An integrated microfluidic platform for negative selection and enrichment of cancer cells⁶

A schematic illustration and a photograph of the CTC negative selection and enrichment chip are shown in figure 1(a). Dimensions of the microfluidic chip were measured to be 45   ×   25 mm (length  ×  width). The chip integrates two different sizes of circular membrane micropumps, including a larger micromixer (diameter: 5 mm, fluidic chamber depth: 150 μm, membrane thickness: 150 μm) and a smaller micropump (diameter: 2 mm, fluidic chamber depth: 150 μm, membrane thickness: 150 μm), and several normally-closed microvalves to control liquid transport and mixing. In addition, the chip contains three reservoirs for storing cell mixture, magnetic beads, and wash buffer, respectively. Electromagnetic valve (EMV)-controlled pressure and vacuum sources were connected to the device via eight air holes to control the actuation of the micromixer, the micropump, and the microvalves. The waste reservoir was connected to a separate vacuum source to collect the waste throughout the process of CTC negative selection and enrichment. An external magnet was placed beneath the micromixer chamber to trap cell-bead complexes, which are then removed during the washing step with the external magnet also removed.

Zoom In Zoom Out Reset image size

An exploded view of the chip is illustrated in figure S1 (stacks.iop.org/JMM/25/084007/mmedia). The chip was composed of an air control layer (upper layer), a liquid transport layer (bottom layer) and a glass substrate. In brief, master molds of microstructures on polymethylmethacrylate (PMMA) plates were first formed by using a computer-numerical-control (CNC) machine (EGX-400, Rolan Inc., Japan) equipped with a 0.5 mm drill bit. Then a polydimethylsiloxane (PDMS) replica-molding process was performed to fabricate the air control layer and the fluidic channel layer. The PDMS layers were first bonded to each other following oxygen plasma treatment (Plasma Cleaning Model CUTE MP-R, UVOTECH Systems, USA), and the bonded PDMS layers were finally bonded to the glass substrate following another oxygen plasma treatment to form the complete microfluidic chip.

In order to pneumatically drive the micromixer, the micropump, and the microvalves on the chip, a pressure source (~25 kPa) and a vacuum source (~90 kPa) were connected to a set of EMVs (S070M-5BG-32, SMC Inc., Japan), which were then connected to the air control holes on the chip via the tubing. Pressure was applied at the end of the mixing sequence to push the membrane down to the bottom glass substrate and squeeze any remaining liquid out of the micromixer chamber to minimize the sample liquid loss in the microfluidic chip (figure 1(b)). Actuation sequences of the micro-components to pump, mix, and transport liquid samples were achieved by controlling the opening and closing of each EMV with a programmable control circuit [18]. The programmble actuation of these micro-components enabled automated liquid handling, thereby obviating manual operation during negative selection and enrichment of CTC.

Figure 2 illustrates the experimental procedure implemented on the CTC negative selection and enrichment chip. Prior to the experiment, a cell mixture containing lymphocytes and cancer cells, anti-human CD45 antibodies-coated magnetic beads, and a wash buffer (Ca²⁺ and Mg²⁺ free PBS, pH 7.4) were loaded into their designated reservoirs. To initiate the experiment, 20 μl of magnetic beads were transported by the micropump to the cell reservoir (figure 2(a)). Cells and beads were mixed gently in the micromixer chamber for 10 min to facilitate cell-bead binding (figure 2(b)). Subsequently, an external permanent magnet was placed underneath the micromixer to capture the cell-bead complexes (figure 2(c)), while unbound cells remaining in the cell suspension were transported back to the cell reservoir (figure 2(d)). Because a significant portion of background cells were depleted from the original cell suspension, target cancer cells in the suspension became enriched. Before starting another round of depletion, the magnet was removed and the cell-bead complexes in the micromixer chamber were washed away by continuously pumping 1 ml of the wash buffer through the chamber (figure 2(e)). Finally, multiple rounds of lymphocyte depletion could be accomplished by repeating the steps 2(a) through 2(e), which were performed in an automated fashion.

Zoom In Zoom Out Reset image size

Colorectal cancer is one of the most common cancers in the world, and thus we chose human colon carcinoma cell line HCT8 as the target rare cells (i.e. model CTCs). Immortalized human T lymphocytes Jurkat served as non-target background cells that would be depleted during the negative enrichment process. HCT8 and Jurkat cells were both cultured in RPMI 1640 medium (Life Technologies, USA) containing 10% fetal bovine serum (Life Technologies) and incubated at 37 °C with 5% CO₂ in air. In this work, Dynabeads^® CD45 (Life Technologies) were used to capture and deplete lymphocytes (i.e. Jurkat cells). Per manufacturer's instructions, an isolation buffer (Ca²⁺ and Mg²⁺ free PBS supplemented with 0.1% bovine serum albumin (BSA) and 2 mM EDTA, pH 7.4) was used throughout this work for the binding between CD45⁺ cells and anti-human CD45 antibody-coated magnetic beads. The bead concentration was adjusted to 3.5   ×   10⁸ per ml prior to use.

To quantify the number of rare cells in our model sample, HCT8 cells were stained with 0.1 M CellTrace^TM Calcein Green AM (Life Technologies), counted using a hemacytometer, and re-suspended in the isolation buffer at 10⁴ cells ml⁻¹. The number of HCT8 in a constant volume (10 μl) was again counted under a fluorescent microscope (Bx43, Olympus, Japan) to verify that the cell suspension contained ~100 cells before the cell suspension was added to 10⁶ Jurkat cells. The volume of the final cell suspension containing both types of cells was 120 μl.

To evaluate the volume loss in the micromixer, 120 μl of water was first weighed and loaded into the cell reservoir. After actuating the micromixer at a pulsation frequency of 0.5 Hz for 5 min, the water sample was collected from the cell reservoir and weighed again. The weights before and after were used to determine the volume loss. The viability of cells after they were subjected to microfluidic mixing at various pumping frequencies was also assessed. In each experiment, 10⁶ HCT8 cells were mixed in the micromixer operating at a particular pumping frequency such as 0 Hz, 0.05 Hz, and 0.5 Hz for 15 min. The cells were then retrieved from the chip and examined by Trypan blue staining.

The depletion efficiency of the CD45⁺ cells on the benchtop and in the microfluidic chip were both investigated in the study. The conventional CD45 depletion on the benchtop was performed in accordance with the manufacturer's instructions. In brief, 10⁶ Jurkat cells and 7   ×   10⁶ magnetic beads were incubated in a 1 mL isolation buffer for 30 min at 4° C with gentle tilting and rotation. Using a benchtop magnetic particle separator (MPC, Dynabeads® MPC^®-1, Life Technologies), the supernatant was collected and the unbound cells remaining in the supernatant were counted by an auto cell counter (TC10, Bio-Rad, USA). On the other hand, on-chip CD45 depletion was performed using the same amount of cells and magnetic beads, but the total reaction volume was decreased to 140 μl. Cells and beads were incubated by actuating the micromixer at a pulsation frequency of 0.5 Hz for 10 min under room temperature. Using an external permanent magnet, the cell-bead complexes were captured and the unbound cells remaining in the cell suspension were transported back to the cell reservoir.

The depletion efficiency of the CD45⁺ cells was calculated as follows:

$$\begin{eqnarray}&&\frac{\text{Input}\text{ Jurkat}-\text{Output}\text{ Jurkat}}{\text{Input}\text{ Jurkat}}\times 100\%\text{.}\end{eqnarray} \tag{ 1 }$$

CTCs recovered from the supernatant after negative selection and enrichment were observed by fluorescent microscope. The recovery rate of CTC was calculated as follows:

$$\begin{eqnarray}&&\frac{\text{Output}\text{ HCT}8}{\text{Input}\text{ HCT}8}\times 100\%\text{.}\end{eqnarray} \tag{ 2 }$$

Notably, the large numbers of input and output Jurkat cells that were counted by the auto cell counter afforded a precise calculation of the depletion efficiency (i.e. many significant figures). On the other hand, the small number of input and output HCT8 cells, which were counted manually via fluorescence microscopy, resulted in a recovery rate with only a few significant figures.

Because the number of CTCs is low, it is imperative for the microfluidic chip to have negligible loss of any sample liquid that may contain CTCs during the enrichment process. The performance in liquid loss was verified by measuring the weight of the liquid samples before loading into the chip and after micromixing and recovering from the chip. Figure 3(a) shows that the liquid weight recovered from the chip was similar to the input weight, suggesting that, under the appropriate membrane actuation, the microfluidic chip has negligible liquid loss.

Zoom In Zoom Out Reset image size

To avoid potential cell damage and even death by shear forces generated from membrane actuation in the mixing step, different pulsation frequencies of the membrane micromixer were tested. To do so, 10⁶ HCT8 cells were first subjected to micromixer actuation at a particular frequency for 15 min. The cells were then retrieved from the chip and examined after Trypan blue staining. As shown in figure 3(b), the cell viability was 89.5   ±   0.7% at a pulsation frequency of 0 Hz (i.e. no membrane actuation), which was similar to the cells placed on the benchtop for the same time (90.5   ±   0.7%). This suggests that cells stay alive while entering the device and in contact with PDMS. Meanwhile, a comparable 90.0   ±   1.4% of cells remained viable under a pulsation frequency of 0.05 Hz. However, the cell viability dramatically decreased to 58.0   ±   1.4% when the micromixer was operated at a 0.5 Hz pulsation frequency. These results suggest that actuating the membrane micromixer at 0.05 Hz, which still achieved effective mixing (figure S2 (stacks.iop.org/JMM/25/084007/mmedia)), was gentle enough to maintain cell viability.

In order to maximize the success of negative selection and the enrichment of CTCs, an efficient depletion of CD45⁺ cells must be achieved. As a parameter to optimize the depletion of CD45⁺ cells, the optimal ratio between magnetic beads coated with anti-human CD45 antibodies and CD45⁺ cells was measured. Here, a cell suspension containing 10⁶ of Jurkat cells and the antibody-coated magnetic beads were loaded into their designated reservoirs in advance. The magnetic beads were transported to the cell reservoir and mixed with cells by activating the micromixer at 0.05 Hz for 15 min. After mixing and incubation, an external permanent magnet was placed underneath the micromixer to capture the cell-bead complexes, while the cells remaining in the suspension were transported back into the cell reservoir and recovered. The cell-bead complexes in the micromixer were washed away. The recovered cells in suspension were counted and the cell depletion efficiency was calculated using equation (1). The results for the different bead-to-cell ratios under this single round cell depletion are shown in figure 4(a). At a low, 2:1 bead-to-cell ratio, only 40.71   ±   6.18% of the original CD45⁺ cell population was depleted. Increasing the bead-to-cell ratio improved the depletion efficiency. For example, when the bead-to-cell ratio was increased to 6:1, a significantly higher 75.13   ±   4.42% of CD45⁺ cells was depleted from the original cell suspension. Further increasing the ratio to 7:1 resulted in a small improvement in the depletion efficiency (79.70   ±   2.07%) without significant difference to the ratio of 6:1, suggesting that the depletion efficiency had likely plateaued. This 7:1 ratio was therefore used in all subsequent experiments.

Zoom In Zoom Out Reset image size

In comparison to depleting the CD45⁺ cells in the conventional benchtop set-up, the efficient micromixing of cells and antibody-coated beads in the microfluidic chip facilitated more efficient binding and hence shortened the required incubation time for a single round of cell depletion. For instance, as shown in figure 4(b), 80.88   ±   1.35% of the CD45⁺ cells were depleted after incubating the sample with a benchtop rotator for an optimal incubation time of 30 min. In contrast, on-chip mixing for only 10 and 15 min led to comparable depletion efficiencies of 78.81   ±   6.40% and 81.08   ±   1.60%, respectively. Of note, the depletion efficiency dramatically decreased to 26.49   ±   0.35% when mixing for 8 min, suggesting an insufficient incubation time. On-chip CD45 depletion efficiency at room temperature or 4 °C was not significantly different (data not shown). Based on these results, the employment of the microfluidic chip at room temperature could shorten the required time for a single round of cell depletion from 30 min to 10 min, or a 67% decrease. The complete on-chip negative selection and enrichment process could be done in ~17 min, while the conventional method needed ~40 min, reducing the time required by 58% and avoiding labor-intensive work.

Similar to a previous demonstration in a benchtop set-up, which showed that the depletion efficiency of CD45⁺ cells in the negative enrichment approach could be increased by performing multiple rounds of cell depletion [15], the microfluidic chip in the current work was also capable of improving the depletion efficiency via multi-round depletion. In microfluidic on-chip multi-round depletion, after transporting the cell suspension that contains unbound CD45⁺ cells to the cell reservoir and washing cell-bead complexes from the micromixer chamber, a new round of depletion was initiated by transporting fresh beads to the cell reservoir and mixing with CD45⁺ cells remaining in the cell suspension. The results for multi-round depletion are shown in figure 5. Compared to single depletion, which achieved a depletion efficiency of 83.99   ±   1.00%, two rounds of depletion yielded a significantly improved depletion efficiency of 98.15   ±   0.28%. Three rounds of depletion further enhanced the performance of 99.84   ±   0.04%. However, the depletion efficiency appeared to have plateaued after three rounds of depletion, as four rounds and five rounds of depletion yielded 99.68   ±   0.05% and 99.64   ±   0.11%, respectively, which were comparable to that from three rounds of depletion. Importantly, performing multiple rounds of cell depletion in microfluidic chips was precise and automated. Carrying out this repetitious operation in microfluidic chips presents a significant advantage over a similar operation in a benchtop set-up, which often involves manual steps, and could therefore become cumbersome and error-prone.

Zoom In Zoom Out Reset image size

To evaluate the CTC recovery rate after multi-round depletion, one hundred fluorescently-stained HCT8 cells were spiked into a 10⁶ CD45⁺ cell suspension. Multiple rounds of negative selection and enrichment were performed, the recovered CTCs in suspension were counted, and the recovery rate was calculated using equation (2). The results for multi-round cell depletion are shown in figure 6. After a single round of CD45 depletion, 70   ±   5% of HCT8 cells remained in the cell suspension. The recovery rate decreased to 53   ±   4% after repeated depletion, while after three rounds of depletion, 32   ±   3% of the spiked HCT8 cells were recovered. Taking the CD45 depletion results and the HCT8 cell recovery results together, the study suggested that the chip could indeed enrich HCT8 cells—and hence CTCs in general—by depleting CD45⁺ cells. Indeed, a single round of negative enrichment resulted in 70% recovery of 100 HCT8 cells and 83.99% depletion of 10⁶ CD45⁺ cells. Of note, although the enrichment of HCT8 cells after two rounds of negative selection (53% recovery of 100 HTC8 cells and 98.15% depletion of 10⁶ CD45⁺ cells) compared less favorably to three rounds of negative selection (32% recovery of 100 HTC8 cells and 99.84% depletion of 10⁶ CD45⁺ cells), the higher recovery could still render this condition more favorable, especially when downstream analyses demand a higher number of target cells.

Zoom In Zoom Out Reset image size