Integrated Microfluidic System for Size-Based Selection and Trapping

Dec 22, 2015 - Vesicles composed of phospholipids (liposomes) have attracted interest as artificial cell models and have been widely studied to explor...
1 downloads 4 Views 3MB Size
Download PDF
Technical Note pubs.acs.org/ac

Integrated Microfluidic System for Size-Based Selection and Trapping of Giant Vesicles Yuki Kazayama,† Tetsuhiko Teshima,‡ Toshihisa Osaki,‡,§ Shoji Takeuchi,*,‡ and Taro Toyota*,† †

Graduate School of Arts and Sciences, The University of Tokyo, 3-8-1 Komaba, Meguro-ku, Tokyo 153-8902, Japan Institute of Industrial Science (IIS), The University of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, Japan § Kanagawa Academy of Science and Technology, 3-2-1 Sakado, Takatsu-ku, Kawasaki City, Kanagawa 213-0012, Japan ‡

S Supporting Information *

ABSTRACT: Vesicles composed of phospholipids (liposomes) have attracted interest as artificial cell models and have been widely studied to explore lipid−lipid and lipid− protein interactions. However, the size dispersity of liposomes prepared by conventional methods was a major problem that inhibited their use in high-throughput analyses based on monodisperse liposomes. In this study, we developed an integrative microfluidic device that enables both the size-based selection and trapping of liposomes. This device consists of hydrodynamic selection and trapping channels in series, which made it possible to successfully produce an array of more than 60 monodisperse liposomes from a polydisperse liposome suspension with a narrow size distribution (the coefficient of variation was less than 12%). We successfully observed a size-dependent response of the liposomes to sequential osmotic stimuli, which had not clarified so far, by using this device. Our device will be a powerful tool to facilitate the statistical analysis of liposome dynamics.

V

including an array of cylindrical microposts and physical trapping structures, have been introduced for the filtration, collection, and pairing of GVs.14−17 However, the previous devices did not control the size dispersity of the GVs, because they were prepared using conventional gentle hydration methods.18 The size dispersity disturbed the trapping efficiency of the devices. Moreover, it is critical to eliminate the influence of the size from the acquired results. An example of the size dependency of GV properties is the membrane raft formation, which has attracted much attention.19 In order to solve these problems, we first propose an integrative platform for producing an array of monodisperse liposomes, which are GVs composed of phospholipids. Note that we apply a polydisperse liposome suspension, prepared using a common gentle hydration method, because of the versatility of this technique. The proposed device combines a hydrodynamic size separation and sorting technique20,21 with physical trapping structures. These enable the size selection of liposomes at the first step and trapping at the second step, and are especially designed for the target size. Immobilized monodisperse liposomes will be useful for exploring sizedependent phenomena such as liposome morphology as well as stochastic analyses that consider the deviation of the lipid

esicles, which are closed bilayer membranes composed of lipids, have been widely investigated both in fundamental studies and for practical applications analyzing lipid−lipid and lipid−protein interactions in industrial and pharmaceutical fields, because their molecular composition and size are compatible with biological membranes. Small unilamellar vesicles (below 100 nm in diameter) have been conventionally used, and some groups recently succeeded to obtain statistical data based on a massive array of small unilamellar vesicles using tethering techniques.1,2 However, they are too small to be regarded as cell models. Giant vesicles (GVs), typically with diameters between 1 and 100 μm, have especially attracted interest as cell models and have been applied in studies that include the membrane raft controversy3−6 and reproduction system.7−9 In these previous studies on GVs, the data acquisition depended on a traditional methodology that manually followed single GVs one by one. Naturally, the data throughput tended to be low, which made it possible to miss minor events. The flow cytometry sorting of GVs using one high-throughput methodology revealed the differences in the dynamics of self-reproducing GVs of different sizes (the size boundary was approximately 10 μm).10,11 However, it cannot monitor and record the individual GV dynamics. Microfluidic devices have been applied for trapping multiple GVs to overcome the above-mentioned throughput problem; these devices were originally developed to control cell pairing and fusion or for single-cell analysis.12,13 Several techniques, © XXXX American Chemical Society

Received: October 7, 2015 Accepted: December 22, 2015

A

DOI: 10.1021/acs.analchem.5b03772 Anal. Chem. XXXX, XXX, XXX−XXX

Technical Note

Analytical Chemistry

by a dashed arrow. An excessive amount of liposomes are collected through the outlet. The separation occurs based on the DLD principle; particles smaller than the critical particle diameter (DC) travel along the initial streamline (zigzag mode), while particles larger than DC are bumped (bump mode). The DC is controlled by the streamline through the microposts, determined by the parameter ελ, where ε and λ are the row shift fraction and post-to-post spacing, respectively. The following method for the theoretical prediction of DC was proposed by Inglis et al.,22 where g is the gap size between the posts.

composition. In the following, we describe the concept of our device and evaluate its size selectivity and trapping efficiency. Finally, we discuss time-course observations of a size-dependent liposome response by sequential osmotic stimuli.

■

DEVICE DESIGN A schematic illustration of our microfluidic device design is shown in Figure 1a. In the front part of the microfluidic

⎡ 1 ⎤ DC = g ⎢1 + 2w + ⎥, ⎣ 2w ⎦ ⎡1 ϵ w=⎢ − + ⎣8 4

⎤1/3⎛ 1 ϵ 3⎞ (ϵ − 1) ⎥ ⎜ − − i ⎟ ⎦ ⎝ 2 16 2 ⎠

(1)

Therefore, it is possible to transfer the target size particles to the target site by combining two types of cylindrical micropost arrays. In the front part of the selection region, the particles larger than DC1 (low ε) are bumped. Subsequently, in the rear part of the selection region, the particles smaller than DC2 (high ε) change their traveling mode to the zigzag mode. As a result, the particles with diameters between DC1 and DC2 are selected and loaded to the trapping region. The parameters and target sizes of the liposomes are listed in Table 1. Table 1. Designed Parameters and Measured Value of Channel Height target diameter (μm)

low ε

high ε

12 16 20

1/17 1/10 1/7

1/16 1/9 1/6

g (μm)

theoretical prediction of selected diameter (μm)

channel height (μm)

40 40 40

11.8−12.2 15.7−16.6 19.0−20.7

14.4 18.2 23.0

The following features are important for the design of the trapping region. First, the gap size of the trap (G) should be small enough to prevent trapped liposomes from passing through the gaps and large enough to increase the trapping flow toward the gaps. In this study, we designed G to be 20−25% of the target diameter x. Second, the pitch of the traps (L) should be designed to ensure high trapping efficiency and prevent the aggregation of liposomes between the traps. We experimentally determined that the value of L should be 2 μm larger than x. We placed dummy resistors in the trapping region, in addition to the physical trapping structures. These dummy traps increased the resistance and prevented the selected liposomes from bypassing the trapping structures and decreasing the trapping efficiency.

Figure 1. Overview of developed device (a). Liposomes with a target diameter are selected from polydisperse liposomes. A typical photograph of the PDMS-on-glass microfluidic device (b) and scanning electron micrographs of PDMS micropost arrays for selection (c), trapping (d), and dummy resistors (e). Scale bars represent 2 mm (a), 1 cm (b) and 50 μm (c−e).

■

EXPERIMENTAL SECTION The microfluidic device shown in Figure 1b was designed and fabricated using standard photolithographic techniques.23,24 The detailed procedure was described in Supporting Information (Text S1). The shape, size, and gap of the micropost arrays were confirmed using scanning electron microscopy (KEYENCE, VE-7800) as shown in Figure 1c−e. The aqueous buffer solution and liposome suspension were infused using two syringe pumps (Harvard Apparatus, Pump 11 Pico Plus Elite) (see the Supporting Information, Figure S1). For loading and trapping liposomes, the flow rates (unit μL/h)

channel, polydisperse liposomes are separated by size using a hydrodynamic separation technique known as deterministic lateral displacement (DLD).22 In the rear part, the selected liposomes are trapped by physical trapping microposts with narrow gaps that prevent the passage of liposomes. An external buffer solution is loaded from inlet 1 as a sheath flow for liposome introduction, and the liposome suspension is loaded from inlet 2 and concentrated in the middle of the channel. The target-size liposomes are separated from the polydisperse liposomes and trapped in the trapping region. The trail of the target size liposomes is depicted in Figure 1a (top) B

DOI: 10.1021/acs.analchem.5b03772 Anal. Chem. XXXX, XXX, XXX−XXX

Technical Note

Analytical Chemistry

20 μm). The mean diameters and coefficients of variation (CVs) of the trapped liposomes were 13.8 μm, 5% (n = 154), 17.7 μm, 10% (n = 105), and 22.1 μm, 12% (n = 67), respectively. The results showed the efficiency of the selection criteria using the designed sequential DLDs. The size dispersions achieved by the device were significantly narrower than the results (15−30%) obtained using the previously reported filtration method.26,27 A few microfluidic technologies using lipid monolayers formed at water−oil interfaces have succeeded in generating monodisperse liposomes,28−30 comparable with the results in our device. However, their methodologies include intrinsic limitations because the oil residue remains within the membrane layers, which potentially influences the properties of the liposome membranes of interest. Therefore, we believe that the device will be useful for statistical and precise analyses of the size-dependent dynamics of liposomes. The mean diameters of the trapped liposomes were slightly larger than the target diameters. We checked for errors in the data treatment used for the analysis; the size distributions of the trapped liposomes were reanalyzed using ImageJ software. However, the mean diameters of the trapped liposomes were similarly found to be larger than the device design (Supporting Information, Figure S3), indicating that this deviation was a result of the characteristics of our system. We focused on the deformability of the liposomes, which was negligible with the plastic/glass particles assumed by theory. The roles of the particle shape and deformability in a DLD device have been studied, especially for red blood cells (RBCs), using both experiments and numerical simulations, to allow their use in the rapid and precise separation of blood components.31,32 The results of these studies indicated that the effective diameter of the RBCs, as determined by their passing and bumping by the DLD microposts, was smaller than the expected diameter. The authors theorized that the effective diameter was affected by the capillary number, which was related to the applied pressure gradient and membrane rigidity of the RBCs. We therefore evaluated the mechanical changes in the liposomes passing by the microposts to quantify the effect of the liposome deformability on the DLD principle. Typical images of a deforming liposome on the surface of a micropost are shown in Figure 2g. The evaluated deformation ratios are listed in Table 2. The deformation ratio increased with the flow

were set to (Q1, Q2) = (1300, 300), and then gradually decreased to (Q1, Q2) = (330−350, 30) to keep the trapped liposomes stable, where Q1 and Q2 were the flow rates of the external solution and liposome suspension, respectively. Morphological changes in the trapped liposomes under osmotic stimuli were observed at flow rates of (Q1, Q2) = (105, −30), where Q2 was in the suction mode. A six-port manual injection valve (Rheodyne, model 7000) was also employed between the device and syringe pump 1 for the quick and facile exchange of the external solution. The exchange rates for the external solution were measured from the fluorescence intensities using a uranine solution, indicating that the exchange of the external solution took 12 min for (Q1, Q2) = (330, −30) and 36 min for (Q1, Q2) = (105, −30) (Supporting Information, Figure S2). Liposomes stained with a fluorescent dye were formed by the gentle hydration of dry lipid films doped with fructose.25 This method enables the efficient formation of liposomes with low lamellarity. The precise descriptions of the preparation method including used lipids, the image acquisition, and data treatment were described in Supporting Information (Text S2 and S3).

■

RESULTS AND DISCUSSION Size Selectivity and Deviation of Trapped Liposome Sizes from Theoretical Values. First, we demonstrated the size selectivity performance of the device. Fluorescence microscopy images and the size distributions of the trapped liposomes are respectively shown in Figure 2a,c,e and Figure 2b,d,f for the three different target diameters (12, 16, and

Table 2. Deformation Ratios for Three Different Flow Rates (Q1, Q2) (μL/h)

deformation ratio (%) (n = 10)

(600, 90) (900, 180) (1300, 300)

4.2 ± 1.9 6.0 ± 2.1 10.1 ± 3.2

rate, up to 10% for (Q1, Q2) = (1300, 300). It is known that the bending modulus of liposomes is typically 1 order of magnitude smaller than that of RBCs, although it depends on the membrane lipid composition.33,34 Therefore, we deduced that the deformability of the liposomes decreased the apparent diameters during the selection process at the DLD and increased the resulting diameters of the trapped liposomes compared to the theoretical sizes. Trapping Efficiency. Second, we examined the trapping efficiency of the device. From a time point when the flow rates were set to (Q1, Q2) = (330−350, 30), the trapping region was monitored and classified using the following three states which

Figure 2. Fluorescence microscopy images and size distributions of trapped liposomes. Target diameters were 12 μm (a, b), 16 μm (c, d), and 20 μm (e, f). Typical images of the deforming liposome traveling in bump mode (g). The images were taken over a time interval of 2.5 ms. Scale bars represent 100 μm (a, c, e) and 20 μm (g). C

DOI: 10.1021/acs.analchem.5b03772 Anal. Chem. XXXX, XXX, XXX−XXX

Technical Note

Analytical Chemistry

ment. In our device, it is preferable to decrease the transitions from successful traps to error traps (k2) in order to achieve better trapping efficiency. We theorized that the flow leak at the gap (G) after trapping the first liposome further attracted the second one. Therefore, three-dimensional trapping structures such as a circular hole instead of the gap should be considered to stop the flow by the first trapping, although their fabrication would demand additional complicated processes.12 Time-Course Observations under Sequential Osmotic Stimuli. We conducted sequential observations of the morphological changes in the trapped liposomes under osmotic stresses. Two devices with target diameters of 12 and 20 μm were used to examine the size dependence on the stimuli. A monodisperse liposome array was produced using an isotonic fructose solution (1 mM) and subsequently exposed to a hypertonic fructose solution (3 mM). The size of the trapped liposomes (n = 10) gradually decreased 10 min after switching the valve to the hypertonic solution, regardless of the target diameter (Figures 4a,c). Confocal fluorescence microscopy revealed that membrane invagination (budding toward the interior) occurred on those contracting liposomes (Figure 4a, inset). Then, a hypotonic stress was applied to the liposomes using the original fructose solution (1 mM). The smaller liposomes (12 μm) showed subtle swelling after 30 min (Figure 4b), whereas most of the larger liposomes (20 μm) exploded within 80 min (Figure 4d, n = 9/10). The change over time of the nominal lamellarity of the trapped liposomes is shown in Figures 4e−h. The contraction ratios (the ratios of sizes of the contracted liposomes to those before contraction) at the first osmotic stress were calculated using three high nominal lamellarity samples and three low lamellarity samples (Figure 4i). Regardless of the liposome size, the contraction ratios seemed to be constant and independent of the initial lamellarity. In contrast, the expansion ratios (the ratios of the sizes of the reswelled liposomes to those before reswelling) at the second osmotic stress were different and depended on the liposome size (Figure 4j). Smaller liposomes exhibited large expansion ratios at 60 min. However, larger liposomes never swelled before exploding (the expansion ratios were calculated from the major axis right before the explosion). Boroske et al. observed a linear decrease in liposome size over time during osmotic contraction and calculated the water permeability of a single bilayer of egg-lecithin liposomes.35 In our case, the water permeabilities were found to be 50 ± 3 μm/ s (n = 8) and 57 ± 7 μm/s (n = 10) for smaller and larger liposomes, respectively. The calculated values were consistent with the literature values.35 Under the second hypotonic stress, smaller liposomes slightly increased in size linearly from 35 to 60 min. The calculated permeability was 27 ± 4 μm/s (n = 8), which was smaller than the value during contraction. This rationally means that the increase in lamellarity caused by the membrane invagination inhibited the osmotic water transport across the membrane. In conclusion, our device enabled both long-term observations of changes in the liposome morphology over time and a stochastic analysis using a monodisperse liposome array. In addition, we revealed the size-dependent response of liposomes to osmotic stimuli. Although the liposomes trapped in our device were exposed to flow stress between the microposts, this had relatively little effect on the morphological dynamics of the liposomes as long as the flow rate was small.

were changed over time due to the instability and continuous loading of the liposomes (Figure 3): blank traps (dashed lines,

Figure 3. Time-course of trapping state change. The target diameters were 12 μm (blue line), 16 μm (green line), and 20 μm (red line). The trapping states were classified as blank traps (dashed lines, triangles), successful traps (solid lines, circles), and error traps (solid line, squares). The size distributions of the trapped liposomes were analyzed at the times represented by the open circles.

triangles), successful traps composed of single liposomes (solid lines, circles), and error traps composed of multiple liposomes (solid line, squares). Three devices with the respective target diameters of 12 μm (blue line), 16 μm (green line), and 20 μm (red line) were used. The trapping efficiencies were calculated as the ratio of the number of successful traps to all the traps. Skelley et al. showed that the trapping efficiencies changed depending on the trap geometry.12 Here, we found that the trapping efficiencies were also related to the observation time. In our system, the blank traps monotonically decreased and the error traps monotonically increased regardless of the liposome size. On the other hand, it seemed that the single-liposome traps slightly decreased for 12 μm, slightly increased for 16 μm, and initially increased and then subsequently decreased for 20 μm. The maximum trapping efficiencies were achieved in 70 min, with efficiencies of 64, 66, and 70% for the target diameters of 12, 16, and 20 μm, respectively. Considering the dynamics of the trapping process, we assumed that the trapping states changed as follows, k1

k2

k1′

k2′

→ [S0(t )]→ ← [S1(t )]← [S2(t )]

(2)

where [S0(t)], [S1(t)], and [S2(t)] are the ratios of the numbers of blank traps, successful traps, and error traps, respectively, and k1, k1′, k2, and k2′ are the proportional constants for each transition (Supporting Information, Figure S4). The ratios of the averaged constants were estimated from nine pairs of neighboring images of the trapping region (Table 3). This result indicates that (i) the transition from successful Table 3. Average Ratios of Proportional Constants for Three Different Target Diameters target diameter (μm)

k1

k1′

k2

k2′

12 16 20

160 120 90

0.8 1 0

60 39 44

62 44 52

traps to blank traps (k1′) was negligible, while (ii) transitions between error traps and successful traps (k2 and k2′) occasionally occurred. In other words, the state of a single liposome in a trap was stable, whereas the state of multiple liposomes in a trap was unstable. The model clarifies the characteristics of the device and provides a hint for improveD

DOI: 10.1021/acs.analchem.5b03772 Anal. Chem. XXXX, XXX, XXX−XXX

Technical Note

Analytical Chemistry

artificial cell models but also a high-throughput screening platform for drug discovery.

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b03772. Texts S1−S3 and Figures S1−S5 (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*Phone: +81-3-5452-6650. Fax: +81-3-5452-6649. E-mail: [email protected]. *Phone: +81-3-5465-7634. Fax: +81-3-5465-7634. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Mr. H. Yoshida and Ms. M. Onuki (The Univ. of Tokyo) for their technical assistance in the microfabrication. This work was partly supported by Platform for Dynamic Approaches to Living System and Grant-in-Aid for JSPS Fellows (Y.K.; Grant 269147) from The Ministry of Education, Culture, Sports, Science and Technology, Japan.

■

REFERENCES

(1) Stamou, D.; Duschl, C.; Delamarche, E.; Vogel, H. Angew. Chem., Int. Ed. 2003, 42, 5580−5583. (2) Bally, M.; Bailey, K.; Sugihara, K.; Grieshaber, D.; Vörös, J.; Städler, B. Small 2010, 6, 2481−2497. (3) Baumgart, T.; Hess, S. T.; Webb, W. W. Nature 2003, 425, 821− 824. (4) Tamba, Y.; Yamazaki, M. Biochemistry 2005, 44, 15823−15833. (5) Tamba, Y.; Yamazaki, M. J. Phys. Chem. B 2009, 113, 4846−4852. (6) Morita, M.; Hamada, T.; Vestergaard, M. C.; Takagi, M. Phys. Chem. Chem. Phys. 2014, 16, 8773−8777. (7) Walde, P.; Wick, R.; Fresta, M.; Mangone, A.; Luisi, P. L. J. Am. Chem. Soc. 1994, 116, 11649−11654. (8) Zhu, T. F.; Szostak, J. W. J. Am. Chem. Soc. 2009, 131, 5705− 5713. (9) Kurihara, K.; Tamura, M.; Shohda, K.; Toyota, T.; Suzuki, K.; Sugawara, T. Nat. Chem. 2011, 3, 775−781. (10) Sunami, T.; Sato, K.; Matsuura, T.; Tsukada, K.; Urabe, I.; Yomo, T. Anal. Biochem. 2006, 357, 128−136. (11) Kurihara, K.; Takakura, K.; Suzuki, K.; Toyota, T.; Sugawara, T. Soft Matter 2010, 6, 1888−1891. (12) Skelley, M.; Kirak, O.; Suh, H.; Jaenisch, R.; Voldman, J. Nat. Methods 2009, 6, 147−152. (13) Eyer, K.; Stratz, S.; Kuhn, P.; Küster, S. K.; Dittrich, P. S. Anal. Chem. 2013, 85, 3280−3287. (14) Nuss, H.; Chevallard, C.; Guenoun, P.; Malloggi, F. Lab Chip 2012, 12, 5257−5261. (15) Paterson, D. J.; Reboud, J.; Wilson, R.; Tassieri, M.; Cooper, J. M. Lab Chip 2014, 14, 1806−1810. (16) Robinson, T.; Verboket, P. E.; Eyer, K.; Dittrich, P. S. Lab Chip 2014, 14, 2852−2859. (17) Yamada, A.; Lee, S.; Bassereau, P.; Baroud, C. N. Soft Matter 2014, 10, 5878−5885. (18) Reeves, J. P.; Dowben, R. M. J. Cell. Physiol. 1969, 73, 49−60. (19) Buboltz, J. T.; Bwalya, C.; Williams, K.; Schutzer, M. Langmuir 2007, 23, 11968−11971. (20) Huang, L. R.; Cox, E. C.; Austin, R. H.; Sturm, J. C. Science 2004, 304, 987−990.

Figure 4. Changes in size (along major axis) (a−d) and nominal lamellarity (e−h) over time of 10 trapped liposomes under osmotic stress. (Inset) Typical confocal fluorescence images of trapped liposomes under sequential osmotic stresses (scale bars = 10 μm). The liposomes experienced (a, c, e, g) a first hypertonic stress and (b, d, f, h) second hypotonic stress (a return to the initial environment). The target diameters were set at 12 μm (a, b, e, f) and 20 μm (c, d, g, h), respectively. Contraction ratios in shrinking process (i) and expansion ratios in the reswelling process (j).

■

CONCLUSIONS We developed a microfluidic platform for the size-based separation and trapping of liposomes. The developed device produced a monodisperse liposome array using a single loading procedure and succeeded in tracing the morphological changes in liposomes under sequential osmotic stresses. Moreover, the device could be reused by flushing the trapped liposomes (Supporting Information, Figure S5). This feature enables multiple assays with a single setup and efficient analyses using rare samples such as liposomes functionalized with membrane proteins or chemical ingredients. We believe that the device will become not only a useful tool for the statistical analysis of E

DOI: 10.1021/acs.analchem.5b03772 Anal. Chem. XXXX, XXX, XXX−XXX

Technical Note

Analytical Chemistry (21) Sajeesh, P.; Sen, A. K. Microfluid. Nanofluid. 2014, 17, 1−52. (22) Inglis, D. W.; Davis, J. A.; Austin, R. H.; Sturm, J. C. Lab Chip 2006, 6, 655−658. (23) McDonald, J. C.; Duffy, D. C.; Anderson, J. R.; Chiu, D. T.; Wu, H.; Schueller, O. J. A.; Whitesides, G. M. Electrophoresis 2000, 21, 27− 40. (24) Osaki, T.; Yoshizawa, S.; Kawano, R.; Sasaki, H.; Takeuchi, S. Anal. Chem. 2011, 83, 3186−3191. (25) Tsumoto, K.; Matsuo, H.; Tomita, M.; Yoshimura, T. Colloids Surf., B 2009, 68, 98−105. (26) Zhu, T. F.; Szostak, J. W. PLoS One 2009, 4, e5009. (27) Tamba, Y.; Terashima, H.; Yamazaki, M. Chem. Phys. Lipids 2011, 164, 351−358. (28) Ota, S.; Yoshizawa, S.; Takeuchi, S. Angew. Chem., Int. Ed. 2009, 48, 6533−6537. (29) van Swaay, D.; de Mello, A. Lab Chip 2013, 13, 752−767. (30) Matosevic, S.; Paegel, B. M. Nat. Chem. 2013, 5, 958−963. (31) Beech, J. P.; Holm, S. H.; Adolfsson, K.; Tegenfeldt, J. O. Lab Chip 2012, 12, 1048−1051. (32) Krüger, T.; Holmes, D.; Coveney, P. V. Biomicrofluidics 2014, 8, 054114. (33) Rawicz, W.; Olbrich, K. C.; McIntosh, T.; Needham, D.; Evans, E. Biophys. J. 2000, 79, 328−339. (34) Tomaiuolo, G. Biomicrofluidics 2014, 8, 051501. (35) Boroske, E.; Elwenspoek, M. Biophys. J. 1981, 34, 95−109.

F

DOI: 10.1021/acs.analchem.5b03772 Anal. Chem. XXXX, XXX, XXX−XXX