The novel centrifugal microfluidic detection device is created by the LoaD, magnet, motor, and control system. Easy operation, lower cost, and higher performance are our key points of novelty. However, it is a challenge to how to control the microfluidic steady flow in the runner of the LoaD (Lin, 2011; Acharya et al., 2023). The rotating device of the motor and the pattern of the LoaD are key points. Therefore, we use a brush motor and servomotor to compare the steady rotating issue for the LoaD. Figure 1 represents the system compositions for the brush motor and the servomotor; their specifications are illustrated in Table 1. The brush motor system components are the relay, speed controller, PLC, encoder, and computer. The servomotor system components are the servomotor, signal converter, speed controller, PLC, and computer.

Lin Y. T. et al. (2021) used CFD software to design and simulate a novel LoaD. The detail specifications and pattern design of the LoaD are illustrated in Figure 2. The novel LoaD can control the micro-particles to flow, stay, and reflow in the runner of the LoaD by the different rotating mode. Therefore, we used the material of the PMMA to process the novel pattern of the LoaD by a CNC machine. The purpose was to test the performance of the microfluidic flow and control in this pattern of the LoaD. We used 95% EtOH and deionized water to wash the pattern of the LoaD and dry it after CNC processing. Then, we used cover and pressure-sensitive adhesive to combine the LoaD (Kuo et al., 2023; Zhang et al., 2018).

Human breast cancer cells MDA-MB-231 (ATCC, Cat# HTB-26) and human noncancer cells HEK293 (ATCC, Cat# CRL-1573) were cultured in a separate culture dish in DMEM-high glucose (Gibco, Cat# 12100046) with 10% fetal bovine serum (Gibco) and 1% penicillin–streptomycin (Gibco). The day before the experiment, MDA-MB-231 and HEK293 were stained with Mitotracker Red (Thermo Fisher Scientific, Cat# M7512) and Mitotracker Green (Thermo Fisher Scientific, Cat# M7514) at 37 °C overnight, respectively. The next day, the cells were washed with PBS and gently scraped off the dish with a cell scraper. They were mixed in different ratios, and the buffer well in each of the four micro-runners was loaded with 120 µL culture media containing 2 × 10⁴ cells. The sample well in the four micro-runners was loaded with 30 µL human CD44 antibody–magnetic bead conjugates (Miltenyi Biotec). We then used two different rotating modes to label and sort the cells by the centrifugal device. The first rotating mode was to let the microfluids flow to the reaction area from the buffer and sample well. The purpose was to let the labeling cells react and combine with the human CD44 antibody–magnetic bead conjugates. Thence, the microfluids of the reaction area flowed to the reservoir area by the second rotating mode. At this time, the labeling cells were adsorbed at the capture well, and the non-labeling cells flowed to the reservoir area. The experimental results achieved the function of sorting between labeling and non-labeling cells.

To image the cells, the LoaD was placed on the stage of a fluorescence microscope (Nexcope, NIB610) (Figure 3). The cells in the reservoir area were imaged first, and the fluid was immediately collected for subsequent flow cytometric analysis (BD Acuri C6 Plus, BD Science). The magnets were gently removed from the back of the LoaD before the cells retained at the magnetic area were imaged. The cells were also subjected to flow cytometric analysis to measure the cells retained at the magnetic area. For all flow cytometric experiments, cells collected from either the magnetic or reservoir areas were resuspended in 1 mL saline buffer, and the first 1000 cells were counted.

Since the LoaD in the present study relied on a combination of microchannel geometry and centrifugation to separate cells within a biological sample, it is essential that the disk maintain its balance and spin at a constant speed over some time without falling off the motor shaft. In addition, the lower cost and higher performance of the centrifugal device are our novel target. Therefore, we tested the rotating performance of the DC motor and servo motor devices (Figures 1a, b). Figure 4 shows the experimental results. Figure 4a shows the relationship of the rotating speed and time when the DC motor is driving at 100, 200, 300, 400, and 500 rpm and 30 s. We found the difference of the maximum and minimum speeds to be 7 rpm when the rotating speed of the DC motor device was set at 100 rpm. The difference of the maximum and minimum speeds was 6 rpm when the rotating speed of the DC motor device was 200 rpm. The difference of the maximum and minimum speeds was 9 rpm when the rotating speed of the DC motor device was 300 rpm. The difference between the maximum and minimum speeds was 8 rpm when the rotating speed of the DC motor device was 400 rpm. The difference of the maximum and minimum speeds was 6 rpm when the rotating speed of the DC motor device was 500 rpm. However, the difference of the maximum and minimum speeds could be controlled to within 1 rpm when the rotating speed of the servo motor device was between 100 and 500 rpm. The experimental results show that the servo motor device performs better than the DC motor device at the same rotating speed . According to Branderhorst (2024), the servomotor has the higher control performance, but its cost is higher than the DC motor.

Figure 4a shows that the DC motor system exhibits poor stability when the motor speed was 100 rpm. Therefore, we tested the performance of the motor system when the rotating speed was 50 and 75 rpm. The experimental results are shown in Figure 5. We found the difference to be about 9 rpm. when the rotating speed of the DC motor was 75 rpm. However, the difference of the servo motor was only 2 rpm. under the same rotating speed. The difference increased to 24 rpm when the DC motor’s rotating speed slowed to 50 rpm, and the difference of the servo motor was 16 rpm at the same rotating speed. We found that although the servo motor had better performance than the DC motor when the rotating speed was set below 100 rpm, the rotating speed stability of the servo motor became worse when the motor speed decreased. This is due to the relationship between the specifications of the motor torque and encoder pulses (Lu et al., 2018; Othman et al., 2018).

The geometric design of the microfluidic pattern will affect the performance of the cell sorting (Glass et al., 2012). In addition, Pishbin et al. (2020) showed that the rotation speed of the LoaD also affects sorting performance. Therefore, different rotating speeds and times may have different sorting performances under the same geometry design of the microfluidic pattern. Thus, we investigated the fluid flow behavior of microfluidics within 120 µL in the runner of the LoaD based on Figure 2 using the servomotor device. The experimental process also tested the optimal angle design of the LoaD’s mixing runner θ.

Figure 6a shows 120 µL Coomassie Brilliant Blue R-250 of microfluidics poured into the buffer well. We then drove the centrifugal device of the servomotor with 200∼250 rpm for 30∼50 min (Lin Y. T. et al., 2021). We tested difference angle of the mixing runner θ at 45°, 50°, 55°, 60°, 65°, 70°, 75°, 80°, 85°, and 90°. Each angle of the mixing runner was tested ten times. The results demonstrated that the microfluids of the buffer well are able to flow to the reaction area passing to the sample well and mixing runner (Figure 6b). We call this the first rotating mode of the centrifugal device. We again drove the centrifugal device of the servo motor at 200∼250 rpm and 20∼25 min. This was the second rotating mode of the centrifugal device. We found that the fluid of the reaction area is able to flow to the reservoir area, passing to the capture well in the second rotating mode of the centrifugal device (Figure 6c). The experimental results verify that the novel centrifugal device of the servo motor can control the fluid flow process with the same geometric pattern design of the LoaD by different rotating modes. The optimal rotating modes of the centrifugal device are in Table 2, and the optimal angle of the mixing runner design was within the range of 55°∼65°.

To demonstrate that our novel centrifugal device has the potential for clinical application, the purification efficiency of the disk was examined as in Figure 6. The buffer area was loaded with antibody–magnetic-bead conjugates recognized specifically for CD44 (CD44 microbeads) in the buffer area. CD44 is a cell receptor known for its role in cancer progression and is a diagnostic marker for early tumor detection (Yaghobi et al., 2021). The sample area was loaded with culture media that contained human breast cancer cells MDA-MB-231 and human non-cancer cells HEK293. CD44 receptors are found abundantly on MDA-MB-231 cell surfaces but not on HEK293 (Kim et al., 2021; Dinthe, 2023).

The MDA-MB-231 cells were pre-stained with Mitotracker Red, whereas HEK293 cells were pre-stained with Mitotracker Green. The two cell types were mixed in culture media and subjected to cell separation in the LoaD (Figure 6a). After being centrifuged counterclockwise for 20 min, the fluid in the buffer and sample areas were mixed at the reaction area (Figure 6b), with no trace of liquid in other compartments. The LoaD was removed from the motor shaft and then placed in a 4 °C incubator for 1 hour, allowing the CD44 microbeads to react with the MDA-MB-231 cancer cells. Afterward, the LoaD was centrifuged at 250 rpm for 25 minutes, and it was noticed that all the liquid was retained in the reservoir area with no trace of fluid in the other compartment (Figure 6c). It was expected that the microbead-bound MDA-MB-231 cells would be retained at the capture well (magnetic area) (Figure 2) while the rest of the biological components would end up at the reservoir area.

Figure 7 shows that the purification efficiency of the LoaD was examined with culture media containing different ratios of MDA-MB-231 to HEK293 cells. In the absence of any MDA-MB-231 cells, no cells were detected in the capture well either by fluorescence imaging or flow cytometric analysis, indicating the CD44 antibody–magnetic-bead conjugates were specific for cells that expressed CD44 (Figure 7a). When exposed to the conditioned media containing 1 × 10⁴ of MDA-MB-231—which represented 50% of the total cell population in the cell culture media—the LoaD managed to capture >99% of the cancer cells in that sample (Figure 7b). Moreover, the number of non-targeted HEK293 cells detected at the magnetic area was extremely low, as revealed by flow cytometric analysis. Likewise, in the culture media that had 0.5 × 10⁴, 0.2 × 10⁴, or 0.05 × 10⁴ MDA-MB-231 cells (equivalent to 25%, 1%, and 0%, 2.5% of the total cell populations, respectively), approximately 85% or greater of the cancer cells were retained at the magnetic area (Figures 7c–e). Additionally, the non-cancerous HEK293 cells were only detected at the reservoir area but not at the magnetic area, indicating that the LoaD has high purification capability.

Studies examining cell separations via magnetic microbeads and centrifugal microfluids have indicated that apart from microchannel geometry, the magnetic strength of magnets also affected separation efficiency (Ma et al., 2019). Assuming that all MDA-MB-231 cells had CD44 microbeads bound to their surfaces, it was suspected that the strength of the magnets installed beneath the magnetic area was not strong enough to capture all CD44 microbead-bound cells. A possible solution would be to increase the area of the magnet (Fu et al., 2016), thus increasing the chance of capturing and retaining the CD44 microbead-bound cells while the LoaD is being centrifuged. Alternatively, instead of increasing the size of the magnets, the magnetic strength could be increased by installing stacked magnets at the back of each four microchannels (Liu et al., 2022). Similarly, increasing the concentration of the CD44 microbeads could also improve the capture efficiency of the microbead-bound cells (Li et al., 2022). Nevertheless, the LoaD in the present study could still retain ∼85%–∼99% of the MDA-MB-231 cells at the capture well while being centrifuged at 250 rpm for 25 minutes. This indicated that the microchannel geometry in the LoaD with the Pattern II design can effectively purify cancer cells from a biological sample.

Large numbers of the non-target cells were also likely to play a role in inhibiting the interactions between the magnets and the CD44 microbead-bound cells. The lowest number of MDA-MB-231 cells that the LoaD exposed to in the present study was 0.01 × 10⁴ MDA-MB-231 cells—equivalent to 0.5% of the total cell number (Figure 7f). Even at such a low number, the LoaD was able to capture ∼75% of the MDA-MB-231 cells, as demonstrated by the flow cytometric data. This indicated that a high concentration of non-target cells could prevent the magnet from capturing the CD44 microbeads bound cells as the sample was spun from the mixing to the reservoir area.

Clinically, the average concentration of circulating tumor cells in a cancer patient’s blood is only one to ten cells per mL (Lin D. et al., 2021), whereas 1 mL of blood contains a few million white blood cells and a billion red blood cells (Yamamoto et al., 2020). Some studies have indicated that diluting biological samples before subjecting them to a diagnostic kit enhances the chance of targeted cells being detected (Galanzha et al., 2019; Endesfelder, 2019). Accordingly, a serial dilution on a biological sample would be needed to justify our LoaD, which could be used as a diagnostic tool for detecting an extremely low number of cells, such as circulating tumor cells in human blood.

Ideally, for any given cell separation technology, it would be best to separate and purify a specific biological target in a label-free manner. Although there have been some promising progress in the development of LoaD to enable the label-free isolation of target cells, most of the reported studies still rely on separation based on size (Zhu et al., 2020b). It had been demonstrated that LoaDs separating cells based on size had a higher chance of contaminating the targeted cells with unwanted cells than those utilizing an antibody-labeling method (Sunkara et al., 2021). Furthermore, the CD44 microbeads bound on the surfaces of MDA-MB-231 could be easily removed from the cell surfaces with a well-defined microbeads removal reagent. Such a reagent does not harm the cells, enabling them to be subjected to further analysis (Seitz et al., 2021).