Sprouting Angiogenesis under a Chemical Gradient Regulated by

Oct 10, 2011 - Microfluidic cell culture assays are versatile tools for studying cell migration, particularly angiogenesis. Such assays can deliver pr...
4 downloads 8 Views 4MB Size
ARTICLE pubs.acs.org/ac

Sprouting Angiogenesis under a Chemical Gradient Regulated by Interactions with an Endothelial Monolayer in a Microfluidic Platform Gi Seok Jeong,† Sewoon Han,‡ Yoojin Shin,‡ Gu Han Kwon,† Roger D. Kamm,§ Sang-Hoon Lee,*,† and Seok Chung*,‡ †

Department of Biomedical Engineering, College of Health Science, Korea University, Seoul, Korea School of Mechanical Engineering, Korea University, Seoul, Korea § Department of Mechanical Engineering and Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America ‡

bS Supporting Information ABSTRACT: Microfluidic cell culture assays are versatile tools for studying cell migration, particularly angiogenesis. Such assays can deliver precisely controlled linear gradients of chemical stimuli to cultured cells in a microfluidic channel, offering excellent optical resolution and in situ monitoring of cellular morphogenesis in response to a gradient. Microfluidic cell culture assays provide a chemical gradient subject to molecular diffusion, although cellular metabolism can perturb it. The actual gradient perturbed by cells has not been precisely described in the context of regulated cellular morphogenesis. We modeled the chemical gradient in a microfluidic channel by simulating the analyte(VEGF) distribution during cellular interactions. The results were experimentally verified by monitoring sprouting angiogenic response from a monolayer of human umbilical vein endothelial cells (hUVECs) into a type 1 collagen scaffold. The simulation provided a basis for understanding a real distribution of the analyte interrupted by cells in microfluidic device. The new protocol enables one to quantify the morphogenesis of hUVECs under a flat, less-steep, or steep gradient.

C

ell migration is essential for a variety of physiological and pathological processes, including new blood vessel formation, cancer metastasis, wound healing, and inflammation.1 Cell migration in the context of sprouting angiogenesis—the initiation of microvessel growth from existing vessels—has been an important feature in vascular biology.2,3 A number of angiogenic proteins, including vascular endothelial growth factor (VEGF) and related factors, have been identified for their potential therapeutic uses in cancer and cardiovascular diseases3,4 Mechanical stimuli, including shear stress, 5,6 interstitial flow, 7,8 and matrix stiffness,5,9 are also key regulators of sprouting angiogenesis, and their precise systematic regulation and characterization poses several challenges.10 Since the development of the first in vitro assay, several attempts have been made to establish assays that better mimic the true in vivo microenvironments; however, the need for quantitative assays and three-dimensional (3D) models remains.2,11,12 Precise regulation of in vitro chemical and mechanical stimuli that quantitatively influence angiogenesis can facilitate successful drug discovery, tissue engineering, and therapeutic strategies.1315 Recent advances in microfluidic cell migration assays have suggested new quantitative methods16 that are versatile and effective for studying endothelial cell biology, including angiogenesis.17,18 Microfluidic cell migration assays present a framework for controlling the complex chemical stimuli imposed upon cells by providing a stable linear gradient.5,19 The assays also r 2011 American Chemical Society

provide excellent real-time optical resolution, thereby enabling precise quantification.20 Recently, 3D microfluidic assays have emerged that use hydrogel scaffolds to mimic the extracellular matrix (ECM) into which cells migrate and proliferate.5,19,2131 Hydrogel scaffolds have been used to advance studies of 3D cellular morphogenesis, including angiogenesis,5,22,29,31 cancer metastasis,21,23,28 neutrophil migration,26 and interactions among multiple cell types.5,30,19 Some studies have even described gelfree schemes that employ alternative methods of constructing 3D tissue organizational structures.32,33 These approaches have spanned the gap between in vitro two-dimensional (2D) cell culture assays and in vivo whole-animal models.3437 The microfluidic approaches described above have mainly involved stable linear chemical gradients introduced over cells cultured in channels or on a hydrogel scaffold. Linear gradients are subject to some degree of diffusion, which may be visualized using fluorescent particles as stand-ins for bioactive molecules without modeling the spatial interruptions or the consumption of the gradient by cells. Microfluidic linear gradients have, therefore, suffered from limited reliability with respect to quantitation of migration and morphogenesis. The 3D microfluidic assays described above have also lacked the statistical reliability required Received: June 14, 2011 Accepted: October 10, 2011 Published: October 10, 2011 8454

dx.doi.org/10.1021/ac202170e | Anal. Chem. 2011, 83, 8454–8459

Analytical Chemistry

ARTICLE

Figure 1. Three cell seeding protocols for the microfluidic cell migration assay are illustrated. (a) The 2-Ch protocol: Cell seeding in the right-hand channel of a two-channel device. Cells were attached to a collagen scaffold under the hydrostatic pressure and were exposed to a VEGF gradient formed by applying a higher-concentration VEGF solution to the left-hand channel. (b) The 3-Ch/1-Cell protocol: Cell seeding in the center channel of a three-channel device. Cells were attached to each side of a collagen scaffold under the hydrostatic pressure and were exposed to a VEGF gradient formed by applying a higherconcentration VEGF solution to the left-hand channel with control medium applied at the right-hand channel. (c) A new 3-Ch/2-Cell protocol: Cell seeding in the center and right channels of a three-channel device. Cells were attached to both collagen scaffolds under gravity while the platform was vertically positioned in the incubator. But some of the cells were still attached to the bottom side. As a result, cells proliferated on both the side and bottom of the channels. A VEGF gradient was formed by application of higher-concentration VEGF solution at the left channel and control medium at the other channels.

to describe cellular response and/or complex 3D cellular morphogenesis in a scaffold in response to an applied chemical gradient. The three-channel (3-Ch) concept was suggested as a means to enhance the quantitative reliability by integrating control and conditioning experiments in a single device with limited knowledge of the actual molecular gradient set up in the incorporated hydrogel scaffolds.5 Cell migration observed on the control side of a 3-Ch system suggested that the integration of the control failed, indirectly proving existence of a gradient, even the control side filled with a uniform supplemented control medium.5 In this study, we developed a precise model of the chemical gradients that are set up in the context of hydrogels used in microfluidic assays, and we experimentally tested the model using cell seeding protocols. The simulations considered diffusion, inhibition, and consumption to precisely estimate the slope of the complex chemical gradient. We additionally investigated the role of the gradient slope on the endothelial monolayer cultured in a microfluidic channel concurrent with sprouting angiogenesis into the hydrogel scaffold.

’ MATERIALS AND METHODS Microfluidic Assay Incorporating a Hydrogel Scaffold. The microfluidic assay incorporating a hydrogel scaffold was developed based on previous studies.5,1921,29,30 Briefly, chips were made of

PDMS (poly dimethyl siloxane, Sylgard 184, Dow Chemical, MI, USA) patterned on an SU-8-patterned silicon wafer (MicroChem, MA, USA) using a conventional soft lithography process. Inlets and outlets for filling with gel or fluid were defined using a punch. After autoclaving, the PDMS chip and coverslide were bonded together by plasma treatment (Femto Science, Korea) and coated with a 1 mg/mL poly- D-lysine (PDL) solution, followed by incubation at 37 °C for 6 h. After PDL coating, the channel was aspirated, washed, and dried at 80 °C for 24 h to render the channel surface hydrophobic. The scaffold region was filled with Type 1 collagen (BD BioSciences, MA, USA) initially at pH 7.4, followed by incubation at 37 °C for an additional 30 min to allow gelation (Supporting Information Figure S1). Endothelial Cell Seeding and Migration into the Collagen Scaffold. Human umbilical vein endothelial cells (hUVECs, LonZa, Basel, Switzerland) were cultured in endothelial cell medium (ECM; Sciencell Research Laboratories Inc., CA, USA), consisting of 500 mL basal medium, 5% fetal bovine serum (FBS), 20 ng/mL VEGF, and 1% penicillin/streptomycin solution. A concentration of 20 ng/mL VEGF, even in the control medium, was the minimum concentration required to achieve the initial proliferation and monolayer formation of hUVECs in the microfluidic channel. The hUVECs were expanded for no more than eight passages. Cell-seeding schemes were created using tiny pressure differences created by droplets of different 8455

dx.doi.org/10.1021/ac202170e |Anal. Chem. 2011, 83, 8454–8459

Analytical Chemistry

ARTICLE

Figure 2. Simulation of the VEGF gradient, with consideration for diffusion, inhibition, and consumption by the hUVEC monolayer. Experimental results under a gradient that provides an active VEGF supply (a high concentration was applied to the left channel). (a) Depiction of the 3-Ch/1-Cell protocol, as described.5,20,29,30 The cells migrated into the collagen scaffolds, attracted by the VEGF gradient. (b) Simulated VEGF gradient applied to the hUVECs cultured in the center channel, and (c) the sprouting angiogenic response of the hUVECs under 5020C20 ng/mL initial conditions. (d) Depiction of the 3-Ch/2-Cell protocol. (e) Simulated VEGF gradient and (f) sprouting angiogenic response of the hUVEC under 5020C20C ng/mL initial conditions. The small figures (top right) in (b) and (e) show the gradient over 24 h. The small figures (below left) of (b) and (e) indicate the initial concentration of VEGF (ng/mL) in the medium. The scale bar indicates 150 μm.

size, shown in Figure 1a,b, as described previously.5,29 In the cellculture protocols illustrated in Figure 1c, the center and right channels were filled with 4050 μL of a hUVEC suspension (2  106 cells/ml), and the device was kept in a 37 °C incubator in a vertical position for 30 min to facilitate attachment of cells by gravity onto the collagen scaffold. hUVECs attached to the sidewalls of the scaffold and channel surface were evenly distributed by the flow induced by the small pressure differential during cell seeding (Supporting Information Figure S1e). After cell seeding, all devices were kept at 37 °C in an incubator containing 5% CO2, and the medium was replaced daily. Cell migration was monitored daily by phase-contrast microscopy (Zeiss, Oberkochen, Germany), and images were collected using the MetaMorph software (Molecular Devices, Inc., CA, USA). The projected area of sprouting angiogenesis was manually measured using ImageJ (http://rsbweb.nih.gov/ij/). Statistical

analysis was implemented using SPSS ver. Twelve (Chicago, IL, USA). The cell seeding protocols are illustrated in Figure 1, and confluence of the hUVEC monolayer was confirmed, as shown in Supporting Information Figure S1f. Computational Fluid Dynamics Analysis. To understand the shape of the complex chemical gradient generated by diffusion into the hydrogel scaffold, which was also influenced by inhibition and consumption by the endothelial monolayer, we performed a two-dimensional simulation using an FEM (finite element method)-based CFD (computational fluid dynamics) code in COMSOL Multiphysics 4.0 (COMSOL Inc., MA, USA). A structured grid system was used to describe the major part of the channel (to yield a high computational performance), and the number of grids was approximately 12 944 (Supporting Information Figure S2). The diffusion coefficient of VEGF in the medium was assumed to be the same as that in water (2.5  8456

dx.doi.org/10.1021/ac202170e |Anal. Chem. 2011, 83, 8454–8459

Analytical Chemistry

ARTICLE

10 9 m 2 /s), and the diffusion coefficients of VEGF in the collagen gel and the cell monolayer were assumed to be 6.6  1011 m2/s19,38 and 1.2  1011 m2/s,29,39 respectively. VEGF consumption due to cellular metabolism was assumed to be 20 ng/day/106 cells.40 The diffusion of VEGF was simulated using Fick’s second law (∂C)/(∂t) + r 3 (Drc) = 0, where c denotes the VEGF concentration (mol/m3), and D denotes the diffusion coefficient (m2/s). All boundaries were set to a no-mass flow-through. The concentration of VEGF was subject to the initial conditions of passive VEGF supply (all channels were filled with the control medium; 202020 ng/mL) or active VEGF supply (a high concentration was applied to the left channel; 502020 ng/mL).

’ RESULTS AND DISCUSSIONS Figure 1 illustrates the various cell culture protocols.5,20,21,29,30 In the 2-Ch protocol, hUVECs were cultured in the right-hand channel in the presence of a VEGF gradient applied to the lefthand channel (Figure 1a). In the 3-Ch/1-Cell protocol, hUVECs were cultured in the center channel (Figures 1b and 2a). Under active VEGF supply, media with a high VEGF concentration were supplied to the left-hand channel, whereas control medium was supplied to the right-hand channel (Figures 1b and 2b: the left-hand channel was filled with medium containing 50 ng/mL VEGF, and the other channels were filled with a control medium containing 20 ng/mL VEGF (5020C20)). In the presence of active VEGF supply, hUVECs attached to the collagen scaffold near the left-hand channel responded to the diffusion of VEGF. The remaining hUVECs attached to the opposite side, near the right-hand channel, were slightly less responsive (Figure 2c). A steeper gradient formed on the left-hand side, although a small notable gradient also formed on the right-hand side (Figure 2b). The simulations of diffusion, inhibition, and consumption of VEGF together successfully estimated the small gradient on the control side, which explained the reduced but apparent sprouting angiogenesis in the scaffolds on the right-hand side. In a new 3-Ch/2-Cell protocol, hUVECs were cultured on the collagen scaffolds both on left-hand side of the center channel and also on the left-hand side of the control channel, as illustrated in Figures 1c and 2d. To attach hUVECs to each side, we left the device left-side-down in an incubator for 30 min after cell seeding. Cells successfully attached to the desired sides of the scaffolds under the force of gravity to form a confluent monolayer within 12 h. During the simulation, a steep gradient formed mainly on the cells in the center channel, which left a small nearly flat gradient on the control side (Figure 2e). The sprouting angiogenic response to the steep gradient was only observed in the center channel, and a small hUVEC response was observed in the control channel (Figure 2f). The simulations of passive VEGF supply (all channels filled only with the control medium) predicted the presence of a gradient derived by VEGF deficiency in the cell-culture channel during the growth and proliferation of hUVECs (Figure 3). The sprouting angiogenic response was quantified, as shown in Figure 4. The migrated cell area, which is an angiogenesis metric,5,19 increased in the collagen scaffold in correlation with the simulated active VEGF supply gradient according to the 3-Ch/1-Cell protocol (Figure 4a). It was impossible to put a hUVEC monolayer in the absence of a gradient. Control experiments for the cellular response under a chemical gradient could not be achieved, as indicated by the non-negligible amount of migration, even under passive VEGF supply. The role of the

Figure 3. Simulation of the VEGF gradient with consideration for diffusion, inhibition, and consumption by the hUVEC monolayer under a passive VEGF supply. (a) Simulated VEGF gradient in the 3-Ch/ 1-Cell protocol, with a hUVEC monolayer cultured in the center channel under 2020C20 ng/mL initial conditions. (b) Simulated VEGF gradient in the 3-Ch/2-Cell protocol, with the hUVEC monolayer cultured in the center and control channels under 2020C20C ng/ mL initial conditions. The small figures (top right) show the gradient over 24 h, while small figures (below left) indicate the initial concentration of VEGF (ng/mL) in the medium.

VEGF gradient in driving sprouting angiogenesis was quantified, although a relatively high p-value was predicted in the t tests (routinely >0.05). However, a new 3-Ch/2-Cell protocol generated an almost flat gradient on the control side, which enhanced the overall reliability of the metrics used to describe sprouting angiogenesis induced by a VEGF gradient (Figure 4b). The enhanced reliability was demonstrated by one-way ANOVA statistical analysis of multiple comparisons, followed by Tukey’s posthoc multiple comparison test. The sprouting angiogenic responses to two VEGF gradients with different slopes— a steep gradient Δ50 formed by 5020C20C initial conditions and a less steep gradient Δ20 formed by 2020C20C initial condition—were successfully quantified and differentiated with good statistical reliability (Figure 4(b),(c)). In the absence of a VEGF gradient, a hUVEC monolayer sprouting response was not observed. We found that a VEGF gradient was essential for sprouting angiogenesis, and interestingly, the slope of the VEGF gradient precisely regulated the extent of the sprouting response. To understand the contributions of the physical structure of 3-Ch system, we simulated and cultured cells in a 2-Ch assay that included a hydrogel scaffold. The simulations showed the presence of a gradient driven by consumption, even under a passive VEGF supply (Supporting Information Figure S3), and hUVEC sprouting angiogenesis was observed in response to the gradient (Figure 4d). Statistical analysis of the reliability of all protocols described (2-Ch, 3-Ch/1-Cell, and 3-Ch/2-Cell protocols) 8457

dx.doi.org/10.1021/ac202170e |Anal. Chem. 2011, 83, 8454–8459

Analytical Chemistry

ARTICLE

Figure 5. Morphological variations during sprouting angiogenesis. (a) The graph shows the relationship between the normalized area and the perimeter, which showed a higher number of sprouts from the hUVEC monolayer on day n in the 3-Ch/2-Cell protocol. The data show that a narrower angiogenic structure formed in the Δ20 gradient under the 2020C20C initial conditions (n = 14) than was observed in the Δ50 gradient under the 5020C20C initial conditions (n = 16). Each plot represents the values on each day, and the error bars indicate the standard error. (b) Examples of sprouts. The white arrowheads indicate tip cells, and empty arrowheads indicate stalk cells migrated from the hUVEC monolayer.

Figure 4. Quantitative analysis of the sprouting angiogenesis. (a) Mean migratory areas per day in the 3-Ch/1-Cell protocol into the left and right scaffolds under 5020C20 ng/mL initial conditions (n = 8). (b) The mean migratory areas per day in the 3-Ch/2-Cell protocol into the scaffolds from the center and control channels under active VEGF supply with 5020C20C ng/mL initial conditions (n = 16). (c) Mean migratory areas per day in the 3-Ch/2-Cell protocol into the scaffolds under passive VEGF supply with 2020C20C ng/mL initial conditions (n = 14). (d) Mean migratory areas per day in the 2-Ch protocol into the scaffold under passive VEGF supply with 2020C ng/mL initial conditions (n = 10). (e) Comparison of the mean migratory areas per day in all three protocols. Statistical analysis was performed using oneway ANOVA followed by Tukey’s posthoc multiple comparison test. The error bars indicate the mean ( standard error (*P < 0.05; **P < 103; ***P < 103; +p < 104; ++p < 105; +++p < 106).

showed enhanced reliability of the 3-Ch/2-Cell protocol, with dramatically reduced p-values. We confirmed the role of the VEGF gradient during the initial sprouting phase using a perfectly controlled system with a flat gradient. Interestingly, we found that the capillaries that formed under the Δ20 gradient were narrower with smaller areas but larger perimeters than those that formed under the Δ50 gradient. The capillaries featured fewer filopodia, and those present were stretched and elongated (Figure 5b). The filopodia dimensions

observed approached those observed in vivo and achieved in vitro by engineered ECM structures.41 The VEGF gradient Δ50, formed by the 5020C20C (ng/mL) initial conditions over a 700 μm wide scaffold, may have been overly steep, initiating multiple tip cells on the top layer and activating too many filopodia at the same time. In summary, this gradient generated short wide capillary structures. The role of the magnitude of the gradient slope on the growth of capillary structures was previously described in the reference, being related to tip cell activation.42 Under a steep gradient, many activated tip cells may invade into the ECM, resulting in short and wide capillary structure. We confirmed the hypothesis that Δ50 gradient was too steep, which made the capillary structure wider, while gradient Δ20 generated long and stable capillary structures. The simulation was used to calculate the magnitude of the global gradient across the length of the scaffold, as well as the local gradient across the scaffold at a depth of 100 μm from the cell monolayer (definitions are provided in Supporting Information Figure S4) in the 3-Ch/2-Cell protocol. Inhibition (blocking) and consumption of VEGF by cells was modeled (+B/+C), and with consideration only for inhibition by the cells (+B/C) as assumed previously.5,29 Neither inhibition nor consumption (B/C) was also modeled as assumed in previous studies19,2123,2528,31 (Supporting Information Figure S5). The initial concentration of the stimulant was 5020C20C ng/mL. In contrast with the other simulated values, the global and local gradients increased only under the +B/+C case, which showed the importance of considering consumption in the simulation and indirectly explained the overall increase in the sprouting area. Which gradient is expected to dominate the process of sprouting angiogenesis: global or local? We cannot answer this question yet; however, we did identify differences between the magnitude of the global and local gradients in +B/+C. By precisely controlling the gradients experienced by the cell monolayers, we hope to gain additional insight into the spatial interactions between cells and the VEGF gradients. We also anticipate obtaining more reliable estimates of the actual diffusion coefficients,29,39 even with consideration for the 8458

dx.doi.org/10.1021/ac202170e |Anal. Chem. 2011, 83, 8454–8459

Analytical Chemistry absorption of molecules on the scaffold fibers, autocrine/paracrine signaling, and consumption rates.40 We showed that the 2D simulation did not generate significant errors relative to the 3D simulation, which considered channel height and the cross-sectional diffusion profile (Supporting Information Figure S6). When the hUVECs were cultured in all three channels in a medium supplemented only with an initial 20 ng/mL VEGF concentration, the simulation predicted formation of a flat gradient across all of the hUVEC monolayers due to the VEGF consumption by the cells (Supporting Information Figure S7a). The sprouting area did not increase both in the conditioned and control sides (Supporting Information Figure S7b), demonstrating that a flat gradient did not induce sprouting angiogenesis. It also confirmed that consumption by the hUVEC monolayer affected the gradient slope. The flat gradient and suppressed sprouting are important for evaluating the effects of physical stimuli on angiogenesis sprouting, for example, stiffness of the scaffold, interstitial flows or shear stress on the cells, removing effect of chemical stimuli imposed by a consumption gradient, even in the control medium. In conclusion, the slope of the VEGF gradient was found to affect the extent of sprouting and the angiogenesis structures, as confirmed using the new 3-Ch/2-Cell protocol. We modeled the chemical gradient in a microfluidic channel by simulating the VEGF distribution during cellular interactions. The new protocol could quantify the morphogenesis of hUVECs under a flat, lesssteep (Δ20) or steep (Δ50) gradients. It also enabled a controlintegrated analysis free from the consumption gradient imposed by cellgradient interactions. The new protocol may be applied to simulations or experimental demonstrations of sprouting angiogenesis, as well as to studies of the migration of various cell types, including neurons, neutrophils, muscle cells, or metastatic cancer cells. Such studies are strongly related to new applications in tissue engineering, and the cell on a chip and drug discovery fields.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Tel: +82-2-940-2880; Fax: +82-2-921-6818; E-mail: dbiomed@ korea.ac.kr (S.-H.L.). Tel: +82-2-3290-3352; Fax: +82-2-9269290; E-mail: [email protected] (S.C.).

’ ACKNOWLEDGMENT This research was supported by the Converging Research Center Program funded by the Ministry of Education, Science and Technology (2010K001179), and Seoul R&BD Program (PA090930) ’ REFERENCES (1) Coultas, L.; Chawengsaksophak, K.; Rossant, J. Nature 2005, 438, 937–945. (2) Folkman, J.; Haudenschild, C. Nature 1980, 288, 551–556. (3) Folkman, J. Annu. Rev. Med. 2006, 57, 1–18. (4) Ferrara, N.; Kerbel, R. S. Nature 2005, 438, 967–974.

ARTICLE

(5) Chung, S.; Sudo, R.; Mack, P. J.; Wan, C. R.; Vickerman, V.; Kamm, R. D. Lab Chip 2009, 9, 269–275. (6) Polacheck, W. J.; Charest, J. L.; Kamm, R. D. Proc. Natl. Acad. Sci. 2011, 108, 11115–11120. (7) Lovchik, R. D.; Bianco, F.; Tonna, N.; Ruiz, A.; Matteoli, M.; Delamarche, E. Anal. Chem. 2010, 82, 3936–3942. (8) Park, J. Y.; Takayama, S.; Lee, S. H. Integr. Biol. 2010, 2, 229–240. (9) Jeong, G. S.; Kwon, G. H.; Kang, A. R.; Jung, B. Y.; Park, Y.; Chung, S.; Lee, S. H. Biomed. Microdev. 2011, 13, 717–723. (10) Hudlicka, O. J. Physiol. 1991, 444, 1. (11) Yamada, K. M.; Cukierman, E. Cell 2007, 130, 601–610. (12) Lee, K. H.; Shin, S. J.; Park, Y.; Lee, S. H. Small 2009, 5, 1264–1268. (13) Salieb-Beugelaar, G. B.; Simone, G.; Arora, A.; Philippi, A.; Manz, A. Anal. Chem. 2010, 82, 4848–4864. (14) Langer, R. Tissue Eng. 2007, 13, 1–2. (15) Jain, R. K.; Schlenger, K.; Hockel, M.; Yuan, F. Nat. Med. 1997, 3, 1203–1208. (16) Auerbach, R.; Auerbach, W.; Polakowski, I. Pharmacol. Ther. 1991, 51, 1–11. (17) Young, E. W. K.; Simmons, C. A. Lab Chip 2010, 10, 143–160. (18) Van Der Meer, A. D.; Poot, A. A.; Duits, M. H. G.; Feijen, J.; Vermes, I. J Biomed Biotechnol 2009, 823148, 110. (19) Sudo, R.; Chung, S.; Zervantonakis, I. K.; Vickerman, V.; Toshimitsu, Y.; Griffith, L. G.; Kamm, R. D. FASEB J. 2009, 23, 2155–2164. (20) Chung, S.; Sudo, R.; Vickerman, V.; Zervantonakis, I. K.; Kamm, R. D. Annals of biomedical engineering 2010, 38, 1164–1177. (21) Zervantonakis, I.; Chung, S.; Sudo, R.; Zhang, M.; Charest, J.; Kamm, R. Int. J. Micro-Nano Scale Trans. 2010, 1, 27–36. (22) Barkefors, I.; Thorslund, S.; Nikolajeff, F.; Kreuger, J. Lab Chip 2009, 9, 529–535. (23) Liu, T.; Li, C.; Li, H.; Zeng, S.; Qin, J.; Lin, B. Electrophoresis 2009, 30, 4285–4291. (24) Kim, M.; Yeon, J.; Park, J. Biomed. Microdev. 2007, 9, 25–34. (25) Frisk, T.; Rydholm, S.; Liebmann, T.; Svahn, H.; Stemme, G.; Brismar, H. Electrophoresis 2007, 28, 4705–4712. (26) Saadi, W.; Rhee, S.; Lin, F.; Vahidi, B.; Chung, B.; Jeon, N. Biomed. Microdev. 2007, 9, 627–635. (27) Wong, A. P.; Perez-Castillejos, R.; Christopher Love, J.; Whitesides, G. M. Biomaterials 2008, 29, 1853–1861. (28) Huang, C.; Lu, J.; Seon, H.; Lee, A.; Flanagan, L.; Kim, H.; Putnam, A.; Jeon, N. Lab Chip 2009, 9, 1740–1748. (29) Chung, S.; Sudo, R.; Zervantonakis, I. K.; Rimchala, T.; Kamm, R. D. Adv. Mater. 2009, 21, 4863–4867. (30) Mack, P. J.; Zhang, Y.; Chung, S.; Vickerman, V.; Kamm, R. D.; Garcia-Cardena, G. J. Biol. Chem. 2009, 284, 8412–8420. (31) Vickerman, V.; Blundo, J.; Chung, S.; Kamm, R. Lab Chip 2008, 8, 1468. (32) Ong, S.; Zhang, C.; Toh, Y.; Kim, S.; Foo, H.; Tan, C.; van Noort, D.; Park, S.; Yu, H. Biomaterials 2008, 29, 3237–3244. (33) van Noort, D.; Hutmacherbgh, D.; Yu, H. Lab Chip 2007, 7, 302–309. (34) Even-Ram, S.; Yamada, K. Curr. Opin. Cell Biol. 2005, 17, 524–532. (35) Pedersen, J.; Swartz, M. Ann. Biomed. Eng. 2005, 33, 1469–1490. (36) Griffith, L.; Swartz, M. Nat. Rev. Mol. Cell Biol. 2006, 7, 211–224. (37) Yamada, K.; Cukierman, E. Cell 2007, 130, 601–610. (38) Helm, C.; Fleury, M.; Zisch, A.; Boschetti, F.; Swartz, M. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 15779. (39) Chang, Y.; Munn, L.; Hillsley, M.; Dull, R.; Yuan, J.; Lakshminarayanan, S.; Gardner, T.; Jain, R.; Tarbell, J. Microvasc. Res. 2000, 59, 265–277. (40) Zandstra, P.; Conneally, E.; Petzer, A.; Piret, J.; Eaves, C. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 4698. (41) Sieminski, A. L.; Was, A. S.; Kim, G.; Gong, H.; Kamm, R. D. Cell Biochem. Biophys. 2007, 49, 73–83. (42) Gerhardt, H. VEGF Dev. 2008, 68–78.

8459

dx.doi.org/10.1021/ac202170e |Anal. Chem. 2011, 83, 8454–8459