Efficient cell capture in an agarose–PDMS hybrid chip for shaped 2D culture under temozolomide stimulation

Abstract

In this work, hybrid microfluidic devices were fabricated by assembling a polydimethylsiloxane (PDMS) mold with an agarose microarray to realize cell capture and patterning in precisely controlled spatial distribution. Microwells with diameter varying from 15 to 30 μm were formed on the agarose hydrogel surface at 15 μm to 40 μm spacing. Cells were efficiently captured in microwells with nearly 100% occupancy, thus achieving cell manipulation in a semi-quantitative manner. The size of the cell population captured on the microwell array is proportional to the patterning area. Further study revealed that the capture process was mainly regulated by fluid dynamics, where liquid was absorbed by the highly permeable agarose substrate and carried target cells into microwells. Our method spared the complex chemical modification steps, and the agarose substrate promised good biocompatibility. By designing PDMS channels with different geometrical layouts, various cell patterning geometries were easily created. Cell culture models with controllable pattern area and population size were successfully developed for a temozolomide stimulation study for as long as 2 days. This work should benefit the study of cancer developing niche and provide a powerful platform for the direct and continuous observation of cell dynamics under drug stimulation.

---

Introduction

Living organisms with structural and biological functionality are comprised of different cell types organized in ordered spatial distribution. The spatial information of cell distribution may give a hint to reveal the inside-body working mechanics, and also affect cellular biology in outside-body culture models.^1,2 For multiple cell type cultures in one system, researchers conventionally employ a specially made transwell dish to study the exchange of cellular factors and their effects on cell biology. This method physically separates different cell types on the two sides of a semi-permeable membrane and prevents cells migrating toward each other.^3–5 However, the transwell method only gives a general picture about intercellular communication and ignores more detailed factors that might influence cell culture results, such as spatial distribution of the cells.⁶ Precise control of cell position is significant in the outside-body mimicking of in vivo bio-system, because spatial manipulation at single cell level would provide an efficient method to construct organ mimicry in more complicated formats and make it more comparable to real organs.^7,8 Morphology imitation may be the next stage in bio-model construction and help researchers to better understand innate connection of different cell types.^9,10 Currently, the negligence of spatial information in cell culture model is due to lack of method for precise position control.

Microfluidic devices have been developed as efficient tools for small volume sample analysis and micro scale objects manipulation. The perfect match of microfluidic chip channel dimension with cell size, both of which vary from several to dozens micrometers, enables precise control of cell behavior and spatial position.^11,12 Since the invention of microfluidic droplet technology, scientists have been dedicated to explore the possibility of applying micro-emulsion technique to cell manipulation and cell differences profiling, especially in single cell analysis manners.^13–15 However, conventional cell encapsulation strategies through T-junction geometry or flow-focusing structure are unable to achieve the certainty of every droplet containing one single cell.^16,17 Normally, dispersion of cells into separate droplets is characterized by Poisson distribution, according to which single-cell droplets can be acquired in no more than 40% proportion.¹⁸ Poisson model is based on the presumption that there is no attraction or repulsion between cells and carrier units. Starting from this theory, many deterministic methods for single cell–single droplet production have been developed, but suffered either from low throughput or poor robustness.^19–22 Another cell manipulation strategy utilizing the flow dynamics in micro-region is microarray capture.^23,24 Generally, microarray cell capture is realized by employing a protruding geometry to stop cells in flowing stream^25–27 or a concave area for cell static docking.^28,29 Microarray method enables simultaneous cell localization and in vitro culture in one platform so that cells can be directly and continuously observed under microscopy with fluorescent probes. Besides, there are other methods realizing the selection and retrieval of certain cell types.^30,31 However, the cell capture efficiency is not always promised if certain surface chemical modification that enhances capture is absent.³² Chen³³ reported a surface modification of microwell array by functional nucleic acids, and the cell occupancy was greatly increased. In addition, the design of capture geometry requires detailed investigation of fluid dynamics and sometimes theoretical simulations, which is time and labor consuming.¹⁹

Patterning technique such as contact printing or other space confined methods has been in rapid development and applied for facile surface modification as well as building cell culture model,^25,34,35 with micro or even nano meter resolution.³⁶ On the other hand, hydrogel material is an ideal candidate for applications such as bio-stamp substrate or 3D cell culture matrix.³⁷ Among all those biocompatible hydrogels, agarose has been considered as an important biomaterial and one of the most frequently used gels for cell studies with tunable rigidity and great integrity.^38,39

Herein, microfluidic chip channel confining combined with agarose microwell array capture was employed to realize cell patterning in precisely controlled spatial distribution. Cells were efficiently captured in microwells with nearly 100% occupancy, thus achieving cell patterning in a semi-quantitative manner. The distance between two cell capture microwells could be as small as 15 μm. We found that the capture process was mainly regulated by fluid dynamics inside PDMS cap–agarose microwell system. As agarose gel gradually absorbed aqueous solution on its surface, cells were swept into microwells at the same time. This physical interaction essence spares the laborious chemical modification steps and increases the robustness of patterning. By designing channels with different geometrical layouts, various cell pattern geometries were easily created. Cell culture model with controllable pattern area and population size was successfully developed for anti-cancer drug stimulated culturing for as long as 2 days. This work should benefit the study of cancer developing niche and provide a powerful platform for direct and continuous observation of cell dynamics under drug stimulation.

Experimental

PDMS channel fabrication

Polydimethylsiloxane (PDMS) channel was fabricated by standard lithography method. Mask for lithography was designed by Adobe Illustrator CS4 and printed to transparent film. Then, photo resist SU-8 2050 (Micro Chem Corp., USA) was spin coated to silicon wafer at 1000 rpm. After 30 min baking in 65 °C oven, the wafer was covered with mask and exposed in UV light for 3 min. Next the wafer was baked again in 65 °C oven for 5 min and developed by developer solution. Then silicone elastomer (SYLGARD 184, Dow Corning, Germany) mixture with initiator at 10 : 1 ratio was poured to the wafer surface to form a cast of 2.5 mm thick and baked overnight in 65 °C. At last PDMS layer was peeled off and cut into uniform size. The PDMS channel for cell loading behavior study was 175 μm high, 3 mm wide and 15 mm long.

Agarose microwell array fabrication

Agarose microwell array was also fabricated by lithography method, except that high resolution chromium mask was used. Chromium mask was designed by AutoCAD 2012 and printed to soda lime glass (Shenzhen Qingyi Precision Mask Making Corporation, Shenzhen, China). SU-8 2050 photoresist was spin coated to silicon wafer at 1000, 1500, 2000, 4000 rpm, separately to achieve layers of different depths as 160, 120, 80 and 40 μm. Then the wafer was baked in oven at 65 °C for 45 min. After cooling in room temperature for 15 min, the wafer was processed to UV exposure against chromium mask for 7 min, and developed by developer. Agarose solution with 2% wt agarose powder (Beijing Dingguo Changsheng Biotechnology Co. Ltd., Beijing, China) was boiled by heater and immediately poured into dish containing wafer mold to form a cast of 2.5 mm thick. Then the dish was transferred to oven to slow down the temperature dropping and stored for 30 min. After that, the dish was preserved at room temperature for 3 hours for agarose gelation. Finally, agarose bulk was peeled from silicon wafer gently and cut into small pieces of 2 cm long and 1 cm wide.

Cell loading and capture on microwell array

PDMS chip was attached to agarose surface to cover the mirowell array. Then cell suspension of 7.9 μl with varied cell densities was loaded by pipette from PDMS channel inlet. After 5–10 min preservation, the suspension was completely absorbed by agarose gel. Then PDMS chip was removed from agarose surface. Cell pattern area was washed by PBS from pipette for 3 times to remove the undocked cells. For demonstration of cell patterning, cells were stained in advance before being added to the microchip.

Theoretical simulation

We carried out numerical simulation in COMSOL Multiphysics 4.2 (COMSOL Co., Ltd, Sweden) to study flow field in PDMS–agarose hybrid chip at the initial status when solution was added just to fulfill the chamber. Flow field model was constructed based on the laminar flow module, finite element method and equilibrium approximation. Diameter of microwell was set as 30 μm, depth and spacing both at 50 μm with the channel width of 1.7 mm, and height of 150 μm. The PDMS boundary was set as water resist and agarose surface was designed to be water permeable.

Fluorescence imaging and cell counting

Fluorescence images were obtained by Leica fluorescence microscopy (DMI4000B, Leica microsystem, Switzerland). Cell counting was accomplished by software Image-Pro Plus (Version 5.1.1.14, Media Cybernetics, USA).

Cell culture under temozolomide

After cell was captured by agarose microwell array, the gel was transferred to Petri dish and cultured with temozolomide (TMZ, analytical grade, Tianlishi Corp, Tianjin, China) conditioned culture medium. The agarose gel was fluorescence stained by live/dead kit (Invitrogen, category number: R37601) and processed to microscopy observation after 48 hours' culture.

Results and discussion

Fabrication of agarose microwell array

Agarose microwell array was fabricated through casting molding of agarose solution onto pre-designed SU-8 template (Fig. S1 and S2†). We used PDMS channels as the introduction pathway of cell suspension. And by designing channels of different geometrical layouts, we were able to control the area which was exposed to target cells (Fig. 1A). The cell capture performance of microwell array was excellent, and nearly every microwell was occupied by cells (Fig. 1C–E). Cells inside microwells were not washed away by the vigorous washing step, and we attributed this result to the barrier and protecting effect of side wall. We noticed that cells exhibited spherical morphologies on microwell array (Fig. 1D). We attributed this result to the surface property of agarose hydrogel. The weight of agarose skeleton accounted for only 2% of the whole hydrogel material, and therefore around cells were mainly water molecules, which resulted in no proper matrix for cell adhesion. In this manner, regionally confined cell patterning with clear boundaries between different cell types was realized (Fig. 1F). U251, HUVEC and U87 cells were employed to demonstrate the cell loading ability with cells confined at different regions on microwell array (Fig. 1G). Owing to the good biocompatibility of agarose hydrogel material, cells were continuously cultured and observed on microwell array for as long as 48 h (Fig. S3†).

Fig. 1 Cell patterned culture under drug stimulation (A) schematic of cell capture on agarose substrate and then being processed for drug stimulated culture. (B) Bright field photographs of agarose microwell array. (C) and (D) HUVEC array in agarose surface recorded at 24 h after capture. (E) Fluorescence image of HUVEC capture array. Cell was stained by calcein-AM. (F) Fluorescence image of three cell types localized patterning. HUVEC was stained by Hoechst 33342 (blue); U251 cell was stained with DiO (green); U87 cell was stained with DiI (red). (G) Regional pattern of different cell types. The first and fifth lines: HUVEC; the second and fourth lines: U87; the middle line: U251 cell. Microwells with diameter of 30 μm and depth of 80 μm were applied, and the spacing between two neighboring wells was 50 μm. Scale bar 200 μm.

Theoretical model

Previous works suggested that cell settlement into microwells could be characterized by Poisson distribution. In our system, the Poisson equation could not properly explain the cell behavior, thus we attempted to propose another mechanism for the cell capture. We found that the high efficiency of cell capture resulted from the unique hydro dynamics inside PDMS chamber bottomed by agarose substrate. COMSOL 4.2 was employed for numerical simulation of flow velocity field at the initial status when the addition of cell suspension was immediately stopped and PDMS chamber was just fulfilled. Agarose gel was highly water permeable and the liquid on gel surface could be completely absorbed in minutes. As demonstrated in Fig. 2A, nearly all of flow streams inside PDMS–agarose hybrid chip were directed into the microwells, which carried cells towards the microwells along with liquid uptake. This velocity field profiling represented an ideal situation which emphasized the water permeability of agarose microstructure and proposed a possible mechanism that might induce cell capture. Furthermore, cell settlement into microwell was studied in arrays of different materials. For the system of PDMS chamber with PDMS microwell array substrate, cells among were affected by gravity and buoyant force. Cell docking into microwells was induced by gravity sediment (Fig. 2B). Dominated by Poisson distribution, cell occupancy should never surpass 40%. For PDMS chamber with agarose microwell substrate, cells were influenced by additional flow dynamics related to agarose water absorption (Fig. 2C). It was suggested that the water permeability together with the microstructure design significantly increased the cell docking performance (Fig. S4 and S5†). Cell capture in PDMS–agarose hybrid system was further enhanced by the formation of air–liquid boundary. Air–liquid boundary formed and shrunk inwards as fluid was absorbed by hydrogel. For cells settling outside of the microwell, air–liquid boundary would sweep them into the inner liquid covered region, and provide chances of redistribution to unoccupied microwells (Fig. 2D–F and ESI movie†). This mechanism has been proposed by previous works and known as receding meniscus.^40,41

Fig. 2 Theoretical model of cell capture process on PDMS–agarose chip. (A) Numerical simulation of flow velocity. Simulation was carried out in COMSOL Multiphysics 4.2 (COMSOL Co., Ltd, Stockholm, Sweden). Stream lines were drawn in black color. Denser stream lines represent larger flow velocity. Cross section views of cell settlement to microwells of PDMS substrate (B) and agarose substrate (C). (D) Air–liquid boundary formed as agarose gel gradually absorbed the liquid on its surface. (E) Cells were piled up and driven forwards by the moving air–liquid boundary. (F) Bright field image of cells gathering on the air–liquid boundary. Red arrows indicate the boundary moving direction. Scale bar: 200 μm.

Study of factors influencing cell capture

Influencing factors of microwell occupancy and cell number per well were thoroughly studied. Firstly, microwell arrays with different diameters and spacings ranging from 15 μm to 40 μm were fabricated (Fig. S6 and S7†). We compared the performance of cell capture in bare agarose microwell array (Fig. 3A) with PDMS channel confined array (Fig. 3B). Capture without PDMS channel showed poor cell occupancy in various diameters and spacings, although in few combinations high occupancy was achieved (above 90%). In contrast, nearly 100% cell occupancy of microwells was realized for PDMS–agarose hybrid chip in all combinations of spacing from 25 to 40 μm and diameters from 15 to 30 μm. This result confirmed the necessity of PDMS channel for cell capture. By providing a PDMS ceiling, air–liquid boundary would form and migrate due to surface tension. However when diameters were larger than 30 μm, we noticed the appearance of bubbles in microwells, which would block cell capturing and reduce the occupancy. As cell suspension went gradually diluted, microwell occupancy could be kept at high level for several different cell densities. But when the density was too low (about 1.5 × 10⁶ ml⁻¹), cell occupancy dropped sharply (Fig. 3C).

Fig. 3 Study of factors influencing cell capture on agarose microwell array. HUVECs were used as model cells. (A) Microwell occupancy when cells were captured without PDMS confinement. (B) Microwell occupancy when cells were captured with PDMS confinement. (C) Microwell occupancy with different cell suspension densities. (D) Microwell occupancy with different microwell depths. (E) Count of cell per well with different microwell diameters and spacings. (F) Count of cell per well with different cell densities. (G) Count of cell per well with different microwell depths. (H) Frequency distribution of cell numbers in one microwell. The microwell array used in (H) was 20 μm in diameter and 35 μm in spacing (cell density of con.1 was 1.25 × 10⁷ ml⁻¹, and was in sequence diluted by 2 folds from con.1 to con.4. If not mentioned, cell density employed was con.1. Infused cell suspension volume was 7.9 μl for all experiments). Error bars were calculated from standard deviation of 3 parallel experiments.

Influences of microwell diameter, spacing and depth on cell number per microwell were also studied. The well diameter dominated the count of cell per well as indicated in Fig. 3E. It is easy to understand that larger microwells captured more cells, because volume exclusion effect defined the upper limit of cell accommodating capacity of microwells. However, well spacing was an ambiguous factor for cell number per well. Small spacing equaled larger density of microwells, which might provide stronger liquid absorption. On the other hand, higher density of microwells also meant more units for cells to distribute into, which resulted in less cell number per well. Therefore the cell number per well was both influenced by these two effects, and a proper selection of well spacing would lead to more cells in each well. As cell density decreased, cell number per well gradually decreased (Fig. 3F). But the depth of microwell didn't significantly influence the average cell number (Fig. 3G). We observed that cells were captured only in single layer despite of the microwell depth. This result could be explained by the barrier effect provided by microwell side walls to protect the bottom layer of cells from being washed away, but allowing other loosely packed cells to dislodge. We then took the 20 μm diameter and 35 μm spacing microwell array as an example to study the frequency distribution of cell counts in one microwell. Poisson fitting with the same mathematical expectation was used as a contrast in Fig. 3H. We found our frequency distribution (S.D. at 0.842) was focused on events that one microwell containing 2 or 3 cells and presented a narrower peak area compared to Poisson distribution (S.D. at 1.567). This result suggested that our method realized a more uniform cell capture.

Regionally confined patterning

Cell patterns with various geometries were formed on agarose substrate, defined by PDMS channels with different shapes. The patterning boundaries were clearly differentiated as shown in Fig. 4. Then we investigated the narrowest pattern width that can be formed. PDMS channels with different widths as 40 μm, 120 μm, 200 μm and 280 μm were fabricated and employed to define cell patterning area. The result indicated that the narrowest linear pattern with acceptable integrity was of 120 μm width (Fig. 4I–L). The width of 40 μm was too narrow for coherent pattern formation, because it reduced the injection volume and cell amount. Also the “wall effect” was significantly enlarged for narrow pattern, where cells were more likely to settle along the channel wall. All these factors led to less occupation of microwells.

Fig. 4 Shaped cell patterns (A)–(H) cell patterns of different morphologies on agarose microwell array. HUVECs were used as model cells, and stained with Hoechst 33342 (blue), DiO (green) and DiI (red) respectively. (I)–(L) Cell patterns of different widths. HUVEC was stained with Hoechst 33342; U251 was stained with DiO; U87 was stained with DiI. (M)–(P) Bright field images of agarose microwell array covered with PDMS channels of different widths (M) 40 μm, (N) 120 μm, (O) 200 μm and (P) 280 μm. Diameter of the microwells was 30 μm and the spacing between two neighboring wells was 50 μm. Scale bar: 200 μm.

Glioma-endothelial culture model under drug stimulation

U251 cells and HUVEC were chosen as the model cells to validate our method because they represent the principal cell types in glioma niche. Considering the vast disparities such as cell size and surface charges cross cell types, it is necessary to study and compare the capture performance of two cell types on microwell array before culture experiment (Fig. 5 and S9†). A gradient of cell densities was established in sequence dilution by 2 folds, among which con.1 was the densest and con.4 was the sparsest. The con.4 was 1.698 × 10⁶ ml⁻¹ for U251, and 1.849 × 10⁶ ml⁻¹ for HUVEC. As the suspension became sparser, cell number per well decreased for both cell types, but the decrease was much slower than the dilution folds. This variety could be explained by increased capture efficiency (which is defined as the ratio of cells captured against amount of cells added) in diluted suspension (Fig. S10†). We found that the cell count per well of U251 were comparable to HUVEC when densities of two suspensions were similar (with no more than 10% difference). And 100% cell occupancy was achieved for both cells when cell suspension density was high enough. Although there were certain differences, for example 100% occupancy of HUVEC was witnessed over a wider cell density gradient than U25, we could still conclude that U251 and HUVEC exhibited similar capture behaviors.

Fig. 5 Comparison of U251 cell and HUVEC capture behaviors. (A) and (C) Count of cell per microwell against successively diluted cell densities. (B) and (D) Cell occupancy of microwells against successively diluted cell densities. (A) and (B) was for U251 cells and (C) and (D) for HUVECs. For U251 con.4 was set as 1.698 × 10⁶ ml⁻¹; for HUVEC con.4 was 1.849 × 10⁶ ml⁻¹. Cell density was diluted by 2 folds in sequence from con.1 to con.4. Infused cell suspension volume was 7.9 μl for all experiments. (E) Different culture models of U251 and HUVEC. The letters M–Q represent models of different relative population size between two cell types; numbers 1–3 represent different position arrangements of two cell types. Diameter of the microwells was 30 μm with 80 μm in depth and the spacing between two neighboring wells was 50 μm. Error bars were calculated from standard deviation of 3 parallel experiments.

Then we studied the culture model of U251 and HUVEC under TMZ treatment and investigated the influencing factors such as relative population size and position between two cell types. TMZ was the most frequently used medicine for glioma chemotherapy, but in some cases TMZ was observed to oppositely aggravate tumor. Studying U251 and HUVEC culture under TMZ stimulation would provide a novel platform for drug resistance assessment. Here, we designed 9 culture models (M1, N1, N2, N3, O1, O2, O3, P1, P2, Q1) of different U251 and HUVEC population sizes and relative positions, as shown in Fig. 5E.

Firstly, we studied the effects of drug concentration and incubation time on cell viability when U251 cells and HUVECs were cultured alone on Petri dish (Fig. 6A and B). According to the previous studies of our group,⁴² TMZ concentration gradient of 0, 300, 600, 900 μM was established. It could be found that U251 was more vulnerable to drug stimulation. U251 viability reached about 95% when no TMZ was added. And increasing the applied TMZ concentration to 900 μM sequentially lowered the viability to below 70%. During the two days' incubation, the viability didn't reflect an evident tendency of increase or decrease, which might result from the cell proliferation and loss of dead cells. For HUVECs, the cell viability was high (90% and above) over all TMZ concentrations, and didn't indicate a time-dependent decrease either. The high viability of HUVEC may be due to their drug resistance. We determined the applied TMZ concentration and incubation time for following array culture experiments to be 900 μM and 48 h respectively. Different culture models were studied under TMZ stimulation, and viability of U251 and HUVEC in each model was analyzed and plotted in Fig. 6C. The result suggested that U251 cell viability was improved when together culturing with HUVECs under TMZ treatment. And when we adjusted the relative population size of HUVECs to U251, more endothelials (2 : 2 or 3 : 1) generally gave the better U251 viability than less ones (0 : 4 and 1 : 3). With live/dead staining, U251 cells in M1 model (Fig. 6E) contained more dead cells (red dots) than U251 in O1 (Fig. 6F). These results indicated that the HUVECs could improve the U251 resistance to chemotherapy, and help U251 to maintain higher viability. In addition, it suggested that there may be an optimized relative population size to realize the best viability of U251. As for HUVEC, the viability was also higher in the models of 2 : 2 and 3 : 1 HUVEC : U251 population size. Although HUVEC was less sensitive to TMZ itself, it was easily affected by the necrosis or apoptosis products from U251 when cultured with larger U251 population.

Fig. 6 Study of U251 and HUVEC culture model under temozolomide stimulation. (A) U251 cells viability on Petri dish for 48 h temozolomide conditioned culture. (B) HUVEC viability on Petri dish for 48 h temozolomide conditioned dish culture. (C) U251 and HUVEC viabilities of different culture models on microwell array after 900 μM temozolomide stimulation for 48 h. (D) Bright field image of U251/HUVEC culture platform. (E) and (F) Fluorescence image of U251 cells stained by live/dead reagent after 48 h temozolomide stimulated culture with or without HUVEC. (E) Corresponds to M1 culture model. (F) Corresponds to O1 culture model. Cell density of HUVEC used on agarose array culture was 1.47 × 10⁷ ml⁻¹, and U251 was 1.35 × 10⁷ ml⁻¹. Scale bar: 200 μm. Error bars were calculated from standard deviation of 3 parallel experiments, *p < 0.05.

Besides, the relative positions of U251 and HUVEC seemed to be an influencing but not dominating factor to cell viability, which could be told by the minor differences of cell viability in the models with same relative population size between HUVEC and U251 but different relative positions (for example O1, O2 and O3). This was quite reasonable, because our culture model allowed the free diffusion of cell factors and other substance through medium over the agarose surface.

Conclusions

In this work, we successfully developed a facile and effective method to realize cell patterning in predesigned area with controllable population size. Theoretical model of cell capture by agarose microwell array was established and studied by numerical simulation. The employment of physical interaction guaranteed the efficient and stable cell capture on agarose–PDMS chip, and spared the laborious chemical modification steps. After we deciphered the influencing factors of cell capture performance such as diameter, spacing and depth of microwells, cell patterning with various geometries was easily fabricated by physical capture.

Tumor-endothelial interaction was implicated in the development of cancer and failure of chemotherapy. Our method was applied to study the together culture model of U251 and HUVEC under TMZ stimulation with various relative cell population sizes and patterning positions between these two cells types. Results suggested that HUVEC enhanced the drug resistance of U251. The PDMS–agarose hybrid chip provides an efficient tool for cell manipulation and may be further applied in other cancer related studies.

Acknowledgements

This work was supported by National Natural Science Foundation of China (No. 81373373, 21435002, 21227006).

Notes and references

1. D. Huh, B. D. Matthews, A. Mammoto, M. Montoya-Zavala, H. Y. Hsin and D. E. Ingber, Science, 2010, 328, 1662–1668.
2. J. Zhang, J. Wu, H. F. Li, Q. S. Chen and J. M. Lin, Biosens. Bioelectron., 2015, 68, 322–328.
3. Q. S. Chen, J. Wu, Q. C. Zhuang, X. X. Lin, J. Zhang and J. M. Lin, Sci. Rep., 2013, 3, 2433.
4. I. Wagner, E. M. Materne, S. Brincker, U. Sussbier, C. Fradrich, M. Busek, F. Sonntag, D. A. Sakharov, E. V. Trushkin, A. G. Tonevitsky, R. Lauster and U. Marx, Lab Chip, 2013, 13, 3538–3547.
5. D. Sticker, M. Rothbauer, S. Lechner, M. T. Hehenberger and P. Ertl, Lab Chip, 2015, 15, 4542–4554.
6. C. T. Ho, R. Z. Lin, R. J. Chen, C. K. Chin, S. E. Gong, H. Y. Chang, H. L. Peng, L. Hsu, T. R. Yew, S. F. Chang and C. H. Liu, Lab Chip, 2013, 13, 3578–3587.
7. B. Derby, Science, 2012, 338, 921–926.
8. G. Villar, A. D. Graham and H. Bayley, Science, 2013, 340, 48–52.
9. D. Huh, Y. S. Torisawa, G. A. Hamilton, H. J. Kim and D. E. Ingber, Lab Chip, 2012, 12, 2156–2164.
10. A. M. Ghaemmaghami, M. J. Hancock, H. Harrington, H. Kaji and A. Khademhosseini, Drug Discovery Today, 2012, 17, 173–181.
11. E. K. Sackmann, A. L. Fulton and D. J. Beebe, Nature, 2014, 507, 181–189.
12. D. J. Collins, B. Morahan, J. Garcia-Bustos, C. Doerig, M. Plebanski and A. Neild, Nat. Commun., 2015, 6, 8686.
13. J. Q. Yu, L. K. Chin, L. Lei, Z. P. Lin, W. Ser, H. Chen, T. C. Ayi, P. H. Yap, C. H. Chen and A. Q. Liu, Lab Chip, 2014, 14, 3519–3524.
14. S. Akbari and T. Pirbodaghi, Lab Chip, 2014, 14, 3275–3280.
15. G. S. Du, J. Z. Pan, S. P. Zhao, Y. Zhu, J. M. J. den Toonder and Q. Fang, Anal. Chem., 2013, 85, 6740–6747.
16. K. Liu, Y. L. Deng, N. G. Zhang, S. Z. Li, H. J. Ding, F. Guo, W. Liu, S. S. Guo and X. Z. Zhao, Microfluid. Nanofluid., 2012, 13, 761–767.
17. L. Mazutis, J. Gilbert, W. L. Ung, D. A. Weitz, A. D. Griffiths and J. A. Heyman, Nat. Protoc., 2013, 8, 870–891.
18. D. J. Collins, A. Neild, A. deMello, A. Q. Liu and Y. Ai, Lab Chip, 2015, 15, 3439–3459.
19. K. Zhang, X. Han, Y. Li, S. Y. Li, Y. L. Zu, Z. Q. Wang and L. D. Qin, J. Am. Chem. Soc., 2014, 136, 10858–10861.
20. J. C. Baret, O. J. Miller, V. Taly, M. Ryckelynck, A. El-Harrak, L. Frenz, C. Rick, M. L. Samuels, J. B. Hutchison, J. J. Agresti, D. R. Link, D. A. Weitz and A. D. Griffiths, Lab Chip, 2009, 9, 1850–1858.
21. E. W. M. Kemna, R. M. Schoeman, F. Wolbers, I. Vermes, D. A. Weitz and A. van den Berg, Lab Chip, 2012, 12, 2881–2887.
22. M. Chabert and J. L. Viovy, Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 3191–3196.
23. Y. Qiao, J. C. An and L. Y. Ma, Anal. Chem., 2013, 85, 4107–4112.
24. K. R. Love, S. Bagh, J. Choi and J. C. Love, Trends Biotechnol., 2013, 31, 16–22.
25. K. Zhang, C. K. Chou, X. F. Xia, M. C. Hung and L. D. Qin, Proc. Natl. Acad. Sci. U. S. A., 2014, 111, 2948–2953.
26. W. M. Liu, J. C. Wang and J. Y. Wang, Lab Chip, 2015, 15, 1195–1204.
27. D. Di Carlo, N. Aghdam and L. P. Lee, Anal. Chem., 2006, 78, 4925–4930.
28. J. C. Love, J. L. Ronan, G. M. Grotenbreg, A. G. van der Veen and H. L. Ploegh, Nat. Biotechnol., 2006, 24, 703–707.
29. W. Y. Xie, D. Gao, F. Jin, Y. Y. Jiang and H. X. Liu, Anal. Chem., 2015, 87, 7052–7059.
30. M. G. Lee, J. H. Shin, C. Y. Bae, S. Y. Choi and J. K. Park, Anal. Chem., 2013, 85, 6213–6218.
31. H. W. Hou, M. E. Warkiani, B. L. Khoo, Z. R. Li, R. A. Soo, D. S. W. Tan, W. T. Lim, J. Han, A. A. S. Bhagat and C. T. Lim, Sci. Rep., 2013, 3, 1259.
32. K. A. Mosiewicz, L. Kolb, A. J. van der Vlies, M. M. Martino, P. S. Lienemann, J. A. Hubbell, M. Ehrbar and M. P. Lutolf, Nat. Mater., 2013, 12, 1071–1077.
33. Q. S. Chen, J. Wu, Y. D. Zhang, Z. Lin and J. M. Lin, Lab Chip, 2012, 12, 5180–5185.
34. S. Gobaa, S. Hoehnel, M. Roccio, A. Negro, S. Kobel and M. P. Lutolf, Nat. Methods, 2011, 8, 949–955.
35. N. Tanaka, H. Ota, K. Fukumori, J. Miyake, M. Yamato and T. Okano, Biomaterials, 2014, 35, 9802–9810.
36. T. Kraus, L. Malaquin, H. Schmid, W. Riess, N. D. Spencer and H. Wolf, Nat. Nanotechnol., 2007, 2, 570–576.
37. T. Billiet, M. Vandenhaute, J. Schelfhout, S. Van Vlierberghe and P. Dubruel, Biomaterials, 2012, 33, 6020–6041.
38. J. Dahlmann, G. Kensah, H. Kempf, D. Skvorc, A. Gawol, D. A. Elliott, G. Drager, R. Zweigerdt, U. Martin and I. Gruh, Biomaterials, 2013, 34, 2463–2471.
39. R. Novak, Y. Zeng, J. Shuga, G. Venugopalan, D. A. Fletcher, M. T. Smith and R. A. Mathies, Angew. Chem., Int. Ed., 2011, 50, 390–395.
40. M. C. Park, J. Y. Hur, H. S. Cho, S. H. Park and K. Y. Suh, Lab Chip, 2011, 11, 79–86.
41. M. C. Park, J. Y. Hur, K. W. Kwon, S. H. Park and K. Y. Suh, Lab Chip, 2006, 6, 988–994.
42. C.-H. Lin, X.-X. Lin, L. Lin, J.-M. Wang, Z.-X. Lin and J.-M. Lin, Chin. Chem. Lett., 2015, 26, 1225–1230.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra15734c

---