3.1 Design of microfluidic device
As shown in Fig. 1A, we designed and constructed a microfluidic operation platform, which was mainly composed of a concentration-gradient generator and 24 single-cell capture devices. The concentration-gradient generator is constituted by a fluid layer, thin polydimethylsiloxane (PDMS) layer, and a glass slide. The flow layer consists of 42 Tai Chi-spiral mixers (50 µm width; 50 µm height), 24 liquid storage chambers (1200 µm width; 2200 µm length; 50 µm height), 3 sample intakes and 24 exits. Three sample entrances were used to fill the source solution. The spiral micromixers were mainly used to diffuse and mix solutions from different sources, resulting in a series of successful drug concentration gradients in the liquid storage chamber. Then 24 outlets were connected to 24 single-cell capture devices, which laid the foundation for subsequent cellular resistance research and analysis. The concentration-gradient generator has very high flexibility, with a design of three inlets and a spread mixing effect of the Tai Chi-spiral mixers. Three concentration gradients can be formed to meet the requirements of a single drug concentration gradient and joint screening of two drugs. Therefore, it can be used to study the possibility of a single drug and multi-drug combination and its optimal dosage [37].
The single-cell capture device consists of five sets of capture matrix columns arranged into a two-dimensional array. Each capture matrix has 180–205 capture units (25µm height) for single-cell capture. As shown in Fig. 1B, each capture unit consists of two adjacent “H” shaped microstructures for single-cell capture. The two adjacent “H” structures constitute a single-cell capture structural unit. The microstructural capture unit has two minimum pores: the first pore is 2 µm wider than the second one. Due to the array differences, the spacing of capture unit is getting smaller and smaller. Cell and reagent portals are used for cell suspension and reagent import (Fig. 1C). Single-cell capture and drug stimulation can be achieved using the H-shaped capture structures. For other design of chip dimensions, it's described in detail in our previous publication [38].
3.2 Three concentration-gradient formation
To determine whether our device is capable of establishing three concentration gradients, numerical simulations and fluorescein experiments were performed to explore the distribution of drug gradients in 24 microcavities. The simulation results showed that three groups of identical drug concentration gradients were formed in the designed device (Fig. 2A). The successive Tai Chi-spiral mixer control and regulation can realize a stable mixing state. The accurate drug-concentration gradients were steadily fabricated in 24 liquid storage chambers (Fig. 2B). The concentration gradients of drug A were distributed in chamber 1–8 and 18–24; the concentration gradients of drug B were distributed in chamber 2–16; the concentration gradients of drug C were distributed in chamber 10–24. The simulation percentages of multiple drug-concentration gradients can be generated.
Furthermore, we used the fluorescein experimental evaluation to verify the distribution of concentration gradients in 24 liquid storage chambers at the same flow rate. Luciferin and PBS solution were injected into the chip from three entrances. As shown in Fig. 2C, a series of solutions containing different concentrations of luciferin were also generated in the liquid storage chambers. The experimental results are consistent with the numerical simulation results (Fig. 2D). More importantly, by comparing and analyzing the numerical simulation and fluorescein experiments under three flow conditions, it was found that the concentration difference of each liquid storage chamber was not obvious at different flow rates, indicating that our device could achieve excellent mixing performance. The average signal intensity obtained from the fluorescence images showed a good linear relationship with the expected numerical simulation concentration (Fig. 2E). These results demonstrated that three sets of stable and symmetrical concentration gradients could be completely constructed in a velocity-insensitive microfluidic system.
To further explain the reasons and verify the reliability of the above experiments, we chose the Tai Chi-spiral mixer used in the concentration gradient generator as the research object. Many microfluidic applications of fluid manipulation have been developed using integration advantage of Dean flow in curving channels [40–42]. The Dean flow simulation for the two sections of the spiral mixers (Fig. 3A) was carried out. Under different flow conditions, the Dean flow field distribution changes on the upper and lower sides of the spiral mixer channel (Fig. 3B, 3C and 3D). As the flow rate increased, the Dean flow intensity in the fluid velocity field gradually increased (Fig. 3E), and the resulting Dean flow could enhance the solution mixing effect in a relatively short period of time. The above results showed that the concentration gradient generator had adequate fluid mixing and uniform fluid splitting at a wide range of flow rates. The stable and effective multi-drug concentration gradient was successfully achieved, which laid a foundation for the subsequent single-cell drug screening based on the formation of three concentration gradients.
3.3 Single cell-drug interaction study
To evaluate the capability of our constructed multiple-concentration platform for single-cell level drug screening, the separate action and combination interaction of anticancer drug A (5-Fluorouracil (5-FU) and drug B (cisplatin (DDP)) on single tumor cell were explored under the same operating conditions. Then drug A with a specific initial concentration of 100 µM and drug B with an initial concentration of 10 µM were routinely selected for chemotherapy. At the same time, normal medium (without drugs) was introduced into the device from drug C inlet. The two anticancer drugs produced three concentration gradients in the fluid storage chambers. In this circumstances, the percentage of 5-FU (drug A) concentrations (0, 12.5%, 25%, 37.5%, 50%, 62.5%, 75%, 87.5%, 100%) change from chamber 17 to chamber 1 in an ascending order. The percentage of DDP (drug B) concentrations (100%, 87.5%, 75%, 62.5%, 50%, 37.5%, 25%, 12.5%, 0) change from chamber 9 to chamber 17 in a descending order. Single tumor cells from chamber 2 to 8 were treated with different combinations of 5-FU and DDP. Thereinto, the ratio of 5-FU and DDP in chambers 2, 3 and 4 was just opposite in chambers 8, 7, and 6. Chamber 17 with drug-free medium was used as the control.
In the following step, the drugs containing medium with different concentrations were applied to HepG2 and MCF-7 cells in the single-cell capture device and traditional Petri dishes (Fig. 4A and Fig. 4B), respectively. The AO/PI double fluorescence staining (living cells/dead cells green/red) was used to characterize the cell viability after microfluidic single-cell capture and culture. The results showed the survival rate of tumor cells cultured by single drug (5-FU or DDP) stimulated increased with the decrease of single drug concentration in Petri dishes and single-cell capture device (Fig. 5A and Fig. 5B). The cell vitality was negatively correlated with drug dose. The untreated single cells (i.e., the control) in
chamber 17 showed normal viability and proliferation within the 2h of culture. When cells were stimulated with the same drug, cell activity at the single-cell level was lower than in conventional plate cultures. It indicates that there may be cell interactions in traditional plate culture with population effect, which can inhibit the influence of anticancer drugs on cells to a certain extent. The average response of the population is usually obtained in traditional flat culture, which covers the heterogeneity between cells. However, tumor cells at single-cell level are not affected by the cell interactions and can be effectively studied for their susceptibility to drugs[43–45]. Furthermore, when the two drugs are combined to act on the cells, the activity of tumor cells in the Petri dishes was lower than that of a single drug, indicating that the two drugs combined formed a synergistic effect on cells stronger than a single drug. The obtained results are similar to those previously reported discoveries and clinical trials in which combination therapy had advantages over monotherapy [43]. Interestingly, this phenomenon can also be found in single-cell capture devices of microfluidic chips while HepG2 cells are observed (Fig. 5A). For MCF-7 cells, the influence of two drugs on cell viability are all important and does not indicate synergy (Fig. 5B). We speculate that this is due to continuous infusion of drugs in the chip. The exact reason of this phenomenon remains unclear and warrants further investigation. This is the first time that multi-gradient dosing of two drug candidates has been demonstrated at the single-cell level, which has the potential to effectively screen mono therapeutic regimens and combination therapies.
Finally, to further investigate and analyze the relationship between biomechanical heterogeneity (such as deformations and size) of tumor cells at the single-cell level and potential drug resistance, we selected individual cells captured by the smallest filter unit (6–8µm) and the largest filter unit (14–16µm) as subjects. It was found that cell viability of small and/or deformable and large and/or less deformable cells all presented a dose-dependent mode after drug stimulation (Fig. 5C and Fig. 5D). The concentration gradient-dependent increase in the mortality of single cells was observed during the experiment. At multi-gradient doses of the two drug candidates, small and/or deformable cells showed higher resistance at the single-cell level than cells with large and/or less deformability, indicating that the biological heterogeneity of cells was correlated with their drug resistance. That is, tumor cells with a
small size or large deformability had stronger drug resistance than tumor cells with large size or poor deformability. This may be due to the small size and high degree of deformation of tumor cells, which is related to the high proportion of tumor-initiating cells and high resistance to chemotherapy drugs [44–45]. In addition, the main manifestations of 5-FU and DDP lethality were DNA synthesis and cell mitosis arrest[46–47], which were weakened by small tumor cells because these cells were predominantly in the G0/G1 phase of their cell cycle. These results indicated that functional gradient-like cell phenotypes in single cells were reconstructed successfully in the microfluidic device, which has advantages over previous microfluidic systems [11, 17]. This indicates that this platform could provide a potential way to study the drug screening by exploring tumor cellular heterogeneity at multiple drug gradients with the required stability and high throughput capability, which would help to facilitate the development of biological and preclinical explorations, such as screening-based cancer stem cell separation and drug discovery.