistive pulse amplitude and width, and the detection efficiency as a function of particle size. For a single population of beads, the μCC detected signals with a consistent amplitude, which changed with bead size (Figure 2c). For beads with different diameters, the μCC showed similar peak widths corresponding to the same transit time and same detector sensing zone size.

The amplitude of the signal increased with increasing size of beads and showed a linear fit as a function of bead volume (Figure 2d). This result is in agreement with Eq. 1, which indicates that the resistance change measured by the μCC is proportional to the cube of the particle diameter. Based on the calibration plot of the μCC resistive pulse amplitude versus particle volume, the analytical sensitivity (ability to differentiate particles based on volume) and detection limit (smallest particles that can be detected) of the μCC could be calculated and used for obtaining size information for cell counting. Based on the background noise obtained from trace data and signal amplitudes, the smallest particle we could detect at S/N = 5 was 5.9 μm.

Cell enumeration and size determination.

SKBR3 and RPMI 8226 cell lines were used for characterization of the μCC. As shown in Figure 3a, signal amplitudes of SKBR3 cells measured by the μCC showed a broad distribution of resistive pulse amplitudes, which reflects the size distribution of this cell line.

Figure 3.

(a) Representative resistive pulse signals for SKBR3 cells measured by the μCC. (b) Histogram of the size distribution of SKBR3 cells obtained by the μCC versus optical microscopy (the diameter of cells was determined from the volume of the cell detected by the μCC). (c) Correlation of the cells counted by μCC with the cell counts obtained from optical microscopy. Different cell concentrations were tested by diluting cells in PBS buffer, pH 7.4. (d) Inter-chip reproducibility of the μCC was tested by connecting two μCCs in series and measuring the same SKBR3 cells.

The calibration plot acquired for the polystyrene bead data was used to determine the SKBR3 cells’ volume and calculated cell diameters. The size distribution of the SKBR3 cells measured by the μCC was correlated with the size determined using an optical microscope (Figure 3b). As shown in the overlay plot, the size distribution histograms acquired from these two methods matched well. No statistical difference between these two methods was observed (p = 0.2682). From the histograms, the average diameter of the SKBR3 cells was 16.5 ±3.5 μm and 16.5 ±4.1 μm as determined by the optical method and μCC, respectively.

To test the cell counting efficiency of the μCC, SKBR3 cells passing through the μCC were collected into a 96-well plate, stained with a nuclear stain (DAPI), and counted by fluorescence microscopy. Comparing the counts obtained from the μCC to that of fluorescence microscopy, the average counting efficiency of the SKBR3 cells was determined to be 93 ±5%. This experiment was also performed for fluorescent beads (d = 20 μm) resulting in a counting efficiency of 95 ±2%. Videos showing the μCC signal changing when beads or SKBR3 cells passed through the aperture can be found in the SI, which demonstrated good correlation between particles passing through the μCC and resistive pulse signal generation.

The counting efficiency of the μCC was further investigated by counting smaller diameter RPMI 8226 cells (average d = 9.5 μm) and comparing the counting efficiency with the SKBR3 cells. Concentrations of cell suspensions (0 – 500,000/mL) were prepared in PBS (pH = 7.4) and 5 μL from each suspension were introduced into the μCC via the rotary injector (see Experimental Section and Figure S2). The μCC counts were compared to the known injected cell number (Figure 3c) to determine counting efficiency from the slope of the μCC cell counts versus cells injected; the counting efficiency was 99% for the SKBR3 cells in this series of experiments and 94% for the RPMI 8226 cells. For about 6% of RPMI 8226 cells, the resistive pulse amplitude was not detected with S/N higher than 5 owing to the smaller size of these cells compared to the SKBR3 cells. However, good linearity was shown for both cell types between the frequency of response and the cell concentrations analyzed; the dynamic linear counting range for both cell lines was estimated to be from a single cell to 8 × 10⁶ cells/mL for a data acquisition rate of 20 kHz as used here.

μCC device-to-device reproducibility was examined by connecting two μCCs in series and counting the SKBR3 cells passing through both detectors (see Figure 3d and Figure S3). Different detectors showed very similar resistive pulse traces. For example, the average ±SD cell volume obtained from μCC #1 and #2 were 1.08 ± 0.68 V and 1.17± 0.86 V, respectively. The inter-device resistive pulse amplitude signal reproducibility was 99 ±2%.

Thermoplastic-based μCC.

The prototype μCC fabricated in glass/PDMS demonstrated utility in cell enumeration as shown in the above sections. The major challenge with PDMS devices is the inability to produce the finished device in a high-scale production mode and with high process yield rates. We therefore proceeded to demonstrate the ability to fabricate the μCC using a thermoplastic (i.e., PMMA) via replication to produce the fluidic network as well as the electrode pattern.

A mold master (Figure 4a) was milled via HPMM and used for embossing the μCC fluidic substrate in PMMA (Figure 4b). The aperture size was designed to have a width of 30 μm, length of 40 μm, and a depth of 40 μm. Ag/AgCl electrodes were embedded into a PMMA cover plate (Figure 4c) by generating 100 μm-deep troughs in the cover plate, which defined the electrode pattern. The finished μCC device (Figure 4d) was realized by thermally fusion bonding the embossed fluidic network containing inlet/outlet holes and the cover plate with electrodes. A micrograph of the μCC with the aperture is shown in Figure 4e.

Figure 4.

Fabrication steps for the PMMA-based μCC. (a) Brass mold fabricated via HPMM containing four μCC devices and a close up of the aperture in the milled brass. (b) μCC chips embossed in PMMA and a close up of one device showing the same architecture as in the PDMS devices. (c) Pictures of a PMMA cover plate (2 mm thick) in which the troughs were milled (100 μm deep) that were filled with Ag/AgCl paste. Silver paste was allowed to dry and excess paste was removed with a razor blade. (d) μCC was assembled by thermally bonding the cover plate containing electrodes with the embossed substrate and gluing inlet/outlet tubing. (e) Picture of the aperture between two inner electrodes following chip assembly. (f) Evaluation of the μCC response with a mixture of 10 and 20 μm beads. (g) Signal amplitudes for single SKBR3 cells flowing through the PMMA-based μCC.

Evaluation of the PMMA-based μCC was first performed with 10 μm and 20 μm beads and then, SKBR3 cells (Figure 4f and g, respectively). The amplitude of the resistive pulse signal for 10 μm and 20 μm beads were 0.5 ±0.15 V and 4.5 ±0.8 V, respectively. These values were used to determine the size of the SKBR3 cells (16.6 ±3.3 μm), which was similar to the values obtained from the PDMS device. Figure 4f, g present data analyzed by a Matlab script (see SI). Each detected cell event was marked and automatically counted.

Integration of μCC with CTC enrichment chip.

In our previous work, the identification of CTCs relied on phenotyping. CTCs were defined as nucleated cells showing positive staining for cytokeratins and negative for CD45 (i.e., leukocyte specific antigen). Based on our previous work, we determined that the purity of the isolated CTCs using our CTC chip was very high, owing to the high shear rates in the sinusoidal channels.⁴⁶ On average, only a few leukocytes (3–6 leukocytes per mL of blood analyzed) were co-isolated with the CTCs, which provided purities >85% for a variety of sample types. Due to the high purity of the isolated CTCs, we also demonstrated an impedance-based sensor for enumeration⁴⁷ and confirmed high concordance between CTCs enumerated via phenotyping and electrical counting. To simplify fabrication, we were motivated to fabricate the μCC via hot embossing.

Here, we demonstrate the integration of the μCC with the CTC enrichment device for counting CTCs isolated from whole blood with the requirement of no blood pre-processing that may result in loss of rare cells. Anti-EpCAM antibodies that cover the CTC enrichment chip surface were attached to the surface via a cleavable linker and the enriched cells were released with USER enzyme.⁴² Released cells were then counted using the μCC sensor. The experimental setup was similar to the one shown in Figure S2, with the rotary injector replaced with the CTC enrichment chip (Figure 5a) and the blood sample flowing serially through the CTC chip followed by the μCC sensor.

Figure 5.

(a) CTC isolation device connected with the μCC sensor. (b) μCC intra-chip reproducibility of the peak amplitude measured for the SKBR3 cell line before and after blood infusion through the μCC. (c) Comparison of CTC and leukocyte diameters obtained from commercial cell counter. (d) Comparison of the μCC signal amplitudes of CTCs and leukocytes. (e) Resistive pulse traces for the counting of CTCs using the μCC following enrichment from blood and release from the CTC chip. Insert: Calcein-AM stained released cells. Table summarizing results from 3 experiments for CTC concentrations of 50 – 250 CTC surrogates per 2 mL of blood.

We tested the stability and signal reproducibility of the μCC sensor before blood infusion and after 1 mL of blood was processed using the system (Figure 5b). Resistive pulse traces for SKBR3 cells before and after passing whole blood through the μCC showed no differences in their signal in terms of amplitudes and counting efficiency demonstrating that blood passing through the μCC during the CTC enrichment step did not foul the electrodes (Figure 5b) nor block the aperture (see SI for statistical analysis).

Even though our microfluidic chip is not prone to non-specific leukocyte adsorption when processing blood samples, we tested the detector response for leukocytes isolated with Histopaque-1119. We compared the signal amplitudes generated for both leukocytes and CTC surrogates (SKBR3). It was observed that the majority of leukocyte peak amplitudes were ≤0.4 V, while the majority of signals for the CTCs were >0.4 V. The signal amplitude overlap between leukocytes and CTC surrogates was ~10% (Figure 5 c, Figure 5d). By applying a threshold, we could efficiently eliminate leukocyte counts from CTC counts following enrichment. For example, when 100 leukocytes and 100 CTCs were co-injected and counted using the μCC, a threshold of 0.4 V eliminated counting of 94% of the leukocytes while CTC counts decreased by only 13%. This would mean that because of the high purity of our CTC isolate following enrichment, very few if any leukocytes will be counted.

As a proof-of-concept, the integrated system was tested with whole blood samples from healthy donors. Blood samples were spiked with pre-stained Calcein-AM CTC surrogates (i.e., SKBR3) and affinity selected using the CTC enrichment chip (see SI). Enriched CTCs were then released using USER enzyme and allowed to flow through the μCC chip for counting. As we have noted from our previous work using a pass through experiment in which 2 similar chips were placed in series with the number of CTCs captured on the first chip divided by the total number of cells captured on both chips, which provided an indication of the recovery of the CTC enrichment chip.⁴ For SKBR3 cells, we have shown that the recovery is 78%.

For verification of the μCC analytical performance, cells were deposited into a 96-well plate, imaged and then counted using fluorescence microscopy. The discrepancy between the μCC and microscopy counting was 9.5 ±9.6% indicating that >90% of the released CTCs from the enrichment device were transferred to the μCC. However, we believe that some of this discrepancy resulted from cell clusters as seen in the fluorescence image of the released cells (Figure 5E). A cell aggregate in the μCC will be counted as a single event with higher amplitude compared to a single cell. However, for microscopy an aggregate of cells are individually counted due to the nuclear staining and the high spatial resolution of the microscope. This can potentially be rectified in the future by noting high amplitude signals as potential aggregates when counting of` CTCs is performed, or applying an algorithm that will allow for the high amplitude signals to be de-convoluted into single cell events. In experiments where single CTCs were analyzed, we observed good correlation between the μCC and microscopic methods (see Table in Figure 5e).

Conclusion

We fabricated a μCC sensor and validated its ability to detect biological cells with diameters above 5.9 μm (S/N = 5). The μCC was calibrated with polystyrene beads and showed a linear response proportional to the particle volume. The sensor figures-of-merit were evaluated with SKBR3 (epithelial breast cancer CTC surrogate) and RPMI 8226 (multiple myeloma CTC surrogate) cell lines. μCC average counting efficiency was 95 ±2% for SKBR3 cells with inter-device reproducibility 99 ±2%. The size distribution of the SKBR3 cells measured by the μCC agreed well with the size determined by optical microscopy. For example, the average diameter of the SKBR3 cells was 16.5 μm ±3.5 μm and 16.5 ± 4.1 μm by hemocytometry and μCC, respectively.

We demonstrated fabrication of both PDMS and PMMA-based μCCs; the PMMA-based device offers high scale production capabilities in comparison with the PDMS device, especially when injection molding is used to produce the device.⁴⁸ The fabrication of the μCC is simple, does not require elaborate microfabrication steps, and devices can be made at low cost.

The μCC also provided the ability for rare cell counting without the need for labeling, and thus retaining the cell’s native state for subsequent analysis such as molecular profiling, FISH, or even cell culturing. In this case, the μCC was integrated to a CTC enrichment microfluidic device so that the low-throughput associated with the sensor could be accommodated because only enriched cells were required to be enumerated. The sensor could thus be used to enumerate rare cells affinity selected without the need for pre-processing of the blood sample that can result in rare cell loss. The μCC provided high accuracy of cell counting following enrichment with high sensor stability.

Future work will focus on developing an integrated CTC analysis system with the μCC unit as a part of a fully automated system; the system will be able to enrich CTCs from whole blood, count the CTCs following release, and molecularly profile the CTCs. Such a system will deliver a disposable, low-cost device, and molecular data on CTCs in a fully automated fashion, potentially at the point-of-care that will enable precision decisions on managing a variety of cancer-related diseases.

Abbreviations

μCC

micro-Coulter Counter

sulfo-SMCC

Sulfosuccinimidyl-4-(N-maleimidomethyl) cyclohexane-1-carboxylate

MES

2-(4-morpholino)-ethane sulfonic acid

PBS

Phosphate buffered saline

CTC

Circulating Tumor Cells

EDC

1-ethyl-3-[3-dimethylaminopropyl] carbodiimide hydrochloride

COC

cyclic olefin copolymer

FBS

Fetal Bovine Serum

PDMS

Polydimethylsiloxane