Concentration gradient generator for H460 lung cancer cell sensitivity to resist the cytotoxic action of curcumin in microenvironmental pH conditions

Abstract

We proposed and demonstrated a concentration gradient generator (CGG) to resist H460 lung cancer cells using curcumin in microenvironmental pH conditions. The CGG was designed as the integration of one-by-two and one-by-three fluidic channels to effectively generate a different concentration gradually. In cell biology, this CGG device was realized to test its operability and practicability. Therein, H460 lung cancer cells could be restrained by the injection of different curcumin concentrations. At a pH = 6.5 microenvironment, the apoptosis rate (AR) of H460 lung cancer cells was not obvious until the curcumin concentration was increased to 57.2 μg mL⁻¹, at which the AR was 47.8%. At pH = 7.4 and 8.5 microenvironments, the ARs of the H460 lung cancer cells were 54.5% and 93.8%, respectively, at a curcumin concentration of 57.2 μg mL⁻¹. In the range of curcumin concentration of 57.2–100 μg mL⁻¹, the average ARs of the H460 lung cancer cells were 53.6%, 58.4%, and 98.0% for pH = 6.5, 7.4, and 8.5 microenvironments, respectively. The resistant efficiency of H460 lung cancer cells could be enhanced substantially at a pH = 8.5 microenvironmental condition. These results pave a way for the use of CGGs in cell biology application.

---

Introduction

Recently, the various controls and quantifications of gradient generation have attracted great attention for biomedical applications such as cell biology,^1–4 tissue regeneration,⁵ and drug screening.^3,6 This is because the gradients play fundamental roles in biological activities and in vivo cellular functions⁷ including cell differentiation⁸ and the immune response,^9,10 as well as cancer, which is affected by the chemical gradients during growth and development.¹¹ The microfluidic-based concentration gradient generator (CGG) provides an opportunity to integrate gradient generation and cell culture^6,12 to mimic the in vivo cellular microenvironment. In general, CGGs can be classified into two categories: static CGGs^13–19 and dynamic CGGs.^12,20–29 The former is dependent on molecular diffusion in a static fluid, which is simple without an extra driving force and can generate a continuous concentration gradient. However, the gradient is usually unpredictable, uncontrollable and has bad durability such that it can be easily changed by spatial and temporal evolution. The latter is formed by diffusion in a flowing laminar flow with a driving force. This kind of CGG has been extensively used because it can generate a stable and predictable gradient. In view of this, such CGGs, integrated with cell culture chambers (CCCs), are frequently utilized in drug screening and chemical reaction optimization to control more factors in cell biology or chemical reactions.^30,31 The “Christmas tree” structure is one of the most conventional CGGs in biomedicine applications,³⁰ but the major disadvantage of these conventional CGGs is their low efficiency of gradient generation.²⁹ To date, the design of a CGG integrated with a CCC in microfluidic chips for biomedicine applications has not been easy to control and operate because both of them are on the same plane. The cells can thus easily flow into the CGG channels when introducing cells.^32,33 Therefore, it is necessary to make the CGG and CCC on different planes to prevent this unwanted situation.

For CGG design, it is crucial to characterize the concentration gradient profiles. Currently, the most common methods to characterize the concentration gradient profile has involved generating a set of concentration gradients of a fluorescent dye, and then taking fluorescent photographs to analyze the corresponding concentration gradient.^20,29,34,35 However, the light intensity of the background and the instantaneous record of the concentration gradients in different areas of the fluorescent photograph are usually different. The fluorescent image thus displays a higher deviation in the concentration gradient characterization. To solve this problem, the use of an optical absorbance gradient solution was proposed to characterize the concentration gradient profiles.³⁶

In this study, we proposed and demonstrated an effective way to generate a concentration gradient, and then further realized its operability and practicability in cell biology tests. In addition, responses of the H460 lung cancer cells to anticancer drug concentration gradients and microenvironmental pH conditions were also characterized.

Experiment

Design and fabrication of the CGG

Our CGG structure was composed of two inlets (diameter: 0.5 mm, depth: 3 mm), eight outlets (diameter: 0.5 mm, depth: 3 mm), and a microchannel network (width: 0.1 mm, depth: 45 μm). The fluidic streams could be repeatedly split and mixed to generate a gradient. The microchannel network was tailored as one-by-two (Design_1) and one-by-three (Design_2) fluidic channels, respectively, as shown in Fig. 1. It can be seen that in Design_2, one fluidic stream is split into three fluidic branches and then both sides of the fluidic branches are mixed with the adjacent one, respectively, to recombine into a new fluidic stream. Simultaneously, the middle fluidic branch forms a new fluidic stream itself directly. One fluidic stream can thus be split into more branches at once compared to in Design_1. Therefore, the designed CGG device in this study can generate multiple concentration gradients efficiently. To utilize this thought of alternative arrangements of Design_1 and Design_2, various CGGs with optional numbers of fluidic channels with different concentration gradients can be obtained.

Fig. 1 Schematic of the CGG composed of two inlets, eight outlets, and a microchannel network.

The fabrication process of the CGG was as follows. First, a prepared silicon mold with a micrometer-sized SU-8 was exposed to chlorotrimethylsilane vapor for 3 minutes, and then released after carrying out the baking process. Secondly, a PDMS prepolymer and curing agent (8 : 1 w/w, RTV615A, RTV615B) were mixed in an 8 : 1 ratio and then poured onto the mold. Thirdly, the sample was degassed under vacuum for 30 minutes and then baked for 45 minutes at 85 °C. Fourthly, the cured PDMS layer with the desired structures was peeled and punched using a metal pin at the terminals of the inlet and outlet channels. Finally, the PDMS structure was bonded on a glass slide.

Characterizations of CGG

In this session, the performance of the CGG was characterized and discussed. The relative characterizations were measured by using a spectrophotometer (Ocean Optics Co., Maya 2000/2000 PRO). Ink solutions (Canon Co., GI-890C) and deionized (DI) water were simultaneously introduced into the inlets at the same flow rate (4 μL min⁻¹) by syringe pump (RISTRON Co., RSP02-B). In the Design_1 structure, the upstream fluid was split into two fluidic branches. Each of the branches met and mixed with the adjacent branch in equal volume and then formed a new concentration solution in a serpentine channel, which was also the same process occurring with both sides of the Design_2 structure. The mixed fluids were diffused in a serpentine channel and then formed an intermediate concentration solution. The middle fluidic branch of the Design_2 structure was kept at the same concentration to that of the upstream fluid. The formation of a new concentration solution will then repeat the separation and mix process in the microchannel network, and can be extracted from the outlets to obtain a set of ink concentration gradient solutions. A schematic of the system setup is shown in Fig. 2. The performance of the CGG is presented by comparing the different concentrations of ink solutions, as measured by using a spectrophotometer to distinguish the tiny absorbance between the different concentrations of ink solutions. Here, the fluidic concentration was proportional to the absorbance in a fixed optical path according to the Beer–Lambert law.³⁷ Therefore, the concentration gradient solutions were extracted in eight centrifuge tubes, and their optical absorption spectra were measured and are shown in Fig. 3(a). When comparing the characterizations as analyzed by using the fluorescent spectrometer, it can be seen that the image is impacted by the ambient light intensity in the different areas of the fluorescent photograph. Also, the image pertains only to the instantaneous gradient at the exact time when the image was taken. However, by using a spectrophotometer, it is possible to simultaneously introduce resource solutions into the CGG and to measure the corresponding absorbance of the concentration gradient solutions. Therefore, this can overcome the problem caused by the use of a fluorescent spectrometer and the resulting measurement deviation. As shown in Fig. 3(b), the experimental absorbance intensities of the concentration gradient solutions are comparable to the theoretical absorbance intensities.

Fig. 2 Schematic of the system setup.

Fig. 3 Characterizations of the CGG. (a) Optical absorption spectra of eight concentration gradient solutions generated by blue ink and DI water. (b) Comparison of the theoretical and experimental absorbance intensities measured by a spectrophotometer.

CGG device for the cell apoptosis assay

To further study the responses of tumor cells to the anticancer drug concentration gradient stimulation, we designed a CGG device composed of CGG and CCC. This CGG device was assembled with three layers: CGG, CCC and glass substrate, respectively, as shown in Fig. 4(a). The geometric sizes of the CCC were as follows: the lengths of the major axis and minor axis of the eight oval structures (observation windows) were 3 mm and 1 mm, respectively, with a depth of 45 μm. The cell inlet and outlet were 2 mm in diameter and 3 mm in depth. The CGG and CCC were located in different layers via eight channels with 1.5 mm diameter and 1 mm depth as indicated by the sapphirine color shown in Fig. 4(b). This design was to prevent the cells flowing into CGG channels and to guarantee that they were unobstructed owing to the fluidic directions of the cells and the concentration gradient solutions located in the different layers, as indicated by the red (cells) and green (concentration gradient solutions) arrows in Fig. 4(c). To verify the characterizations of the CGG device, different inks were introduced into the CGG device, as shown in Fig. 5. Symmetrical overlapping concentration gradients were generated by the two inks, such as the blue and yellow, blue and red, and red and yellow inks shown in Fig. 5(a–c), respectively.

Fig. 4 (a) Schematic of the CGG device. (b) Enlarged image of the vias between CGG and CCC. (c) Cross-sectional view of the CGG device. The red and green arrows are the fluidic directions of the cells and concentration gradient solutions, respectively.

Fig. 5 Photographs of the CGG device with (a) blue and yellow, (b) blue and red, and (c) red and yellow inks concentration gradients.

Cell culture in three pH microenvironments

The human non-small H460 lung cancer cell line was kindly donated by the Key Laboratory for Nano–Bio-Interfaces, Chinese Academy of Sciences. H460 lung cancer cells were cultured at the Roswell Park Memorial Institute (RPMI) in 1640 medium (Hyclone Co., RPMI-1640) supplemented with 10% fetal bovine serum (FBS), 100 U mL⁻¹ penicillin, and 100 U mL⁻¹ streptomycin, and then stored in an incubator with 5% CO₂ at 37 °C for 2–3 days. The cells were harvested by trypsin for 25 seconds at 37 °C to obtain the cells suspension, when they reached 80% confluence in the flask. Sequentially, the supernatant was removed by centrifugation and the precipitated cells were resuspended in fresh medium at a density of 5 × 10⁶ cells per mL. The channels of CCC were flushed with phosphate buffer saline (PBS) to remove bubbles before seeding the cells in CCC, and then the medium was introduced into CCC to balance the culture environment for 30 minutes. The suspended cells were dropped into the CCC inlet and extracted from the CCC outlet by using negative pressure, as shown in Fig. 6(a). The merit of this introducing procedure was that the suspended cells could be controlled with a slow flow rate to stop in CCC uniformly. Therefore, it could prevent the cells being deposited at the bottom of the syringe by using positive pressure to introduce the cells into the CCC. The CCC with seeding cells was stored in a CO₂ incubator, allowing the cells to adhere and spread for a duration of 4–5 hours. For cell perfusion culture, a pipette tip was inserted into the CCC inlet as a simple reservoir and then supplemented with 300 μL fresh medium in the reservoir with a cycle time of 12 hours, as shown in Fig. 6(b). In this experiment, the cells were cultured in three pH (6.5, 7.4, and 8.5) microenvironments, respectively, by changing the pH value of the medium, which was immediately adjusted with HCl and NaOH solutions with a 1 mol mL⁻¹ concentration.

Fig. 6 Experimental procedures of cell culture in the CGG device. (a) Suspended cells were dropped into the CCC inlet and extracted from the CCC outlet. (b) Supplemented fresh medium in the CCC inlet was added via a pipette tip for the cell perfusion culture.

Cell apoptosis assay in three pH microenvironments

The cells were cultured for 24 hours in the CGG device, and were then stimulated by different curcumin concentration gradients. The media with and without curcumin were simultaneously introduced into two inlets of CGG with a continuous flow rate of 0.5 μL min⁻¹ in a CO₂ incubator. The device was taken out from the CO₂ incubator after 24 hours. The fluorescent dye (Life Technologies Co., LIVE/DEAD Viability/Cytotoxicity Kit) was dissolved in PBS and then 150 μL was introduced into the device via the CGG inlets to stain the live cells with green fluorescence and the dead cells with red fluorescence. The cells were incubated in a dark environment for 30 minutes at 37 °C and then observed under a fluorescence microscope (Nikon Co., Eclipse Ti-U) for characterization of the cell apoptosis assay. The cells fluorescence photographs were taken by a CCD camera (Nikon Co., DS-Fi2/DS-U3) and analyzed by using Image Plus-Pro software. Herein, cell apoptosis assays were performed by introducing curcumin solutions with three pH (6.5, 7.4, and 8.5) values into the CGG device. The pH value of the medium with and without curcumin was adjusted immediately.

Results and discussions

Consideration in the design and characterization of the CGG

For the realization of an in vivo cellular microenvironment, CGG was designed to have eight concentration gradient channels connected with eight vias to eight observation windows of the CCC. The microchannel network of the CGG was composed of Design_1 and Design_2. In comparing the microchannel networks only composed of Design_1 or Design_2, Design_1 was a conventional CGG with two inlets that could generate eight concentration gradient solutions, which required six levels of horizontal channels, while Design_2 only required three levels of horizontal channels for eight concentration gradient solutions. Design_2 was expected to generate concentration gradient solutions efficiently and to allow shrinking of the device size, i.e., saving generation time of a concentration gradient. The saving time was 48.48% compared to that of Design_1. However, as there were nine outlets when using Design_2, this would imply a waste of one outlet. This is a tradeoff to save the generation time and fluidic volume. In this study, the combination of Design_1 and Design_2 in the CGG design could generate eight concentration gradient solutions and could save 33.33% of the generation time compared Design_1 alone and save 11.11% of fluidic volume compared Design_2 alone, respectively.

Considerations in the design of the CGG device

CGG is a conventional platform to study the cell response to chemical concentration gradients. The integration of CGG and CCC could greatly simplify the experimental procedure of the cell assay. In the past decade, many researchers have utilized this integration device to study cell responses to chemical concentration gradients. The major disadvantage has been that CCC and CGG were typically located on the same plane, whereby cells would easily flow into the CGG and then grow in it when they were introduced into the CCC. Therefore, the generation of a concentration gradient would be affected if cell growth and reproduction happened in the CGG. In this study, a CGG device was proposed and demonstrated with CGG and CCC on different planes. CGG was a top layer above CCC but connected by eight vias to guarantee that the concentration gradient solutions indeed flow into CCC and to prevent the cells flowing into the CGG.

Cell apoptosis assay induced by different curcumin concentration gradients in the three pH microenvironments

Currently, it has been demonstrated that curcumin could induce H460 lung cancer cells apoptosis.^38–40 It has reported that the acidic microenvironment of tumors plays an important role in cancer evolution and progression, especially in drug resistance.^41–45 Solid tumors involve extracellular acidosis, with extracellular pH values as low as 6.5.⁴⁶ In in vitro tumor experiments, the pH value of the medium is around 7.4 generally, but this pH value is not consistent with that of the tumor microenvironment in vivo. Therefore, tumor cells cultured in this environment cannot accurately reflect the actual condition of tumor cells in vivo. Therefore, we observed the responses of H460 lung cancer cells in acidic microenvironments (pH = 6.5) compared to neutral (pH = 7.4) and alkaline (pH = 8.5) microenvironments. In this study, the resistance of H460 lung cancer cells was studied for the response to curcumin concentration gradients in pH = 6.5, 7.4, and 8.5 microenvironments by using a CGG device. Due to the design of the CGG device, the H460 lung cancer cells' response to the curcumin concentration gradient in one microenvironmental pH condition could be obtained and used to guarantee that the experimental results of the eight observation windows were also in agreement. The fluorescent photographs of the H460 lung cancer cells were taken at the observation windows of the CCC to analyze the apoptosis rate (AR) under different curcumin concentrations in pH = 6.5, 7.4, and 8.5 microenvironments, respectively, as illustrated in Fig. 7. At lower curcumin concentration (i.e., below 28.6 μg mL⁻¹), the viability of live H460 lung cancer cells was obvious in the three tested pH microenvironments. In increasing the curcumin concentration, the viability of the live H460 lung cancer cells gradually disappeared. This indicated that the generation of a curcumin concentration gradient by the CGG was successfully and had affected the H460 lung cancer cells. The relationships of the H460 lung cancer cells' AR to the different curcumin concentration gradients in the three pH microenvironments are summarized in Fig. 8. H460 lung cancer cells could be effectively resisted under the curcumin concentration of 42.9 μg mL⁻¹ in the pH = 8.5 microenvironmental condition, in which the AR was 61.7% higher or 4.71-fold and 1.17-fold than that in the pH = 6.5 (AR = 13.1%) and 7.4 (AR = 52.9%) microenvironments, respectively. At pH = 6.5, the AR was not obvious until the curcumin concentration increased to 57.2 μg mL⁻¹, in which the AR was 47.8%. At pH = 7.4 and 8.5, the ARs were 54.5% and 93.8%, respectively, at the curcumin concentration of 57.2 μg mL⁻¹. The endurance capability of the H460 lung cancer cells to the curcumin concentration gradient was highest in the pH = 6.5 microenvironment (where the average maximum AR was 53.6%, in the range of 57.2 μg mL⁻¹ to 100 μg mL⁻¹), while it was lowest in the pH = 8.5 microenvironment (where the average maximum AR was 98.0%, in the range of 57.2 μg mL⁻¹ to 100 μg mL⁻¹). At pH = 7.4, the average maximum AR was 58.4%, in the range of 57.2 μg mL⁻¹ to 100 μg mL⁻¹. These results indicate that it was more beneficial for H460 lung cancer cells living in the pH = 6.5 microenvironment to resist curcumin stimulation. The resistant efficiency of H460 lung cancer cells in the pH = 8.5 microenvironment could be enhanced 1.68-fold and 1.83-fold on average compared to in the pH = 7.4 and pH = 6.5 microenvironments, respectively. This means that the cytotoxic action to H460 lung cancer cells at pH = 8.5 is higher when the concentration of curcumin is more than 57.2 μg mL⁻¹. Here, the AR is almost 100%. Each condition was derived and we took the average of three test data results as the error bar, as shown in Fig. 8. These results prove the influence of the anticancer drug concentration and pH condition on the cancer cells resistance in the extracellular microenvironment by using a CGG device. This would be very useful for clinical applications in anticancer treatment. In this study, the research into the CGG device demonstrated it to be a viable platform to control multi-factors in the extracellular microenvironment. We believe this study could be further applied to study other cancer cells resistance by using CGGs.

Fig. 7 Fluorescent photographs of H460 lung cancer cells apoptosis induced by different curcumin concentration gradients in (a) pH = 6.5, (b) pH = 7.4, and (c) pH = 8.5 microenvironments.

Fig. 8 The relationships of H460 lung cancer cells AR to different curcumin concentration gradients in the pH = 6.5 (blue curve), 7.4 (red curve), and 8.5 (black curve) microenvironments.

Conclusions

In conclusion, we demonstrated an effective approach to realize a concentration gradient and investigated the resistance of H460 lung cancer cells in microenvironmental pH conditions by using a CGG device. The concentration gradient profiles were characterized using a spectrophotometer to observe the corresponding optical absorbance. This method was more accurate and had less error than characterizations carried out by fluorescent measurements. At a curcumin concentration of 42.9 μg mL⁻¹, the ARs of H460 lung cancer cells were 13.1%, 52.9%, and 61.7% in pH = 6.5, 7.4, and 8.5 microenvironments, respectively. H460 lung cancer cells could be effectively resisted under a curcumin concentration of 42.9 μg mL⁻¹ in the pH = 8.5 microenvironmental condition. In the range of curcumin concentration of 57.2–100 μg mL⁻¹, the average ARs of the H460 lung cancer cells were 53.6%, 58.4%, and 98.0% for the pH = 6.5, 7.4, and 8.5 microenvironments, respectively. The resistant efficiency of H460 lung cancer cells in the pH = 8.5 microenvironment could be enhanced 1.83-fold compared to that in the pH = 6.5 microenvironment. This opens an opportunity to investigate tumor cells' resistance using a CGG device. Furthermore, this research concluded that two factors played key roles: the concentration of anticancer drugs and the microenvironmental pH conditions.

Acknowledgements

This research work was supported by National Natural Science Foundation of China (No. 21271182), Major National Science Research Program (973 Program, No. 2013CB933000), Jiangsu Natural Science Youth Foundation (No. BK20160398) and Suzhou Science and Technology Project (No. SYG201633).

Notes and references

1. S. G. Uzel, O. C. Amadi, T. M. Pearl, R. T. Lee, P. T. So and R. D. Kamm, Small, 2016, 12, 612–622.
2. B. G. Chung, L. A. Flanagan, S. W. Rhee, P. H. Schwartz, A. P. Lee, E. S. Monuki and N. L. Jeon, Lab Chip, 2005, 5, 401–406.
3. B. Hong, P. Xue, Y. F. Wu, J. N. Bao, Y. J. Chuah and Y. J. Kang, Biomed. Microdevices, 2016, 18, 21–28.
4. J. Y. Cheng, M. H. Yen, C. T. Kuo and T. H. Young, Biomicrofluidics, 2008, 2, 24105–24116.
5. H. Xu, M. M. Ferreira and S. C. Heilshorn, Lab Chip, 2014, 14, 2047–2056.
6. N. N. Ye, J. H. Qin, W. W. Shi, X. Liu and B. C. Lin, Lab Chip, 2007, 7, 1696–1704.
7. K. M. Scully and M. G. Rosenfeld, Science, 2002, 295, 2231–2235.
8. H. L. Ashe and J. Briscoe, Development, 2006, 133, 385–394.
9. R. A. Firtel and C. Y. Chung, BioEssays, 2000, 22, 603–615.
10. D. Q. Wu, Cell Res., 2005, 15, 52–56.
11. S. A. Eccles, Curr. Opin. Genet. Dev., 2005, 15, 77–86.
12. H. Wang, C. H. Chen, Z. L. Xiang, M. Wang and C. Lee, Lab Chip, 2015, 15, 1445–1450.
13. V. V. Abhyankar, M. A. Lokuta, A. Huttenlocher and D. J. Beebe, Lab Chip, 2006, 6, 389–393.
14. J. P. Diao, L. Young, S. Kim, E. A. Fogarty, S. M. Heilman, P. Zhou, M. L. Shuler, M. M. Wu and M. P. DeLisa, Lab Chip, 2006, 6, 381–388.
15. S.-Y. Cheng, S. Heilman, M. Wasserman, S. Archer, M. L. Shuler and M. Wu, Lab Chip, 2007, 7, 763–769.
16. V. V. Abhyankar, M. W. Toepke, C. L. Cortesio, M. A. Lokuta, A. Huttenlocher and D. J. Beebe, Lab Chip, 2008, 8, 1507–1515.
17. Y. Du, J. Shim, M. Vidula, M. J. Hancock, E. Lo, B. G. Chung, J. T. Borenstein, M. Khabiry, D. M. Cropek and A. Khademhosseini, Lab Chip, 2009, 9, 761–767.
18. D. Kim, M. A. Lokuta, A. Huttenlocher and D. J. Beebe, Lab Chip, 2009, 9, 1797–1800.
19. A. A. Ahmad, Y. Wang, C. E. Sims, S. T. Magness and N. L. Allbritton, RSC Adv., 2015, 5, 74881–74891.
20. N. Jeon, S. K. W. Dertinger, D. T. Chiu, I. S. Choi, A. D. Stroock and G. M. Whitesides, Langmuir, 2000, 16, 8311–8316.
21. S. K. W. Dertinger, D. T. Chiu, N. Jeon and G. M. Whitesides, Anal. Chem., 2001, 73, 1240–1246.
22. C.-W. Li, R. Chen and M. Yang, Lab Chip, 2007, 7, 1371–1373.
23. B. Mosadegh, C. Huang, J. W. Park, H. S. Shin, B. G. Chung, S.-K. Hwang, K.-H. Lee, H. J. Kim, J. Brody and N. L. Jeon, Langmuir, 2007, 23, 10910–10912.
24. J. Atencia, J. Morrow and L. E. Locascio, Lab Chip, 2009, 9, 2707–2714.
25. J. Y. Park, S. J. Yoo, C. M. Hwang and S.-H. Lee, Lab Chip, 2009, 9, 2194–2202.
26. Y.-C. Toh, T. C. Lim, D. Tai, G. Xiao, D. van Noort and H. Yu, Lab Chip, 2009, 9, 2026–2035.
27. J. He, Y. Du, J. L. Villa-Uribe, C. Hwang, D. Li and A. Khademhosseini, Adv. Funct. Mater., 2010, 20, 131–137.
28. T. M. Keenan, C. W. Frevert, A. Wu, V. Wong and A. Folch, Lab Chip, 2010, 10, 116–122.
29. C.-G. G. Yang, Y.-F. F. Wu, Z.-R. R. Xu and J.-H. H. Wang, Lab Chip, 2011, 11, 3305–3312.
30. A. G. G. Toh, Z. P. Wang, C. Yang and N. T. Nguyen, Microfluid. Nanofluid., 2014, 16, 1–18.
31. D. Friedrich, C. P. Please and T. Melvin, Chem. Eng. J., 2012, 193, 296–303.
32. D. Liu, L. Wang, R. Zhong, B. Li, N. Ye, X. Liu and B. Lin, J. Biotechnol., 2007, 131, 286–292.
33. J. Kim, M. Hegde, S. H. Kim, T. K. Wood and A. Jayaraman, Lab Chip, 2012, 12, 1157–1163.
34. C. Kim, K. Lee, J. H. Kim, K. S. Shin, K.-J. Lee, T. S. Kim and J. Y. Kang, Lab Chip, 2008, 8, 473–479.
35. K. Lee, C. Kim, B. Ahn, R. Panchapakesan, A. R. Full, L. Nordee, J. Y. Kang and K. W. Oh, Lab Chip, 2009, 9, 709–717.
36. W. S. Yao, R. H. Byrne and R. D. Waterbury, Environ. Sci. Technol., 1998, 32, 2646–2649.
37. A. Beer, Ann. Phys., 1852, 162, 78–88.
38. S. H. Wu, L. W. Hang, J. S. Yang, H. Y. Chen, H. Y. Lin, J. H. Chiang, C. C. Lu, J. L. Yang, T. Y. Lai, Y. C. Ko and J. G. Chung, Anticancer Res., 2010, 30, 2125–2133.
39. Y. Wang, J. Xiao, H. Zhou, S. Yang, X. Wu, C. Jiang, Y. Zhao, D. Liang, X. Li and G. Liang, J. Med. Chem., 2011, 54, 3768–3778.
40. I. T. Chiang, W. S. Wang, H. C. Liu, S. T. Yang, N. Y. Tang and J. G. Chung, Oncol. Rep., 2015, 34, 1853–1874.
41. S. Harguindey, G. Orive, J. L. Pedraz, A. Paradiso and S. J. Reshkin, Biochim. Biophys. Acta, Rev. Cancer, 2005, 1756, 1–24.
42. R. A. Gatenby, E. T. Gawlinski, A. F. Gmitro, B. Kaylor and R. J. Gillies, Cancer Res., 2006, 66, 5216–5223.
43. D. Neri and C. T. Supuran, Nat. Rev. Drug Discovery, 2011, 10, 767–777.
44. B. A. Webb, M. Chimenti, M. P. Jacobson and D. L. Barber, Nat. Rev. Cancer, 2011, 11, 671–677.
45. M. Damaghi, N. K. Tafreshi, M. C. Lloyd, R. Sprung, V. Estrella, J. W. Wojtkowiak, D. L. Morse, J. M. Koomen, M. M. Bui, R. A. Gatenby and R. J. Gillies, Nat. Commun., 2015, 6, 8752–8764.
46. A. I. Hashim, X. M. Zhang, J. W. Wojtkowiak, G. V. Martinez and R. J. Gillies, NMR Biomed., 2011, 24, 582–591.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra20804e

‡ These authors contributed equally.

---