Acoustofluidic Droplet Sorter Based on Single Phase Focused Transducers

Abstract

Droplet microfluidics has revolutionized the biomedical and drug development fields by allowing for independent microenvironments to conduct drug screening at the single cell level. However, current microfluidic sorting devices suffer from drawbacks such as high voltage requirements (e.g., >200 Vpp), low biocompatibility, and/or low throughput. In this article, we introduce a single-phase focused transducer (SPFT) based acoustofluidic chip, which outperforms many microfluidic droplet sorting devices through high energy transmission efficiency, high accuracy, and high biocompatibility. The SPFT based sorter can be driven with an input power lower than 20 Vpp and maintain a post-sorting cell viability of 93.5%. The SPFT sorter can achieve a throughput over 1,000 events per second and a sorting purity up to 99.2%. We utilized our SPFT sorter here for the screening of doxorubicin cytotoxicity on cancer and noncancer cells, proving its drug screening capability. Overall, our SPFT droplet sorting device shows great potential for fast, precise, and biocompatible drug screening.

Graphical Abstract

Single Phase Focused Transducer (SPFT) is a novel acoustofluidic tool for droplet sorting, using a unique interdigital transducer design. The capability of the SPFT to generate high acoustic intensities enables high efficiency cell-laden droplet sorting with a low input voltage and a high biocompatibility. This new sorting technology shows significant potential as a tool for droplet-based screening and analysis.

1. Introduction

The introduction of droplet microfluidics to the biomedical and pharmaceutical fields has significantly enhanced the drug screening process by reducing sample consumption and improving throughput.^1–10 Compared with conventional methods, which utilize 96-well plates as the independent units and require several microliters of cell suspensions containing hundreds of cells per drug response assay,^10–14 droplet microfluidics reduce the volume of independent units to picolitre level.¹⁵ With the unit volume reduced, the processing time for every unit is also shortened, improving the throughput of droplet-based drug screening.^{16, 17} With these two improvements, droplet microfluidics has become an important drug screening tool.

The first step in droplet microfluidics is droplet generation. Current cell-laden droplet generation strategies use low-concentration cell suspensions for droplet production to ensure a high proportion of single cell occupancy in droplets. This technique, however, leaves the majority (>85%) of droplets empty.^18–21 If no measure is taken to remove empty droplets when conducting drug screening, an enormous amount of time will be wasted testing empty droplets, leading to low screening efficiency. Therefore, a microfluidic platform that can efficiently sort and enrich cell-laden droplets is particularly important during drug screening.^{22, 23}

A variety of microfluidic technologies have been developed to sort droplets effectively, such as dielectrophoretic sorters,^{24, 25} optofluidic sorters,^26–29 and magnetically activated droplet sorters.^{30, 31} Although these methods often achieve accurate droplet sorting, the high electric pulse, optothermal effect,²⁶ and magnetic additives^30–32 are potentially harmful to cells and influence cell viability. For example, the voltage applied to the dielectrophoretic sorter is often above 500 Vpp,^{19, 33, 34} causing cell viability to immediately drop from 95% to 87% right after exposure to the electric pulses.

Acoustofluidics, the fusion of acoustics and microfluidics, can effectively manipulate and sort picolitre droplets via noninvasive, mild, and highly biocompatible acoustic forces.^35–53 Therefore, acoustofluidic sorters often have advantages in biocompatibility.^54–56 However, current acoustofluidic droplet sorters have deficiencies in their energy conversion rate, which hinders their further development.⁵⁷ The standard interdigital transducer design enables the bidirectional transmission of acoustic waves, causing inefficient energy utilization and excess heat.⁵⁸ Due to this energy conversion deficiency, stronger input power must be used for high-throughput sorting, resulting in more heat generation, weakening the biocompatibility advantages of acoustofluidic sorters.

In this work, we present a single-phase focused transducer (SPFT) based droplet sorter. This SPFT design reduces energy waste by ensuring all acoustic waves propagate unidirectionally and focus at the sorting region. With the same input power, the energy transmitted into the microchannel by SPFT is around 4.83 times higher than by that generated by typical interdigital transducer based acoustofluidic sorting devices. Therefore, the SPFT-based droplet sorter can apply the same force as traditional interdigital transducer based devices with a significantly lower input power. Our SPFT sorter can be driven by input voltages as low as 16 Vpp and due to a low energy waste, a high post-sorting cell viability around 93.5% is achieved. In addition, the SPFT based droplet sorter enables fluorescence activated droplet sorting at a throughput of 1,182 events per second. We verified the capability of the SPFT sorter by utilizing it to detect the cytotoxicity of doxorubicin⁵⁹ on MCF-7 cells (tumor cells) and HEK-293T cells (non-cancer cells). Cell viability differences between the two types of cells verified selective cytotoxicity properties of doxorubicin. Overall, we describe a SPFT droplet sorter that achieves high purity, high biocompatibility, and low input voltage requirements, simultaneously. We believe that with these advantages and further development, the SPFT based sorter will become a powerful technology in the fields of droplet microfluidics, cell sorting, and drug screening.

2. Results

2.1. Design of the SPFT-based droplet sorter

Figure 1(A) shows the schematic of the SPFT based droplet sorter, which is composed of three major components: the microfluidic channel, optical detection part, and acoustic sorting part. The photo of a fabricated SPFT droplet sorter (Figure 1 (B)) shows the arrangement of the device. A lithium niobate (LiNbO₃) substrate, with a pair of SPFT deposited on it, was attached to a single-layer polydimethylsiloxane (PDMS) microfluidic channel. The microfluidic channel features two inlets, two outlets, and an operating region. The inlet in the middle is used for droplet injection, and the peripheral inlet is loaded with fluorinated oil (HFE-7500, RAN Biotechnologies, USA) to adjust the spacing between droplets. After the distance between droplets is adjusted to proper ranges and were steady, the droplets sequentially flow through a tapered channel where an optical fiber pair detectes fluorescence signals and then, according to the fluorescence information, pulse acoustic signals would be generated to trigger the SPFT pair. The droplet would be manipulated by standing surface acoustic waves (SAW) and guided to the collection outlet if identified as a cell-laden droplet. Otherwise, the droplet maintains its route to flow into the waste outlet.

Figure 1.

(A) Schematic of the SPFT based droplet sorter. Droplets containing the single green ball: single-fluorescence-cell-laden droplets; Droplets containing the single orange ball: non-fluorescence/dead cell laden droplets; bare droplets: empty droplets. The oil inlet injects HFE-7500, which functions as the spacing adjustment oil. (B) An optical image of the SPFT droplet sorter (optical fibers are not plugged in).

The core of the entire chip is the acoustic sorting part, which is composed of the SPFT pair. The SPFT pair is placed perpendicular to the flow direction and the focal point of the pair is on the side of the collection outlet, so that the pressure node line of the standing SAW is close to the collection outlet.^{60, 61} Compared with the regular interdigital transducer (Figure 2 (A)), the SPFT utilizes both focused sector design and unidirectional design. As shown in Figure 2 (B), one unit of SPFT design is an assembly of five fingers, including two connected fingers that are connected to the positive and negative electrodes, two floating fingers, and one “U shape” floating finger. The acoustic impedance of the floating fingers is larger than the LiNbO₃ substrate, thus the SAW propagating away from the focused center would be reflected. In addition, the acoustic impedance of the U shape floating finger is smaller than the substrate, causing the amplitude toward the focused center larger than the reverse direction.⁵⁸ The synergy of floating and U shape floating fingers makes a higher proportion of energy spread to the focused center (Figure S2). In addition, the fan-shaped focused design helps to further concentrate energy. When the floating and U shape floating fingers are placed at the octave point of the resonant wavelength, the phase of generated waves and reflected waves could be perfectly matched, reaching a maximal amplitude. In contrast, SAW generated by the regular parallel transducer would propagate bi-directionally, leading to half energy loss. Figure 2 (D) shows the simulation of standing SAW amplitude in this condition. Compared with the standing SAW amplitude generated by regular interdigital transducers (Figure 2 (C)), the energy transmission efficiency of SPFT is 4.83 times higher than the regular interdigital transducer.

Figure 2.

The geometry schematic of (A) the interdigital transducer pair and (B) the SPFT pair (only paint 3 pairs as a display in (A) and 1 pair in (B)). Simulation results for energy distribution rate of (C) the interdigital transducer pair and (D) SPFT pair, respectively.

2.2. Optimization of SPFT

The geometric dimensions of SPFT is performed to improve the device performance on energy utilization. Considering the attenuation during wave transporting, a proper distance from SPFT to channel would ensure the strength of standing SAW field in the microchannel. The bandwidth of the standing SAW field would determine the length of time that the standing SAW can perform at a certain droplet flow rate. The radiation force generated by standing SAW is expressed as^{54, 62, 63}

$$F_r=-\left(\frac{\pi p_0^2{V}_d{\beta }_w}{2\lambda }\right)\phi \left(\beta ,\rho \right)\sin \left(2kx\right)$$ (1)

$$\phi \left(\beta ,\rho \right)=\frac{5{\rho }_d-2{\rho }_w}{2{\rho }_d+{\rho }_w}-\frac{{\beta }_d}{{\beta }_w}$$ (2)

where p₀, λ, V_d, ρ_d, β_d, ρ_w, and β_w are the acoustic pressure, acoustic wavelength, droplet volume, density and compressibility of the droplet, and density and compressibility of the liquid, respectively. The values of the radian (θ) would influence the aggregation of standing SAW.⁶² We compared the performance of SPFT with different θ by adding same input power to SPFT and calculating droplet-shifting distances. Compared with SPFT of 10°, 15°, and 25° angles, SPFT with 20° angle performes best in the following aspect. When input power is lower than 24 Vpp, 20° SPFT pushes droplets to the furthest distance; after input power exceeded 24 Vpp, droplet-shifting distance driven by 20° SPFT achieves the maximum (Figure S3). The comparison result indicates that 20° SPFT shows the best response.

The diameter (R_s) determines the distance between the SPFT pair. In order to maintain the sector shape and the radian of the transducers, while ensuring that the pair of transducers still has a common focus, the arc length of the innermost finger (l) should increase proportionally with the increase of R_s, under the rule of l= θ×R_s/2. Therefore, R_s not only determines the shortest distance from standing SAW to the operation region, but also determines the bandwidth of standing SAW. Since SAW attenuate when they propagate, the distance that must traverse from the point of SAW generation to contact with the droplets should be as short as possible. In addition to the attenuation of SAW, the bandwidth of standing SAW must also be taken into account when choosing R_s. If the bandwidth is too wide, the spacing between droplets has to be increased to ensure that there is one and only one droplet in the operation region. In this case, the control flow rate would be particularly high to separate droplets far enough. If the bandwidth of the standing SAW field is too narrow, the time for the droplet to pass through the standing SAW field would also be shortened. Under high-throughput conditions, the execution time for sorting is not enough to push the droplets to the pressure node line, resulting in unsuccessful sorting. Our design finally determined R_s at 1 mm, which ensured the SPFT pair was as close to the channel as possible.

2.3. Experimental comparison between SPFT and regular interdigital transducers

In addition to theoretical analysis and simulation comparison, experiments were conducted to prove the advantage of SPFT over regular interdigital transducers. Based on optimization results, 20° SPFT based chip was selected for the comparison. Before conducting the assay, both types of transducer pairs (SPFT and regular interdigital transducers) were measured (Figure S4) for their respective resonance frequencies using a Network Analyzer^® (Vector Network Analyzer 2180, Array Solutions, USA). Both chips (SPFT and regular interdigital transducers) were running at their resonance frequencies, respectively. Figure 3 (A) and (C) show the trajectories of droplets when the standing SAW is turned off in each chip; all droplets flow into the left outlet (waste). After input power is applied to the two chips, the shifting of droplet trajectories in each chip is recorded by a microscope. As shown in Figure 3 (B) and (D), there is a significant difference in the ability of the two transducers to change the droplet trajectories, indicating that the SPFT possesses stronger standing SAW under the same input power. The quantitative relationship of shifting distance versus input voltage of two types of transducers is shown in Figure 3 (E). When the input power reaches 8 Vpp, the shifting distance of droplets in the SPFT based chip reaches 13 μm. Parts of the droplets start to be pushed into the sorted outlet under this input power. After increasing the input power to 16 Vpp, the shifting distance of droplets in the SPFT based chip increases to 47 μm and all droplets in the SPFT based sorter shift into the sorted outlet when standing SAW is turned on. In contrast, the regular interdigital transducer based sorter only makes droplets shift 19 μm at 32 Vpp, pushing only part of the droplets (~60%) into the sorted outlet. For further confirmation, the grayscale changes in the detection region (Figure S1) under 32 Vpp input voltage are detected. The response of the two chips is shown in Figure 3 (F). From the results, we could easily distinguish that after SPFT is turned on, all droplets are pushed into the sorted outlets successfully, while only about half of the droplets are successfully sorted by traditional interdigital transducers. From the assay, the SPFT based droplet sorter achieves the same effect of interdigital transducers based chip at a lower input voltage. The high energy utilization of the SPFT based sorter makes it possible to achieve high-throughput sorting with a relatively acceptable input voltage.

Figure 3.

Snapshots of the droplets flow track at the SPFT based droplet sorter when standing SAW was turned (A) off and (B) on. Input voltage: 32 Vpp. In comparison, snapshots of the droplets flow track at the regular interdigital transducer based sorter when standing SAW was turned (C) off and (D) on with the same experiments were conducted. Scale bar: 200 μm. (E) Comparison of droplet-shifting distances perpendicular to the flow, driven by the SPFT pair (red dots) and the regular interdigital transducer pair (blue dots). The results show that when the input voltage increases from 0 to 24 Vpp, droplet-shifting distance increases much more rapidly by SPFT than by interdigital transducers, indicating that the SPFT design possesses a higher energy utilization capability. (F) comparison between the SPFT design and the interdigital transducer design at 32 Vpp. After standing SAW were turned on at 0 s, droplets were continuously pushed into the sorted outlet by the SPFT pair (red line), while four out of ten droplets were missed in the regular interdigital transducer based sorter (blue line). The detection region was selected at the entrance of sorted outlet.

2.4. Periodic droplet sorting at high throughput

Herein, we verified the capability of the SPFT based sorter for high-throughput sorting. Under high-throughput conditions, the flow rate of the droplets is fast, which limits the pulse time to apply standing SAW onto each droplet. If the pulse time is longer than the duration that one droplet flows through the operation channel, more than one droplet might be sorted at once, causing errors. In order to prove that the standing SAW generated by the SPFT are strong and precise enough to sort single droplets at high throughput, we add a periodic pulse signal with 145 Vpp of input voltage, 1, 000 cycles of duration, and 5 ms of interval, onto the transducer pair. Figure 4 shows the on-and-off of SPFT pair within 3 ms. The SPFT successfully pushes the second droplet into the sorted outlet, while the adjacent droplets (droplets 1,3, and 4) maintains their trace. The droplets in Figure 4 are injected with the flow rate of 520 μL/min, or 1,014 events/s. This result proves the high-throughput sorting ability of our SPFT sorter. For further verification, a video is recorded through the microscope (Movie S1) and the gray value change over time at the sorted outlet entrance is detected (Figure S5). Subsequently, the intensity change of the pulse signal (blue line) is also introduced into the same graph. As can be seen from the graph, after one signal pulse is released, about 2.65 ms later, one gray value dip would form, indicating one successful sorting event occurred. During the interval between two pulses, no dip is detected. From this experiment, we prove that our design is capable of high-throughput (>1,000 events/s) droplet sorting with high accuracy.

Figure 4.

(A)~(F) Snapshots of five droplets flowing through the SPFT device within 3 ms. Scale bar: 200 μm. The periodic pulse signal applied on the SPFT pair: resonant frequency: 9.696 MHz; interval time between each pulse: 5 ms; and wave numbers of each pulse: 1000. Droplet #3 was moved by standing SAW and flowed into the sorted outlet while droplets #1, 2, 4, and 5 remained in the original trajectory.

2.5. Sorting of single-cell-laden droplets

The pulse signal activated sorting assay demonstrated the capability of our SPFT sorter for single droplet sorting under high throughput. In single-cell droplet related experiments, fluorescence technology is widely used, acting as the signals for identifying, screening, and sorting. Therefore, we designed a fluorescence activated single-cell droplet sorting experiment.

In droplet-based drug screening, single-cell-laden droplets are ideal research objects. A low percentage of cell-laden droplets greatly reduce the efficiency of drug screening since significant time and labor must be devoted to distinguishing between cell-laden and empty droplets. Therefore, rapid and efficient enrichment of single-cell laden droplets is necessary for droplet-based biological analysis. The efficiency of single-cell laden droplet generation is largely dependent on the cell suspension concentration. According to the distribution theory, the number of cells distributed in one droplet satisfies the Poisson distribution,³³

$$P\left(x=k\right)=\frac{{\mathrm{e}}^{-n}{n}^k}{k!}$$ (3)

where x stands for the number of cells in one droplet, n is the intrinsic parameter for Poisson distribution, and k is an integer, ranging from zero to infinity. Based on this principle, to avoid generating multi-cell-laden droplets as much as possible, an inevitable consequence is that a majority (about 85%) of the droplets have no cells. Here we used MCF-7 cell line to perform fluorescence activated single-cell droplet sorting through our SPFT sorter. Before sorting, about 15% of droplets contain fluorescent cells (Figure 5 (A)). Then, droplets are re-injected into the SPFT based sorter with four representative throughputs (~100, 300, 500, and 1,000 events/s) to validate the sorting purity under different throughputs. The sorting results under every throughput are collected and counted under a microscope (Figure 5 (B)). After sorting, the purity of cell-laden droplets increases from 14.8% to 99.2%, indicating a high sorting accuracy for cell-laden droplet enrichment. Moreover, by comparing Figure 5 (A) and (B), no droplet debris or other impurities are observed after sorting, which indicates that the sorting process also helps to filter debris to a certain extent, such as smaller droplets that may be generated by the compression of droplets during the transfer process. With the increase in throughputs, the sorting purity slightly decreases, but is still higher than 85% at 1,187 events/s. The result in Figure 5 (C) suggests that our SPFT based droplet sorter achieves precise cell-laden droplet sorting at high throughputs. For further verification, we count the number of cells encapsulated in the droplets after sorting. The cell distribution in Figure 5 (D) shows that most of the droplets are single-cell-laden droplets, consistent with the result from the Poisson distribution. According to these results, the SPFT based droplet sorter proves to be sufficient for the enrichment of single-cell-laden droplets, making them useful for subsequent droplet based single-cell cultivation and analysis applications.

Figure 5.

(A) Image of single-cell-laden droplets right after droplet generation, before acoustic sorting. (B) Image of droplets from the sorted outlet, after acoustic sorting. Scale bar: 200 μm. Both two images are composed of the green fluorescence picture and bright-field picture. (C) The purity of fluorescence-cell-laden droplets from the sorted outlet after sorting at different throughputs. (D) The ratio of the distribution of different numbers of cells per droplet after sorting.

2.6. Single-cell-laden droplet cytotoxicity assays

To further showcase the utility of our SPFT sorter, we developed a cytotoxicity screening assay. Through this assay, the SPFT based droplet sorter is demonstrated to be capable of completing droplet-based drug screening experiments. We chose doxorubicin as the target and selected MCF-7 wild type cell line as the model cancer cells and HEK-293T cell line as the non-cancer cells. The two cell types underwent the same cytotoxicity assay and received live/dead cell detection after 0, 4, 8, and 12 hours of incubation. Figure 6(A) shows the fluorescence signal changes after 8 hours incubation of MCF-7 cells with 5 μM doxorubicin. The green fluorescence dye stains live cells while the red fluorescence only marks dead cells. By periodically counting the ratio of live to dead cells in the droplets, we obtained the MCF-7 and HEK-293T cell viabilities’ change over time (Figure 6(B) and (C)). By comparing the results in Figure 6(B) and (C), the decrease in cell viability of HEK-293T cells is much lower than that of MCF-7 cells. At the end of 12 hours of incubation, the cell viability of MCF-7 cells in single-cell droplets drops to 48.9%, while 75.8% of single HEK-293T cells inside droplets keep alive. The results indicate that 5 μM doxorubicin is more cytotoxic to MCF-7 cells than to HEK-293T cells, which is consistent with other research.⁶⁴ Through this experiment, we prove that doxorubicin indeed possesses certain targeting properties, although it still has cytotoxicity to non-cancer cells. With the help of the SPFT based sorter, more targeted carriers or drugs can be screened out.

Figure 6.

(A) A plot of the states of MCF-7 cells in droplets after the doxorubicin assay. Live cell dye is green fluorescent and dead cell reagent is red fluorescent. Scale bar: 50 μm. (B, C) The comparison of MCF-7 cell and HEK-293T cell viability after the doxorubicin assay. The results indicate that 5 μM doxorubicin is more toxic to cancer cells (MCF-7 cells) than to non-cancer cells (HEK-293T cells).

As demonstrated in the doxorubicin cytotoxicity assay, the SPFT based droplet sorter not only exhibits an excellent capability for high-throughput and precise droplet sorting, but also shows a great potential for droplet-based drug screening and single-cell-level cellular response screening applications.

3. Conclusion

In this work, we have developed a high-throughput droplet sorter which utilizes a SPFT pair to sort target droplets with high speed and accuracy. The core concept of the design is to maximize energy use to meet the requirements of biocompatible, high-throughput sorting. The key part of the sorter is the SPFT pair, which combines the advantages of both the focused transducer design and U-shaped structure. With this design, the SPFT not only has the capability to focus acoustic waves but also possesses structures to convert generated waves into unidirectional signals. Compared to the regular transducer design, the SPFT could focus more energy to the focal point, allowing it to generate a much stronger standing SAW field with the same input power. In contrast to the recently released sorting strategy,^{33, 65–68} the SPFT based sorter is able to conduct precise cell-laden droplets under a low input pulse, and still maintains a high sorting purity of up to 99.2%. Overall, the SPFT based droplet sorter is able to continuously perform accurate droplet sorting with a throughput as high as 1,187 events per second. Moreover, from the doxorubicin assay results, the platform is proven to be biocompatible. After SPFT sorting, the average cell viability remained higher than 90%. The biocompatibility of the SPFT sorter makes it possible for use in droplet-based drug screening,⁶⁹ single cell analysis,^70–72 and many other applications in biology and medicine.^{20, 73–75} With continuous improvement, it will become a powerful platform for fields in biomedical science, pharmacology, and cell biology.

4. Methods and materials

4.1. Device fabrication

The SPFT based droplet sorter chip includes a pair of SPFT and a bonded PDMS channel in between the SPFT (Fig. 1). The SPFT (5 nm Cr and 150 nm Au) is fabricated via standard photolithography, e-beam evaporation, and liftoff process⁴⁵ on a 0.5 mm thick Y-128° cut lithium niobate (LiNbO₃, Precision Micro-Optics, USA) substrate. The PDMS channel is fabricated through soft lithography. Here, the PDMS channel includes a microfluidic channel with a height of 60 μm and two chambers for guiding incident and detection optic fibers with a height of 130 μm. The chamber for the incident fiber is arranged perpendicular to the microfluidic channel while the chamber for the detection fiber is 45° tilted to the microfluidic channel for better optical signal detection. After the fabrication, the PDMS channel is bonded on the LiNbO₃ substrate following a 6 min surface plasma treatment. Before performing a sorting task, the microfluidic channels are coated with Aquapel^® Glass Treatment hydrophobic reagent (Aerosol Glass Windshield Cleaner, Aquapel, USA) and the device is tested for tightness to prevent from leakage.

4.2. Experimental setup

In addition to the SPFT sorting chip, the experimental setup includes an optical signal detector (Oscilloscope, Keysight Technologies, USA), a function generator (AFG3102C, Tektronix, USA), and an amplifier (25A250A, Amplifier Research, USA). The function generator and amplifier are used to process signal from the oscilloscope and drive the SPFT pair. To set up the sorting system, the chip was firstly mounted on the workbench of an inverted microscope (TE2000-U, Nikon Ti, Japan). Then, an incident fiber, which was connected with a 488 nm laser source (CrystaLaser, USA) on one end as the source to activate fluorescence signals, was plugged in the perpendicular chamber. On the other side, a detection fiber was plugged into the chamber at a 45° tilted angle to the incident light path. The fluorescence signals collected by the detection fiber are filtered by a filter set (525/50 nm) and then translated to analog electric signals by a photomultiplier tube (PMT, C6780–20, Hamamatsu, Japan). The analog electric signals are monitored by a mixed signal oscilloscope (InfiniiVision MSOX2024A, Keysight, USA). When a fluorescent droplet flows across the incident light path, a dip in the electric signal will be detected by the oscilloscope. If the signal dip is below the preset threshold value, the oscilloscope will trigger the function generator to generate RF pulse signals. To perform droplet sorting, the triggered RF pulse signals are amplified and excite the SPFT for generating standing SAW. By adjusting the response delay time, signal frequency, voltage, and cycles of the RF pulse signal, standing SAW could be modulated for precisely sorting droplets based on the detected fluorescence.

4.3. Cell preparation

Two major cell lines, MCF-7 (ATCC^® HTB-22^™, USA) cells and HEK-293T (ATCC^® CRL-11268^™, USA) cells, were used in the assay. These cell lines were cultured in DMEM medium with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin in petri dishes. The cultivated cells were maintained in the 37°C, 5% CO₂ cell culture incubator (HERAcell VIOS 160i CO₂ Incubator, Thermo Fisher Scientific, USA) until the amount occupied 70% of the bottom area of the petri dish. Before assay, cells were resuspended in DMEM to a final concentration of 2 million per mL. The assayed cells were stained with Calcein-AM™ (Thermo Fisher Scientific) with the volume ratio of 1.5: 1000 (Calcein-AM: cell suspension) and coated for 15 minutes.

4.4. Droplet generation

The water-in-oil droplets were generated by an independent microfluidic generator.^{59, 64} In the droplet generator, the inner inject was connected to the cell suspension syringe and the peripheral inject was connected to the HFE-7500 oil. Two syringes were pumped by the syringe pump (neMESYS 290N, Germany) at the flow of 10 μL/min (inner) and 70 μL/min (peripheral), respectively. The droplet size was influenced by the flow focusing junction width, where the oil flow and cell suspension flow first meet. The flow focusing junction width of our droplet generator was set to 50 μm. Under this experimental setup, the diameter of generated droplets was 90 μm. When both the cell suspension and oil flows were stable, we used a 1.5 mL tube to collect generated droplets. The whole droplet generation process lasted about 10 minutes. For doxorubicin assay, since the drug was added before droplet generation, to minimize the impact of the difference in cultivation time, the single-cell-laden droplets were sorted out within 30 minutes and divided into four subgroups to incubate for 0-, 4-, 8-, and 12-hours with drugs.

4.5. Device operation and result interpretation

The sorting process was initiated when the platform was set up with generated droplets. Before sorting, the SPFT chip and all four connected tubings were filled with HFE-7500. Then, droplets kept in the tube were drawn by syringe and connected to the inner inlet. Another syringe filled with HFE-7500 was connected to the peripheral inlet. To test the device at different throughputs, the syringe pumps were turned on to inject droplets at the rate of 2.5, 9, 20, and 30 μL/min, respectively. At these flow rates, the flow of HFE-7500 was tuned to have the final throughput of 100+, 300+, 500+, and 1000+ events per second, respectively. The spacing between two droplets was stabilized to be 65 μm. When the droplet flow was stable, the incident fiber was turned on to generate 488 nm laser, and the optical fiber on the other side collected fluorescent signals to trigger the SPFT pair for droplet sorting.

The sorting results were either transferred into the tube for cultivation or directly observed and analyzed under the microscope. To calculate the sorting purity, we counted the number of cell-laden droplets, divided by the total number of droplets coming from the sorted outlet.

Apart from obtaining information from the sorting results, we could also evaluate the performance of the acoustic transducers during operation by measuring the shifting distance of droplets when standing SAW were turned on. The shifting distance of droplets was defined as the displacement of the droplet in the x direction at the midpoint of the operation channel in the y direction. To calculate the shifting distances of droplets under different input voltages, droplet trajectories in the operation channel would be recorded before and after input voltage was added. Then, the x coordinates of the center of droplets at the operation channel midpoint were extracted and the displacement would be calculated.

To verify whether the pushed droplets successfully entered the sorted outlet, we used the method of detecting the gray value at the sorted outlet. The interface of droplets was in a dark circle shape under the microscope, which would reduce the average gray value in the detection region. In the assay, we selected the entrance of the sorted outlet as the detection region (Figure S1) to record the average grayscale change over time. A gray value dip indicates one droplet entering the sorted outlet.

4.6. Doxorubicin cytotoxicity assay

Doxorubicin is a well-developed chemotherapy medicine that is commonly used to treat various types of cancer. It is believed to specifically target tumor cells while having little influence on somatic cells.^{59, 70} For the doxorubicin assay, cells were firstly treated with Live/Dead™ Cell Imaging Kit (488/570, Thermo Fisher Scientific) for 15 minutes. Then, the cells were suspended in DMEM to a final concentration of two million/mL and highly concentrated doxorubicin (D1515, Sigma-Aldrich) was added into the cell suspension to a final concentration of 5 μM. Next, the cell suspension was immediately encapsulated in HFE-7500 to generate cell-laden droplets of 90 μm in diameter. Then the SPFT based sorter was applied to enrich single-cell-laden droplets. The cell viability changes of the sorted droplets after doxorubicin treatment would be calculated under microscope.