High-throughput single-cell quantification using simple microwell-based cell docking and programmable time-course live-cell imaging

Abstract

Extracting single-cell information during cellular responses to external signals in a high-throughput manner is an essential step for quantitative single-cell analyses. Here, we have developed a simple yet robust microfluidic platform for measuring time-course single-cell response on a large scale. Our method combines a simple microwell-based cell docking process inside a patterned microfluidic channel, with programmable time-course live-cell imaging and software-aided fluorescent image processing. The budding yeast, Saccharomyces cerevisiae(S. cerevisiae), cells were individually captured in microwells by multiple sweeping processes, in which a cell-containing solution plug was actively migrating back and forth several times by a finger-pressure induced receding meniscus. To optimize cell docking efficiency while minimizing unnecessary flooding in subsequent steps, circular microwells of various channel dimensions (4–24 µm diameter, 8 µm depth) along with different densities of cell solution (1.5–6.0 × 10⁹cells per mL) were tested. It was found that the microwells of 8 µm diameter and 8 µm depth allowed for an optimal docking efficiency (>90%) without notable flooding issues. For quantitative single-cell analysis, time-course (time interval 15 minute, for 2 hours) fluorescent images of the cells stimulated by mating pheromone were captured using computerized fluorescence microscope and the captured images were processed using a commercially available image processing software. Here, real-time cellular responses of the mating MAPK pathway were monitored at various concentrations (1 nM–100 µM) of mating pheromone at single-cell resolution, revealing that individual cells in the population showed non-uniform signaling response kinetics.

---

Introduction

Cellular responses to external signals have been mostly investigated by utilizing bulk-scale methods that measure average outcomes for a population of cells. For example, commonly used methods for high-throughput, cell-based assays rely on a well-plate (96-, 384- and 1536-well) format.^1–3 Another experimental approach such as western blotting analysis has been extensively performed to monitor protein expression in cells.^4,5 Despite the success of these assays, such population-wide studies mask the behavior of individual cells and are often insufficient for characterizing biological processes in which cellular heterogeneity plays a key role.⁶ Because the living cell is a complex system in which thousands of biochemical processes occur simultaneously, the cells under apparently identical environmental conditions often display a distribution of heterogeneous responses.^7–9

For the measurement of cellular responses at single-cell level, flow cytometry^10,11 and automated microscopy^12,13 have been widely used. Flow cytometry has been most successful for single-cell analysis because of its massive throughput combined with fluorescent labeling, which allows the quantitative determination of various protein levels in a population of cells. Although flow cytometry can be used to obtain biological information from thousands of cells, it provides only a snapshot of cellular responses at single time points. Also, traditional microscopy experiments can track biological responses in individual cells, but can only monitor a relatively small population of cells.¹⁴

To overcome the limitations of conventional approaches described above, microfluidics or “lab-on-a-chip” technologies have been introduced, in which array-based techniques were incorporated with automated time-course live-cell imaging. This platform can be used to track biological responses in individual cells and monitor large populations of cells, allowing for high-throughput experimentation to study single-cell information over time. More specifically, a number of cell docking strategies have been developed to isolate and capture single cells including physical trapping,^15,16 chemical patterning,^17,18 electrical positioning^19,20 and microwell-based docking.^21–23 In conjunction with these methods, programmable fluorescent microscopy can provide the ability to revisit individual cells over time, collect emitted fluorescence and capture cell images in a high-resolution and high-throughput manner. Various commercial or non-commercial software packages are available such as Metamorph (Molecular Devices Corporation), ImagePro (Media Cybernetics, Inc.), ImageJ,²⁴ CellProfiler,²⁵ and Cell-ID²⁶ to operate the microscopes, and to collect and analyze the images. Therefore, the combination of microfluidic cell docking with automated fluorescent microscope can provide a versatile platform for high-throughput single-cell quantification.

For high-throughput analyses of biological responses to environmental stimulations, DNA²⁷ and protein microarrays²⁸ have been frequently used because of their well-defined methodology. Similarly, cell-based microarrays are useful to probe gene function at cellular level in a highly parallel fashion.²⁹ To create cell-based microarrays, a robotic arm is typically used to print onto a surface that is compatible with cell attachment and proliferation. Because of the limitation in feature size of spots which is on the order of a few hundred micrometres, the resolution of cell cluster could not guarantee the single-cell resolution. In contrast with the robotic arm-based cell microarray, a microwell-based cell array system is able to address the limitation of inadequate spotting resolution. For example, sedimentation of cells into engraved microwells has been employed for its easy and simple implementation.^21,23 The sedimentation-based cell docking is especially useful for single-cell level docking in the case of relatively large cells (∼10 µm to ∼100 µm); if the cell size becomes less than ∼10 µm such as yeast or bacterial cells, however, the docking process is not efficient to obtain single cell arrays.

Recently, we have reported a simple cell docking method by exploiting a patterned microfluidic channel, surface tension driven cell seeding, and a receding meniscus.³⁰ Using this method, single to multiple yeast cells were accurately deposited onto microwells while a meniscus was receding towards the outlet that is generated by one-side natural evaporation. It turned out that subsequent cellular analysis was hindered by two major obstacles: (i) reduced cell viability with a long-term cell docking (>30 min) and (ii) flooding issues in the following steps of washing and filling medium. Here, we introduce an improved version of such meniscus-aided cell docking scheme by employing an active movement of cell solution plug by a finger pressure (instead of natural evaporation) and by optimizing geometrical and processing parameters with various channel dimensions (4–24 µm diameter, 8 µm depth) and cell densities (1.5–6.0 × 10⁹cells per mL). Furthermore, we combined the present cell docking method with high-throughput automated fluorescent microscopy to monitor cellular responses of the MAPK signaling pathways in the budding yeast, Saccharomyces cerevisiae. The MAPK signaling pathways are critical for many biological processes including cell growth, cell differentiation and survival in eukaryotes.^31,32 In S. cerevisiae, the pheromone response pathway, which is mediated by Fus3 MAPK signaling pathway, is highly homologous to the ERK MAPK signaling pathway in mammals.^33–38 The matingMAPK signaling pathway was analyzed using green fluorescent protein fused to signal specific promoter (P_fus1). The massive fluorescent cell images acquired by computerized microscope were processed using a commercially available image software package. Using this platform, we were able to monitor the real-time response patterns of the mating pathway at single-cell resolution.

Methods and materials

Yeast strains and materials

The S. cerevisiae strain used in this study was SH129 (MATa, leu2, trp1, met15, fus1::yEGFP-HIS3, gpd1::Tdimer2-URA3). A peptide corresponding to the mating pheromone, α-factor, was chemically synthesized using F-moc chemistry and was purified by HPLC. The SH129 cells were grown at 30 °C in YPD medium to an A₆₀₀ of ∼0.4 (1.5 × 10⁷cells per mL).

Fabrication of patterned microfluidic channels

We used UV-curable polyurethane acrylate (PUA) polymer (MINS-311RM, Minuta Technology Co. Ltd.) for fabricating microwells. The PUA microwell structures were fabricated onto glass substrate by capillary moulding.^39,40 Positive PDMS stamps with circular pillar patterns (4–24 µm diameter, 8 µm in height) were replicated from the negative silicon master. A little amount of the PUA polymer solution (∼2 µL) was carefully dropped onto the glass substrate, and the PDMS stamp was then immediately placed in conformal contact with the polymer solution. The assembly was subsequently exposed to UV (dose = 150 mJ cm⁻²) for a few tens of seconds to cure. The masters for microfluidic moulds had protruding (positive) features with the impression of microfluidic channels (1 mm in width and 100 µm in height). For the replicated PDMS microfluidic moulds, holes were punched through the inlets and the outlets as reservoirs. Each reservoir had a hole of 4 mm in diameter, which allows for easy injection of the cell suspension or mating pheromone, and sufficient area to remove the residual cell suspension around the reservoir and microfluidic channel with a tissue paper. Once the PUA microstructures were fabricated, the patterned substrate and PDMS microfluidic mould were plasma cleaned at the same time for 45 s (60 W, PDC-32G, Harrick Scientific Products Inc.). After plasma treatment, the microfluidic mould was carefully aligned on the patterned substrate, brought in conformal contact with the substrate and firmly pressed to form an irreversible seal.

Cell docking within patterned microfluidic channels

An overall schematic illustration of cell docking procedure is shown in Fig. 1a–f. The yeast cell solution was introduced into the patterned microfluidic channel by surface tension driven capillary flow due to hydrophilicity (a). To easily manipulate the generation and migration of a meniscus, the seeding amount was adjusted in such a way that it was 30–40% less than the actual volume of the channel. Since the microfluidic channel has the dimension of 1 mm in width, 100 µm in height and 10 mm in length, the seeding amount was about 0.6–0.7 µL (channel volume: 1.0 µL). After seeding the solution, a solution plug was located in the middle of the channel with distinct menisci at both ends (b).

Fig. 1 (a–f) Schematic illustration of the cell docking method within a patterned microfluidic channel. (a) Green and red fluorescent protein modified yeast cells are introduced into the channel. (b) A solution plug containing yeast cells is located in the middle of the channel with distinct menisci at both ends. (c and d) The meniscus is sweeping over the microwells by slightly pushing the PDMS cover. The whole procedure is repeated twice to ensure high-efficient cell docking. (e) Remaining cells are washed with the SC media via capillary action. (f) After washing process, the microfluidic channel is filled with the SC media. (g) An overview of the cell chip. The cell chip has one simple straight microfluidic channel and two punched reservoirs. The channel dimension is 10 mm in length, 1 mm in width, and 100 µm in height. (h) A representative SEM image of PUA microstructures fabricated onto glass substrate showing high-density PUA microwells (8 µm diameter, 8 µm depth, and diameter : spacing = 1 : 1). (i) A higher-magnification (×3500) SEM image showing well-defined PUA microwells with good physical integrity. Scale bar: 40 µm (h) and 20 µm (i).

To manipulate the solution plug, the opening of one side of reservoir was covered with a small PDMS piece and then a slight pressure was carefully applied with a finger, resulting in slow migration of the cell solution towards the opposite reservoir (c). After finishing this “sweeping” process, the other opening was covered with the same PDMS piece and the solution plug was pushed in the opposite direction (d). As the meniscus recedes over the microwells, the yeast cells were spontaneously captured into the microwells by lateral capillary force at single-cell level. By repeating these sweeping processes, the cell docking was completed. After cell docking, the remaining cells inside the channel and around the two reservoirs were washed by flowing synthetic complete (SC) media via capillary action from one side of reservoir to another and ultimately removed using a soft tissue paper (e). This step was repeated to completely evacuate the channel except for the medium and cells captured within microwells. After the washing process, the microfluidic channel was filled with the SC media (f).

Tracking cellular response to mating signals using programmable live-cell imaging

To initiate the mating response of MAPK signaling pathways, various concentrations (1 nM–100 µM) of mating pheromone (α-factor) were used. The α-factor was flowed into the channel from the inlet via capillary filling until it completely filled the inlet and outlet reservoirs. Subsequently, the cell chip was equipped with a live-cell imaging system (DeltaVision, Applied Precision Inc.) which allows for automated multi-position time-course imaging owing to a precisely motorized X–Y–Z stage. To operate the system, we used a commercial software package (softWoRx, Applied Precision Inc.) which is capable for programmed time-course experiments. 20 spatial points, each covering 49 microwells at a ×60 oil-immersed objective, were arbitrarily selected to track the cellular response for 2 hours. The differential interference contrasts (DICs), green (GFP) and red fluorescent (RFP) images were taken every 15 minutes (9 time points) at every 20 spatial points.

Image processing and extraction of quantitative information

The 20 sets of captured images, in which each set contains 9 DIC, 9 GFP, and 9 RFP images, were scaled to adjust with a reference time and respective spatial point intensity using the softWoRx program. For more extensive processing of images such as background subtraction, sharpen filtering, and extraction of fluorescence intensity, an ImagePro (Media Cybernetics, Inc.) script was used, which was composed of a series of measurements including average green and red fluorescence intensities, lengths of major and minor axes of eclipse, and cell area. We calculated total green and red fluorescence intensities by multiplying average intensity and area of cells using Excel (Microsoft Co.) scripts. The additional data were used for the compensation of slight measurement errors in the calculated area and data filtering (by the noise of average red intensity and calculated area).

Results and discussion

Microwell-based single-cell docking inside patterned microfluidic channel

To capture and analyze single cells, we fabricated a patterned microfluidic channel, in which the bottom of the microfluidic channel was physically modified with hollow microwells.⁴¹ An overview of the channel is shown in Fig. 1g along with representative SEM images of the PUA microwells (Fig. 1h and i). The cell chip had one simple straight channel and two punched reservoirs (g). The channel dimension was 10 mm in length, 1 mm in width, and 100 µm in height. The circular microwells presented here contained hollow cylindrical holes of 8 µm diameter and 8 µm depth (h), yielding a feature density of 3906 wells per mm² (∼400 000 wells per cm²) which is nearly the same with that of a typical GeneChip microarray (Affymetrix Inc.). The higher-magnification (×3500) SEM image shows well-defined PUA microwells with good physical integrity and edge definition (i).

Using the fabricated patterned channels, a small amount of the cell solution (0.6–0.7 µL) was introduced such that the channel was nearly filled while the inlet and outlet reservoirs were remained empty. By carefully pushing the PDMS cover on one side of reservoir, the solution plug was slowly migrating towards the other reservoir. In our previous report,³⁰ we observed that the optimum migration speed was 5–8 µm s⁻¹ for effective single-cell docking. Although we used an active finger pressure to push the liquid plug instead of passive natural evaporation, the speed was maintained in an adequate region, and the entire time for cell docking dramatically reduced to 5–10 minutes. The present docking mechanism is simple, efficient, easy to be automated, and more importantly enhances cell viability by minimizing undesirable high osmolarity through solvent evaporation.

In order to determine the optimal geometry of the microwells for high-efficiency cell docking and minimal flooding, various channel dimensions were tested with the diameter ranging from 4 to 24 µm for the same depth of 8 µm. Fig. 2 shows bright field images of the microwells at initial state, after cell seeding, and after washing, respectively, with various microwell dimensions. For 4 µm microwells, no cells were captured after the docking presumably due to small size of the microwell. Although the size of S. cerevisiae is relatively small (3–8 µm), the cell docking process appears to require some gap spacing where the solution first wets the wall and generates a lateral capillary force that is need to induce cell docking (a).³⁰ For 8 µm microwells, most of the microwells were filled with the cells at single-cell resolution (docking efficiency >90%), and the captured cells still remained even after washing step (b). As with the 8 µm microwells, larger microwells (12 to 24 µm microwells) enabled cell docking with a similar efficiency, but failed to provide the single-cell resolution with the deposition of multiple cells (c). Moreover, large microwells were observed to suffer from flooding in the subsequent washing step, so that not many cells remained at the final stage as can be seen from the optical images in Fig. 2d–f. This suggests that although larger microwells are capable of capturing cells at high efficiency, they can compromise single cell resolution as well as stable positioning of the captured cells.

Fig. 2 Bright field images of various microwells at initial state, after cell seeding, and after washing step, respectively: (a) 4, (b) 8, (c) 12, (d) 16, (e) 20, and (f) 24 µm. Scale bar: 40 µm.

To elaborate on the flooding effect in washing step, we performed additional experiments with a syringe pump to control the flow rates inside the microfluidic channel. Here, the sizes of box-shaped microwells were selected as 8 µm and 24 µm for the same depth of 8 µm, rendering the aspect ratio (diameter to depth) of 1 and 3, respectively. The flow velocity was controlled at 10, 100, and 1000 µm s⁻¹. In the case of 8 µm microwells, the captured cells slowly rotated at 10 µm s⁻¹ (ESI Movie S1a†), rotated more vigorously at 100 µm s⁻¹ (ESI Movie S1b†), and finally were flooded at 1000 µm s⁻¹ (ESI Movie S1c†). In the case of 24 µm microwells, the captured cells rotated actively at 10 µm s⁻¹ (ESI Movie S2a†), and were flooded at a relatively intermediate velocity of 100 µm s⁻¹ (ESI Movie S2b†). It is noted that actual solution velocities during the cell docking process including cell seeding, manipulating meniscus, and removing residual cell solution, are in the range of 5–200 µm s⁻¹, suggesting that the 8 µm microwells can sufficiently protect the flooding while nurturing the cells with enhanced mass transport. Based on these results, the microwells of 8 µm diameter and 8 µm depth would offer the best docking performance and thus the subsequent single-cell analysis is carried out using these microwells.

Next, we investigated the effect of cell density on optimal cell docking efficiency. To this aim, we prepared three different densities of cell solution by centrifuging the original A₆₀₀ of ∼0.4 (1.5 × 10⁷cells per mL) solution. The resulting densities were 1.5, 3.0, and 6.0 × 10⁹cells per mL, respectively. For each density, 0.6 µL of the cell solution was introduced into the microfluidic channel, and the cells were captured within the 8 µm microwells by the same procedure. Fig. 3a–c show the results of cell docking for each cell density. For the ×100 concentrated solution (1.5 × 10⁹cells per mL), the cells were rarely captured because of their lower population (a). When the ×200 concentrated solution was used, 40–60% of cells were captured into the microwells (b). For the highest cell density (6.0 × 10⁹cells per mL), most of the wells were filled with the cells at single-cell level, indicating that the cell seeding density should at least be higher than ∼5.0 × 10⁹cells per mL. Fig. 3d and e represent a snapshot image of the transient cell docking process and a typical image of the captured cells at single-cell resolution. As shown in Fig. 3d, some cells were already deposited by sedimentation in the course of cell seeding and the remaining microwells were spontaneously filled with the cells while the meniscus passes over the microwells. It is worthwhile noting that the curvature of meniscus is not easily seen from the figure, since the channel width is significantly increased to 1 mm to enable high-throughput single-cell analysis.

Fig. 3 (a–c) Bright field images of captured yeast cells at three different densities of cell solutions: (a) 1.5 × 10⁹cells per mL (×100 concentrated), (b) 3.0 × 10⁹cells per mL (×200 concentrated), and (c) 6.0 × 10⁹cells per mL (×400 concentrated). (d) A snapshot image of transient cell docking process, showing spontaneous capture of the cells by lateral capillary force. (e) A representative SEM image showing the captured cells at single-cell resolution over a large area. Scale bar: 24 µm.

Programmable time-course live-cell imaging to monitor the cellular response

For tracking the cellular response to mating signals in MAPK signaling pathways, the captured yeast cells were treated with α-factor of various concentrations (0.001, 0.01, 0.1, 1, 10, and 100 µM), which was introduced into the chip after the washing process. This experiment was performed using yeast cells in which genes for green and red fluorescent proteins were fused to response-specific promoters, P_fus1 and P_gdp1, respectively. The green fluorescence corresponds to mating response and the red fluorescence was used to identify the cell boundary in successive image processing steps because red fluorescence intensity corresponds to high osmolarity response and was relatively high enough for detection of cell periphery.

We arbitrarily selected 20 spatial points which are capable of covering 980 microwells for satisfying statistical significance. DIC for bright field images, yEGFP for green fluorescent (GFP) images and Tdimer2 for red fluorescent (RFP) images were taken every 15 minutes (9 time points) at every 20 spatial points. Fig. 4 shows representative images for 0, 30, 60, 90, and 120 minute time points in the case of 10 µM α-factor treatment. As shown in the first column (DIC), the captured cells were freely mobile within the microwells indicating well-circulation of the mating pheromone into the microwells. With the help of some un-captured cells that were slowly flowing inside the channel, we were able to validate an even distribution of mating pheromone inside the whole microfluidic channel. In the second column (yEGFP), at a basal state the green fluorescence was rarely shown, but as time goes by the GFP intensity was dramatically increased by triggering the mating response. In parallel, the RFP intensity (Tdimer2) was relatively strong even at a basal state, and was not triggered by mating pheromone because the mating signaling pathway is not related with the high osmolarity signaling pathways. In the merged and stitched images shown in the fourth and fifth columns, we observed diverse colours stemming from an overlap of green and red fluorescence, and non-uniform cellular response over a large area.

Fig. 4 Representative microscopic images for 0, 30, 60, 90, and 120 minute time points in the case of 10 µM α-factor treatment (DIC: bright field images, yEGFP: green fluorescent, and Tdimer2: red fluorescent images). The merged and stitched images show diverse colours from a mixture of green and red fluorescence.

Image analysis for high-throughput quantification of the cellular response

To extract quantitative information from the captured images, we performed additional steps which were composed of serial image processing procedures such as time and spatial scaling, background subtraction, and sharpen filtering. Because the fluorescence intensity is not absolute but a relative value, scaling each fluorescent image to the reference intensity level is of paramount importance for quantification of the fluorescence. Detailed image processing steps are described in ESI† along with a step-by-step procedure with a background subtraction step. These image processing and cell boundary identifying procedures were recorded to a macro for easy conduction at every 20 image sets which contains 9 GFP and RFP merged images per each set. Using the macro scripts, we measured the green and red fluorescence intensities of each identified cell objects. To compensate the defects especially in the case of abnormal identifications, the additional length of major and minor axes of eclipse was measured to calculate the eclipse area along with the area of identified cell boundary.

With the massive time-course measurements, the cellular response pattern to mating signals in MAPK signaling pathways was displayed by plotting each time-course data together. Fig. 5 represents the results of time-course measurements with x-axis for time (minute) and y-axis for yEGFP fluorescence intensity (arbitrary unit). Also, each inset indicates the normalized time-course average of green fluorescence intensity which means the ratio of each time point value to initial value. In the case of 1 nM α-factor (a), the mating pheromone was not capable to trigger the signal response for most cells. As the concentration of α-factor was increased to 10 nM (b), the mating signaling response became noticeable. As the concentration was further increased to 10 µM (c–f), the mating response was dramatically increased. When the concentration was higher than 10 µM, the final steady-state fluorescence intensity did not show notable difference probably due to the saturation of signaling capability (e and f). The normalized average of green fluorescence intensity presented in this study showed similar expression levels with typical bulk tests by flow cytometry (e.g., FACS), suggesting that the environment of the microfluidic cell chip ensures similar conditions for cell viability and proper reaction to mating signals (manuscript in preparation). As shown in Fig. 5, the yeast cells in the population generated a non-uniform response behavior in the matingMAPK signaling. This result means that individual time-course traces have a unique response pattern and kinetics.

Fig. 5 Time-course measurements of mating responses of individual cells at various concentrations of mating pheromone: (a) 1 nM, (b) 10 nM, (c) 100 nM, (d) 1 µM, (e) 10 µM, and (f) 100 µM, respectively. The x-axis indicates time (minute) and the y-axis indicates yEGFP fluorescence intensity (arbitrary unit). Each inset shows the normalized time-course average of yEGFP fluorescence intensity.

Conclusions

We have developed a simple yet robust microfluidic platform for quantifying time-course single-cell information in a high-resolution and high-throughput manner. The method combined simple microwell-based cell docking within a patterned microfluidic channel, with programmable time-course live-cell imaging and software-aided fluorescent image processing. Utilizing multiple sweeping processes of a cell-containing solution plug induced by a slight finger pressure, the yeast cells were efficiently arrayed in well-defined PUA microwells at single-cell resolution over a large area. To optimize the cell docking scheme, the dimension of the microwell and cell seeding density were determined by using various microwell sizes and different density of cell solutions. In addition, the effect of flooding was verified at various flow velocity conditions.

We confirmed that the 8 µm microwell (8 µm depth) was an optimal design for ensuring the single-cell resolution and protecting the flooding effect. Also, the cell density should at least be higher than 5.0 × 10⁹cells per mL to render a higher cell docking efficiency (>90%). By the programmable time-course live-cell imaging, each captured cells were repeatedly monitored at many different positions for massive fluorescence intensity data acquisition. And the massive data were analyzed through various image processing steps such as scaling, background extraction and so on. The resulting images were used to analyze the matingMAPK signaling responses of yeast cells stimulated by various concentrations of mating pheromone. Using this platform, we were able to monitor and quantify the real-time cellular responses and kinetics of the mating signaling in yeast cells. We hope that this simple microwell-based cell chip combined with the programmable time-course live-cell imaging potentially provides a valuable tool for high-throughput quantification of cellular response as well as for determining and analyzing the “signaling dynamics” instead of “noise” in heterogeneous responses.

Acknowledgements

This work was supported by the World Class University program on multiscale mechanical design (R31-2008-000-10083-0) and the Basic Science Research Program (2010-0027955) through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (MEST) to KYS. This work was also supported in part by the SRC program of MEST/KOSEF (R11-2005-009-02004-0) and Seoul R&BD Program (10543) to SHP.

References

1. S. B. Chen, Q. S. Zhang, X. Wu, P. G. Schultz and S. Ding, J. Am. Chem. Soc., 2004, 126, 410–411.
2. R. P. Hertzberg and A. J. Pope, Curr. Opin. Chem. Biol., 2000, 4, 445–451.
3. G. Sorg, H. D. Schubert, F. H. Buttner and R. Heilker, J. Biomol. Screening, 2002, 7, 11–19.
4. S. Matsuoka, B. A. Ballif, A. Smogorzewska, E. R. McDonald, K. E. Hurov, J. Luo, C. E. Bakalarski, Z. M. Zhao, N. Solimini, Y. Lerenthal, Y. Shiloh, S. P. Gygi and S. J. Elledge, Science, 2007, 316, 1160–1166.
5. K. Takahashi, K. Tanabe, M. Ohnuki, M. Narita, T. Ichisaka, K. Tomoda and S. Yamanaka, Cell, 2007, 131, 861–872.
6. M. Kaern, T. C. Elston, W. J. Blake and J. J. Collins, Nat. Rev. Genet., 2005, 6, 451–464.
7. M. B. Elowitz, A. J. Levine, E. D. Siggia and P. S. Swain, Science, 2002, 297, 1183–1186.
8. J. M. Raser and E. K. O'Shea, Science, 2004, 304, 1811–1814.
9. A. Colman-Lerner, A. Gordon, E. Serra, T. Chin, O. Resnekov, D. Endy, C. G. Pesce and R. Brent, Nature, 2005, 437, 699–706.
10. J. P. Nolan and L. A. Sklar, Nat. Biotechnol., 1998, 16, 633–638.
11. J. P. Nolan, S. Lauer, E. R. Prossnitz and L. A. Sklar, Drug Discovery Today, 1999, 4, 173–180.
12. G. Rabut and J. Ellenberg, J. Microsc. (Oxford, U. K.), 2004, 216, 131–137.
13. R. Pepperkok and J. Ellenberg, Nat. Rev. Mol. Cell Biol., 2006, 7, 690–696.
14. D. Di Carlo and L. P. Lee, Anal. Chem., 2006, 78, 7918–7925.
15. A. M. Skelley, O. Kirak, H. Suh, R. Jaenisch and J. Voldman, Nat. Methods, 2009, 6, 147–152.
16. D. Di Carlo, L. Y. Wu and L. P. Lee, Lab Chip, 2006, 6, 1445–1449.
17. K. Y. Suh, A. Khademhosseini, S. Jon and R. Langer, Nano Lett., 2006, 6, 1196–1201.
18. K. Y. Suh, J. Seong, A. Khademhosseini, P. E. Laibinis and R. Langer, Biomaterials, 2004, 25, 557–563.
19. J. Voldman, M. L. Gray, M. Toner and M. A. Schmidt, Anal. Chem., 2002, 74, 3984–3990.
20. C. J. Flaim, S. Chien and S. N. Bhatia, Nat. Methods, 2005, 2, 119–125.
21. J. R. Rettig and A. Folch, Anal. Chem., 2005, 77, 5628–5634.
22. A. Khademhosseini, J. Yeh, S. Jon, G. Eng, K. Y. Suh, J. A. Burdick and R. Langer, Lab Chip, 2004, 4, 425–430.
23. J. C. Love, J. L. Ronan, G. M. Grotenbreg, A. G. van der Veen and H. L. Ploegh, Nat. Biotechnol., 2006, 24, 703–707.
24. J. B. Sheffield, Microsc. Microanal., 2007, 13, 200–201.
25. A. E. Carpenter, T. R. Jones, M. R. Lamprecht, C. Clarke, I. H. Kang, O. Friman, D. A. Guertin, J. H. Chang, R. A. Lindquist, J. Moffat, P. Golland and D. M. Sabatini, GenomeBiology, 2006, 7, R100.
26. A. Gordon, A. Colman-Lerner, T. E. Chin, K. R. Benjamin, R. C. Yu and R. Brent, Nat. Methods, 2007, 4, 175–181.
27. M. Schena, D. Shalon, R. W. Davis and P. O. Brown, Science, 1995, 270, 467–470.
28. G. MacBeath, Nat. Genet., 2002, 32, 526–532.
29. J. Ziauddin and D. M. Sabatini, Nature, 2001, 411, 107–110.
30. M. C. Park, J. Y. Hur, K. W. Kwon, S. H. Park and K. Y. Suh, Lab Chip, 2006, 6, 988–994.
31. M. C. Gustin, J. Albertyn, M. Alexander and K. Davenport, Microbiol. Mol. Biol. Rev., 1998, 62, 1264–1300.
32. D. E. Levin and B. Errede, Curr. Opin. Cell Biol., 1995, 7, 197–202.
33. E. A. Elion, Curr. Opin. Microbiol., 2000, 3, 573–581.
34. I. Herskowitz, Cell, 1995, 80, 187–197.
35. J. Y. Hur, G. Y. Kang, M. Y. Choi, J. W. Jung, K. P. Kim and S. H. Park, Mol. Cells, 2008, 26, 41–47.
36. S. H. Park, A. Zarrinpar and W. A. Lim, Science, 2003, 299, 1061–1064.
37. R. J. Taylor, D. Falconnet, A. Niemisto, S. A. Ramsey, S. Prinz, I. Shmulevich, T. Galitski and C. L. Hansen, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 3758–3763.
38. P. J. Lee, N. C. Helman, W. A. Lim and P. J. Hung, BioTechniques, 2008, 44, 91–95.
39. K. Y. Suh, Y. S. Kim and H. H. Lee, Adv. Mater., 2001, 13, 1386–1389.
40. K. Y. Suh, M. C. Park and P. Kim, Adv. Funct. Mater., 2009, 19, 2699–2712.
41. S. M. Kim, S. H. Lee and K. Y. Suh, Lab Chip, 2008, 8, 1015–1023.

---

Footnotes

† Electronic supplementary information (ESI) available: Supplementary figure S1, supplementary movie sets S1, S2. See DOI: 10.1039/c0lc00114g

‡ Published as part of a LOC themed issue dedicated to Korean Research: Guest Editors: Professor Je-Kyun Park and Kahp-Yang Suh

§ These two authors equally contributed to this work.

---