Centrifugation-Assisted Single-Cell Trapping in a Truncated Cone

Nov 18, 2015 - significant loss of cells due to the truncated cone shape of the microwells. ... chips have been widely utilized for large-scale single...
0 downloads 0 Views 4MB Size
Download PDF
Article pubs.acs.org/ac

Centrifugation-Assisted Single-Cell Trapping in a Truncated Cone-Shaped Microwell Array Chip for the Real-Time Observation of Cellular Apoptosis Lu Huang,†,‡,§ Yin Chen,‡,§ Yangfan Chen,† and Hongkai Wu*,†,‡ †

Department of Chemistry, The Hong Kong University of Science and Technology, Hong Kong, China Division of Biomedical Engineering, The Hong Kong University of Science and Technology, Hong Kong, China

‡

S Supporting Information *

ABSTRACT: Microfluidic devices have been extensively used in single-cell assays. However, most of them have complicated structures (multiple layers, valves, and channels) and require the assistance of a pump or pressure-controlling system. In this paper, we present a facile centrifugation-assisted single-cell trapping (CAScT) approach based on a truncated cone-shaped microwell array (TCMA) chip for real-time observation of cellular apoptosis. Our method requires neither a pump nor a pressure-controlling system, and it greatly reduces the complexity of other cell-trapping devices. This method is so fast and efficient that single-cell occupancy could reach approximately 90% within a few seconds. Combined with modern fluorescence microscopy, CAScT makes the highly ordered and addressable TCMA a high-throughput platform (104−105 single-cell trapping sites per cm2) for singlecell analysis. Cells trapped in it could be exposed to various chemicals by directly immersing it in bulk solutions without the significant loss of cells due to the truncated cone shape of the microwells. As a proof of concept, we demonstrated the ability of our chip for the real-time observation of the apoptosis of single HeLa cells induced by the common anticancer drug doxorubicin. This simple, robust, and efficient approach possesses great potential in diverse applications, such as drug screening, biosensing, and fundamental biological research. system for fluid introduction and guidance, and some need specialized electronic or optical equipment. These auxiliary facilities are not universal in a common medical or biological laboratory and induce extra obstacles for operation. Second, the structures of many microfluidic chips are complicated due to the need for multiple layers, channels, and valves, which are difficult to be fabricated and/or manipulated.10,12,22 Third, the auxiliary parts occupy much space and thus greatly lower the area density of single-cell arrays. Last but not least, many of them show compromised performance. For example, some devices with dense microhurdles or dams are high-throughput but not sample-economic;10,11,23 some based on gravity-induced cell sedimentation have simple structures but only show low or moderate single-cell resolution with a time-consuming isolation process;18−21 and others are low-throughput with a small number of cell-trapping sites.10,12 Construction of a platform concurrently possessing high throughput, high single-cell resolution, and a low expertise requirement remains a challenge.

B

ulk-cell assays such as ELISA, Western blot, and real-time PCR have been extensively applied in cellular experiments. Despite the success of these methods, they are incapable of disclosing cellular heterogeneity, which is important as indicated by more and more research.1−3 Recently, single-cell assays have been proposed for cellular studies.4 In this context, cellular contents, behaviors, and responses to external stimuli are analyzed at the single-cell level leading to a more comprehensive understanding of cell-to-cell variations. Microfluidic chips have been widely utilized for large-scale single-cell isolation and manipulation due to their compatible dimensions.5−8 Isolation and localization of large amounts of cells with high single-cell resolution are the primary functions for those devices. To date, microfluidic chips with different separation principles have been developed. For example, some of these chips used physical structures such as microhurdles or dams to capture single cells in microfluids;9−12 some employed finecontrolled microdroplets to encapsulate single cells;13−15 some utilized integrated optoelectronic tweezers or dielectrophoresis to isolate cells;16,17 and others applied cell-sized microwells or chambers to dock gravity-induced sedimentary cells.18−21 For those devices, practical applications in clinics or general biological laboratories are limited due to four major drawbacks. First, most of them require a pump or a pressure-controlling © 2015 American Chemical Society

Received: August 7, 2015 Accepted: November 18, 2015 Published: November 18, 2015 12169

DOI: 10.1021/acs.analchem.5b03031 Anal. Chem. 2015, 87, 12169−12176

Article

Analytical Chemistry

methods mentioned above, this method is not only highthroughput but also fast and can be handily operated with minimal expertise requirements. The capability of the developed TCMA as a platform for real-time observation of the apoptosis of single cancer cells induced by an anticancer drug doxorubicin was evaluated. Different from previous devices that introduced stimuli by complicated microchannels, valves, or pumping systems,10,11,23 cells trapped in an addressable TCMA could be directly immersed in bulk incubation solutions with almost no cell loss. Here, a HeLa-C3 cell line expressing a yellow fluorescent protein (YFP)/cyan fluorescent protein (CFP) Förster resonance energy transfer (FRET) probe was used as a convenient cell model for investigating single-cell apoptotic behaviors.28

Live-cell imaging has been widely incorporated with microfluidic chips for real-time single-cell observation.11,23,24 However, most of the previous works could not track massive amounts of individual cells precisely due to the small field of view provided by a microscope, undistinguishable cell-trapping units, and possible cell transposition caused by fluid fluxion or external interference. Herein, we present a simple but efficient centrifugation-assisted single-cell trapping (CAScT) approach based on a truncated cone-shaped microwell array (TCMA) chip for single-cell isolation (Figure 1a). Combined with automatic live-cell imaging, the TCMA chip provided a high-throughput and high-resolution platform for real-time single-cell observation. In our work, photolithography based on the underexposure of negative photoresist led to the formation of highly dense and ordered TCMA on cured poly(dimethylsiloxane) (PDMS) chips. Each truncated cone-shaped microwell has an opening with a diameter significantly smaller than that of its bottom. This unique structure can effectively prevent cell transposition caused by possible flow disturbance. The diameter of the opening and depth of the microwell have been optimized to obtain a high single-cell trapping efficiency. A centrifugal machine is routine equipment in both research laboratories and clinics. It has been successfully employed for single-cell loading in several previous works.25−27 In our study, the method was further systematically optimized to ensure its best performance in cell trapping. After systematic optimization of CAScT, cells could be trapped into microwells within a few seconds at high single-cell occupancy and viability. Distinct from the other

■

EXPERIMENTAL SECTION Reagents. Negative photoresist (SU-8 3025) and its developer were purchased from MicroChem (Newtown, MA). Trichloro(1H,1H,2H,2H-perfluorooctyl)silane (catalog no. 448931) was obtained from Sigma−Aldrich (St. Louis, MO). PDMS prepolymer kit (RTV615) was provided by Momentive Performance Materials (Waterford, NY). Bovine serum albumin (BSA, catalog no. A7906), Dulbecco’s modified Eagle’s medium (DMEM, catalog no. D7777), fluorescein diacetate (FDA, catalog no. F7378), paraformaldehyde (PFA, catalog no. P6148), and gelatin (catalog no. G1890) were purchased from Sigma-Aldrich. Fetal bovine serum (FBS, catalog no. 26140-079), Hoechst 33342 (catalog no. H1399), 4′,6-diamidino-2-phenylindole diacetate (DAPI, catalog no.

Figure 1. (a) Bright-field images showing the cross-sectional view of microwells on a PDMS replica, a nonaddressable TCMA chip, an addressable TCMA chip, and a representative number-coded block on an addressable TCMA chip (from left to right). (b) Procedures for the preparation of a PDMS chip with a truncated cone-shaped microwell array (TCMA). (c) Steps for centrifugation-assisted single-cell trapping (CAScT). (d) Schematic illustration of the principles for the fabrication of inverse truncated cone-shaped microposts (left) and normal columnar microposts (right). 12170

DOI: 10.1021/acs.analchem.5b03031 Anal. Chem. 2015, 87, 12169−12176

Article

Analytical Chemistry

surface and 60° inclined plane of metal stubs. After that, the specimens were observed under a scanning electron microscope (SEM, JSM 6490, JEOL), and images were taken at a 10 kV accelerating voltage at 200× , 800×, and 1000× magnifications. Optimization of CAScT. A 5 mL HeLa cell suspension was loaded into each conical centrifuge tube with a PMDS chip. The tubes were centrifuged (Centrifuge 5810 R, Eppendorf) with varying physical parameters (size of the microwell, centrifugation rate, cell concentration, centrifugation duration, and centrifugation cycles). After centrifugation, PDMS chips loaded with cells were carefully detached from the cushions and transferred into cell culture dishes filled with fresh medium. Chips were gently rinsed with fresh medium to remove untrapped cells and then replaced with 5 mL of new medium to maintain cell vitality. Cells trapped in TCMA were observed under an upright epifluorescent microscope (BX41, Olympus), and bright-field images (4× and 10× magnifications) were captured with a computerized charge-coupled device (CCD) camera (12.0 Monochrome w/o IR-18, Diagnostic Instruments). Single-cell occupancy (rate of microwells filled with a single cell) and double-cell occupancy (rate of microwells filled with double cells) were calculated accordingly. Cell Viability Test. Hoechst 33342 (10 mg/mL in ultrapure water) and FDA (2 mg/mL in methanol) solutions were added together in cell culture medium (1:1,000) for cell staining. After incubation for 30 min, the medium was renewed to remove residual dyes, and the fluorescence images of cell nuclei (blue color) and hydrolyzed FDA (green color) inside living cells along with bright-field images were taken sequentially. Cell numbers were obtained by performing particle analysis on each fluorescence image using ImageJ software (NIH). Cell viability was determined by dividing the number of living cells (stained by FDA) by the total quantity of cells (stained by Hoechst 33342). Inspection of Cell Morphology. A laser scanning confocal microscope (LSCM, LSM 710, Zeiss) was utilized for inspecting morphologies of single HeLa-C3 cells trapped in the TCMA. After CAScT, cells were fixed with PFA solution (4% in PBS) for 30 min and stained with DAPI (5 μg/mL in PBS) for 1 h. The chip was then gently washed with PBS and placed onto a clean coverslip with a layer of PBS. Fluorescence images of YFP (green color) and DAPI (blue color) along with bright-field images were taken using LSCM and processed with ZEN 2009 software (Zeiss). Real-Time Observation of Single-Cell Apoptosis. A live cell observation station (LCOS, equipped with a Nikon Ti-E-PFS microscope, Chamlide TC chamber, and Andor Zyla sCMOS camera) was used for real-time observation of single-cell apoptosis. Sterilized PDMS chips in the centrifuge tubes were soaked in 1 mL of gelatin solution (0.2 wt % in PBS) for 30 min and then rinsed with sterilized PBS three times. After CAScT of HeLa-C3 cells, the chips were gently taken out and immediately immersed into fresh medium in new culture dishes. The specimens were further rinsed with fresh medium to remove untrapped cells and then placed into different wells of a 12-well plate (TPP, Switzerland). Each loaded well was filled with 1 mL of cell culture medium to maintain cell viability. The well plate was maintained in the incubator for 30 min to warm the cells, and then it was mounted into the incubation chamber (37 °C, humidified 5% CO2/95% air at a flow rate of 50 mL/min) on the motorized stage of the LCOS. Five coded blocks on each PDMS chip were randomly selected for time-lapse observation. Prior to anticancer drug treatment, initial images of CFP

D3571) and penicillin/streptomycin (P/S, catalog no. 15140) were provided by Life Technologies (Carlsbad, CA). Trypsin (catalog no. 0458) and ethylenediaminetetraacetic acid disodium salt (EDTA·2Na, catalog no. LE118) were bought from Amresco (Solon, OH) and Genview (El Monte, CA), respectively. Doxorubicin (DOX, catalog no. D4000) was supplied by LC Laboratories (Woburn, MA). The rest of the chemicals were acquired from Sigma-Aldrich unless otherwise stated. All reagents were used as received unless otherwise mentioned. Cell Culture. HeLa (Human cervix adenocarcinoma cell line, CCL-2TM, ATCC) and HeLa-C328 were used in this study. Cells were cultured in DMEM supplemented with 3.7 g/L NaHCO3, 10% FBS, and 1% P/S, which was renewed every other day.29 They were maintained in 10 mL of medium incubated at 37 °C in a humidified atmosphere of 5% CO2 and 95% air. To harvest the cells, the medium was replaced by 5 mL of phosphate-buffered saline (PBS, pH = 7.4) when approximately 80% confluence had been reached. After 5 min in the incubator, PBS was replaced by 1 mL of 0.05% trypsin/0.53 mM EDTA·2Na solution (in PBS), and the cells were gently shaken and placed back in the incubator until they were thoroughly detached. Afterward, 5 mL of medium was added, and the cell density was determined by a hemocytometer. Most of the cell suspension was diluted to the desired concentration for CAScT while the remaining part was transferred to a new culture dish (60.1 cm2, catalog no. 93100, TPP, Switzerland) for proliferation. Fabrication of SU-8 Molds and PDMS Chips. SU-8 molds bearing an inverse truncated cone-shaped micropost array were fabricated by modified photolithography. SU-8 3025 photoresist was spin-coated (800 rpm for 10 s and 2 400 rpm for 30 s) onto a clean silicon wafer (525 mm in thickness) to form a uniform film with a thickness of approximately 30 μm. After soft bake (12 min at 95 °C), UV light underexposure (6 s, 18.9 mW cm−2) under a chrome mask with a designed pattern; and post exposure bake (1 min at 65 °C and 10 min at 95 °C), the photoresist was developed (1 min in a SU-8 developer and then gently flushed with fresh developer 3−4 times) and dried by pressurized air. Post UV light exposure and hard bake (10 min at 120 °C) were further applied to thoroughly solidify the SU-8 patterns. Subsequently, the SU-8 mold was treated by oxygen plasma (PDC-32G, Harrick Scientific Products, Inc.) for 2 min and silanized with trichloro(1H,1H,2H,2H-perfluorooctyl)silane vapor in a vacuum container overnight. Afterward, a mixture of PDMS prepolymer (10:1 oligomer to curing agent) was poured onto the master mold and cured at 110 °C for 2 h following degassing. Finally, the cured PDMS membrane was peeled off the wafer and cut into smaller chips for use. (Major procedures are shown in Figure 1b). Assembly of the Device for CAScT. A 10 g mixture of PDMS prepolymer (10:1 oligomer to curing agent) was poured into a 50 mL conical centrifuge tube (Falcon, Fisher Scientific), which was kept vertical overnight (at room temperature) and then heated (20 min at 110 °C) to completely cure the rubber before use. A clean PDMS chip was laid on the flat top of the PDMS filling in the tube and carefully pressed to make a tight contact to avoid floating during the addition of cell suspension. Finally, the assembled device was sterilized by ethanol and then rinsed by autoclaved PBS before use. Characterization of SU-8 Molds and PDMS Chips. A 15 nm gold layer was deposited onto SU-8 molds and PDMS chips in a gold sputter coater (S150B, Edwards). Then, they were mounted with double-side sticky tape on the horizontal 12171

DOI: 10.1021/acs.analchem.5b03031 Anal. Chem. 2015, 87, 12169−12176

Article

Analytical Chemistry

contact with developer, thus resulting in the inverse truncated cone-shaped microposts after development. Thus, the exposure time was critical for obtaining this unique feature. Exposure times of 2−240 s were tested, and our results revealed that when the exposure time was too short (2−5 s), it was difficult to keep most of the microposts erect after development. This is reasonable because a too short exposure time resulted in such a low cross-linking degree at the bottom of the microposts that they were unable to stand straight with the erosion and flushing of developer. In the worst case, microposts with smaller diameters collapsed and were even flushed away by developer. A longer exposure time (>10 s) could guarantee that the microposts would be upright, but it reduced the gradient of the degree of cross-linking within them, thus generating microposts with vertical sidewalls. Given a fixed UV light intensity (18.9 mW cm−2), the exposure time was optimized to be 6 s, which led to the generation of robust microposts with distinct inclined sidewalls, as shown in Figure 1a. Characterization of SU-8 Molds and PDMS Chips. After fabrication of SU-8 molds and PDMS chips, SEM was applied for the quantitative characterization of their sizes. Measuring the actual dimensions of microwells would help to select the proper TCMA for trapping cells. For each type of TCMA, the diameter of the designed micropattern on the chrome mask, the diameter of the opening (d), and the diameter of the bottom (determined with the diameter of the top of the corresponding microposts) (D), were measured using the scale bars in the SEM images (Figure 2 and Figure S1a) for reference. The actual sizes

(blue color) and YFP (green color) were captured sequentially using a CFP filter cube (excitation, 436 ± 20 nm; emission, 480 ± 20 nm) and a CFP/YFP filter cube (excitation, 436 ± 20 nm; emission, 535 ± 15 nm), respectively, with a 400 ms exposure time. Bright-field images were also taken to identify the code number of each block. A 20× objective (Plan Fluor Dry Phase contrast, NA 0.45) was applied so that nearly a complete block of TCMA could be observed in the field of view. Subsequently, cells were treated with the anticancer drug by replacing the normal medium with 1 mL of DOX-containing medium (10 μM, 20 μM, or 50 μM) and incubated for 1 h, after which the culture medium was changed back to the normal one. CFP, YFP, and bright-field images of the selected blocks were recorded every 2 h after DOX treatment. These images were collected after the culture medium had been renewed to reduce the background caused by C3-sensor leakage and to avoid contamination by microorganisms. Fluorescence intensities of YFP and CFP within single cells with background deduction were quantified using ImageJ software. A normalized YFP/CFP (Y/C) emission ratio was calculated by dividing the Y/C ratio at each time point with the initial one (right before drug treatment). Statistical Analysis. Experimental data were presented as mean ± std. Statistical significance was calculated by Welch’s t-test with a homemade program in Matlab R2013b (The MathWorks).

■

RESULTS AND DISCUSSION Design and Fabrication of TCMA. Two types of TCMA chips were designed: nonaddressable and addressable. Nonaddressable chips have a simpler array consisting of 15 625 (125 × 125) truncated cone-shaped microwells within 1 cm2 area as shown in Figure 1a. The distance between two neighboring microwells was designed to be 80 μm, which can be shortened to obtain a higher density up to 105 microwells/cm2. To trap single mammalian cells (normally 10−20 μm in diameter) in the microwells, a chrome mask with arrays of roundshaped micropatterns whose diameters range from 10 to 24 μm was made. Addressable chips were designed for the precise tracking of single cells. They have 100 blocks within a 1 cm2 area, and each block is composed of 10 rows and 10 columns of microwells. A block contains 96 microwells and a unique code number in the middle, as shown in Figure 1a. The distance between neighboring microwells remains 80 μm, and the spacing between adjacent blocks is 200 μm. Both parameters could be adjusted to obtain the desired microwell density. Standard photolithography has been widely used for photoresist mold fabrication. The appropriate UV exposure dose is very critical in making fine structures. The principle using negative photoresist (SU-8) is drawn in Figure 1d. The part of SU-8 beneath the transparent region of the mask was exposed to UV light, which became cross-linked and resistant to developer. The unexposed part covered by the opaque region of the mask was removed by developer. Standard photolithography with normal UV light exposure creates microstructures featuring vertical sidewalls. However, we found that underexposure could lead to the formation of inverse truncated cone-shaped microposts, which is believed to be caused by incomplete cross-linking. The cross-linking degree of the microposts decreases gradually from the top layer to the bottom layer, as shown in Figure 1d (the brighter the color, the lower the degree of cross-linking).30,31 The part with a lower degree of cross-linking was less resistant to developer, and erosion was more prominent at the edges of microposts that had direct

Figure 2. Characterization of the SU-8 mold and PDMS chip. (a) Scanning electron microscopy (SEM) images of SU-8 microposts designed with sizes of 10−24 μm (800× , top view). (b) SEM images of PDMS microwells designed with sizes of 10−24 μm (1000× , top view).

of the microwells versus the designed sizes are summarized in Figure S1b. Our results demonstrated that the diameters of the microwell bottoms were slightly larger than the designed sizes, whereas the corresponding openings were 3−4 μm smaller than the designed sizes. The difference between D and d was normally between 4 and 5 μm for a TCMA. Optimization of CAScT. In our study, the CAScT method was systematically optimized for single-cell isolation, and its procedure is illustrated in Figure 1c. Briefly, a TCMA chip was laid on the surface of solidified PDMS filling in a centrifuge tube and pressed to make a tight contact with it. After sterilization, 12172

DOI: 10.1021/acs.analchem.5b03031 Anal. Chem. 2015, 87, 12169−12176

Article

Analytical Chemistry cell suspension was added and the cells were deposited into the microwells under centrifugal force. To obtain high single-cell occupancy, optimal centrifugation conditions were explored. Five factors are most likely to affect the cell occupancy, i.e., the size of microwells, the centrifugation rate, the centrifugation duration, the cell concentration, and the centrifugation cycle. HeLa cells were used for in our preliminary experiment, and it showed that when TCMA chips with a designed size of 24 μm were used, high single-cell occupancy could be obtained at a centrifugation rate of 1 500 rpm (453g), a cell concentration of 200 000 cells/mL, a duration of centrifugation of 1 s, and a single centrifugation cycle. On the basis of these initial conditions, we started to optimize the CAScT method by varying only one physical parameter at a time while keeping the others fixed (Figure 3).

cells were left to settle under gravity (without a centrifugal force), no cells were found to be trapped (Figure S2). When the centrifugation rate was raised to 1 250 rpm (314g), a high cell occupancy was achieved, consisting of 78.5 ± 2.5% singlecell occupancy and 7.1 ± 1.6% double-cell occupancy. To our surprise, 1 500 rpm (453g) seemed to be the optimal centrifugation rate, and further increasing it (2 000 rpm, 805g) caused higher double-cell occupancy by sacrificing single-cell occupancy. For microwells with smaller openings, a lower cell occupancy was achieved under the same centrifugation rate (Figure S2), suggesting that a stronger centrifugal force was required to trap the cells. On the basis of these observations and cross-section views of a TCMA chip filled with cells (Figure S7), we can conclude that the sizes of most HeLa cells were bigger than the selected microwell openings; therefore, only when the centrifugal force was large enough could they be deformed and pushed into the wells. Consequently, we believe that most cells could not escape from the microwells in the subsequent processing without an external force. Cell nuclei are generally believed to be rigid and nondeformable;32 thus, it is reasonable that a threshold value for the size of microwell openings exists, below which (14.8 μm for HeLa cells in this study) cells could not be trapped using centrifugal force. In addition, too large of a centrifugal force was detrimental to cell trapping because it may not only lead to an increase in doublecell occupancy but also cause damage to the cells. Compared with the method that traps cells via passive gravity-driven settling (0−4 cells per microwell in 5−10 min with a singlecell occupancy of approximately 40−50%),20 CAScT using a 24 μm-sized TCMA at a centrifugation rate of 1 500 rpm (453g) demonstrated an improvement of approximately 40% in singlecell occupancy, and it requires less time. Because of these findings, we adopted these parameters for further optimization of CAScT. Another important factor is the cell concentration. As displayed in Figure 3c, the cell occupancy exhibited an increase when the cell concentration increased from 50 000 cells/mL to 200 000 cells/mL, but then it decreased slightly at 400 000 cells/mL. We roughly estimated the area density of cells on the surface of the PDMS filling to be approximately 3250 cells/mm2, which means that there were approximately 1.2 cells per microwell opening area for 24 μm-sized TCMA chips. It is hypothesized that at such a high cell concentration, the crowding effect of cells may block them from being trapped in the microwells. Considering this phenomenon, we set the cell concentration to be 200 000 cells/mL. The centrifugation duration may also influence the cell occupancy. We evaluated durations of 1 s, 1 min, 2 min, or 5 min and then explored their effects on cell trapping. It was found that increasing the duration of centrifugation could not enhance single-cell occupancy. Instead, double-cell occupancy rose significantly (p < 0.001) from 4.4 ± 1.2% for 1 s to 11.0 ± 1.9% for 5 min. It is likely that the cell-trapping process was completed within a very short time so that increasing the time for centrifugation would create a better chance to force an additional cell to enter an occupied microwell, thus raising the double-cell occupancy. Lastly, we investigated the effect of centrifugation cycles on CAScT. After the first centrifugation cycle, centrifuge tubes were taken out and gently shaken to resuspend untrapped cells. Then, the centrifugation process was repeated for one or two cycle(s). As anticipated, although the total cell occupancy increased to nearly 100%, additional centrifugation cycles could not effectively increase the single-cell occupancy.

Figure 3. Optimization of cell trapping: (a−e) Influence of five major physical parameters in CAScT on the cell occupancy. The initial conditions were set as follows: TCMA with a designed diameter of 24 μm, a centrifugation rate of 1 500 rpm (453g), a cell concentration of 200 000 mL−1, a centrifugation duration of 1 s, and 1 centrifugation cycle. The physical parameters were optimized one by one following the as-shown order by keeping the remaining four parameters fixed. Results were presented as the mean ± standard deviation (std) based on 9 randomly chosen areas (each contains about 250 microwells) from triplicate experiments. (f) Representative bright-field image showing the TCMA filled with cells (*p < 0.05, **p < 0.01, ***p < 0.001).

As shown in Figure 3a, the cell occupancy was enhanced in accordance with an increase in the size of microwells, which reached the highest level (>90%, 86.1 ± 2.0% for single-cell occupancy and 4.4 ± 1.2% for double-cell occupancy) when 24 μm-sized TCMA was used. The cutoff value for the designed size of microwells that allowed the smallest HeLa cells to enter was 18 μm, which corresponded to openings with a diameter of 14.8 μm. For the centrifugation rate (Figure 3b), using the centrifugal machine in our study, a threshold value of 1 000 rpm (201g) was also discovered, below which almost no cells were trapped in the microwells. As a control, when the 12173

DOI: 10.1021/acs.analchem.5b03031 Anal. Chem. 2015, 87, 12169−12176

Article

Analytical Chemistry On the contrary, double-cell occupancy rose significantly (p < 0.001) from 4.4 ± 1.2% with a single centrifugation cycle to 14.1 ± 0.8% with three centrifugation cycles. Therefore, one centrifugation cycle was enough to achieve optimal singlecell occupancy. Cell Viability and Three-Dimensional (3D) Morphology. Because the centrifugal force may cause damage to the cells, we analyzed their viabilities after CAScT under different conditions. Our results suggested that except for CAScT performed on 18 μm-sized TCMA chips, more than 90% of cells were still alive after centrifugation (Figure 4 and Figures S3−S6).

Figure 5. Inspection of the 3D morphology of HeLa-C3 cells trapped in a 24 μm-sized TCMA by a laser scanning confocal microscope (LSCM). (a) Fluorescence image of cell nuclei (stained by DAPI, blue color). (b) Fluorescence image of YFP in C3 probes (green color). (c) Bright-field image of TCMA on a PDMS chip. (d) Merged image of parts a−c.

Real-Time Observation of Single-Cell Apoptosis. As a proof of concept for the application of TCMA chips in singlecell analysis, we applied this platform for the real-time observation of single-cell apoptosis induced by an anticancer drug. With a live cell observation station (LCOS) consisting of an incubation chamber and a fluorescence microscope with an automatic tracking function, addressable TCMA chips can work as a high-throughput platform for single-cell analysis. HeLa-C3 cells, which express a FRET probe (named as C3) composed of a yellow fluorescent protein (YFP) and a cyan fluorescent protein (CFP) linked by a sequence containing a caspase-3 specific cleavage site, were used as a convenient model for the apoptotic study.28,33 Without activation of caspase-3, fluorescence emitted from the C3 probe is mostly green under the excitation of CFP due to the FRET effect. After activation of caspase-3 by external stimuli, such as UV irradiation, TNF-α and anticancer drugs,34,35 the fluorescence is shifted to blue due to the cleavage of the C3 probe. Activation of caspase-3 is a major pathway in activating cell apoptosis. It has been considered to be a key mechanism for the action of antitumor drugs, such as DOX.36−39 Thus, the shift in fluorescence color inside HeLa-C3 cells can be used as an index for monitoring the progress of cell apoptosis induced by DOX. HeLa-C3 cells in addressable TCMA were treated with DOX for 1 h, before and after which fluorescence images of YFP, CFP, as well as brightfiled images, were captured. The intensities of YFP and CFP fluorescence were obtained from the images using ImageJ software. The normalized Y/C emission ratio (Y/C ratio for short thereafter) was calculated by normalizing Y/C ratios at designated time points to the initial value. A lower value of the ratio indicates a more advanced progress of cell apoptosis. Figure 6 shows real-time changes in the Y/C ratio from single cells as a function of the DOX concentration (marked in the upper right corner of each scattered plot). For each DOX concentration, results were obtained from 100 to 200 single cells in duplicate experiments. As expected, heterogeneous apoptosis was observed among the cell population. For some cells hypersensitive to DOX, apoptosis occurred so fast that fluorescence signals were gone soon due to the fragmentation

Figure 4. Cell viability test: (a) representative fluorescence microscopy images of cell nuclei (stained by Hoechst 33342, in blue color) and living cells (stained by FDA, in green color) at optimal conditions. (b) Effects of different physical parameters in CAScT on the cell viability. The initial conditions were set the same as those in Figure 3. The physical parameters were optimized one by one following the asshown order by keeping the remaining four fixed. Results were presented as the mean ± standard deviation (std) based on 9 randomly chosen areas (each area contains about 250 microwells) from triplicate experiments.

The lower cell viability (approximately 80%) for 18 μm-sized TCMA chips may be explained by the trapping of some dead cells. As mentioned before, most cells were too large to enter the microwells of 18 μm-sized TCMA chips, as reflected by the low cell-trapping efficiency (