Biomed Microdevices (2010) 12:443–455 DOI 10.1007/s10544-010-9401-x
Measurement of single-cell adhesion strength using a microfluidic assay Kevin V. Christ & Kyle B. Williamson & Kristyn S. Masters & Kevin T. Turner
Published online: 9 March 2010 # Springer Science+Business Media, LLC 2010
Abstract Despite the importance of cell adhesion in numerous physiological, pathological, and biomaterialrelated responses, our understanding of adhesion strength at the cell-substrate interface and its relationship to cell function remains incomplete. One reason for this deficit is a lack of accessible experimental approaches that quantify adhesion strength at the single-cell level and facilitate large numbers of tests. The current work describes the design, fabrication, and use of a microfluidic-based method for single-cell adhesion strength measurements. By applying a monotonically increasing flow rate in a microfluidic channel in combination with video microscopy, the adhesion strength of individual NIH3T3 fibroblasts cultured for 24 h on various surfaces was measured. The small height of the channel allows high shear stresses to be generated under laminar conditions, allowing strength measurements on well-spread, strongly adhered cells that cannot be characterized in most conventional assays. This assay was used to quantify the relationship between morphological characteristics and adhesion strength for individual well-spread cells. Cell adhesion strength was found to be positively correlated with both cell area and circularity. Computational fluid dynamics (CFD) analysis was performed to examine the K. V. Christ : K. S. Masters : K. T. Turner Materials Science Program, University of Wisconsin, Madison, WI 53706, USA K. B. Williamson : K. S. Masters : K. T. Turner Department of Biomedical Engineering, University of Wisconsin, Madison, WI 53706, USA K. T. Turner (*) Department of Mechanical Engineering, University of Wisconsin, 1513 University Avenue, Madison, WI 53706-1572, USA e-mail:
[email protected]
role of cell geometry in determining the actual stress applied to the cell. Use of this method to examine adhesion at the single-cell level allows the detachment of stronglyadhered cells under a highly-controllable, uniform loading to be directly observed and will enable the characterization of biological events and relationships that cannot currently be achieved using existing methods. Keywords Cell adhesion . Cell mechanics . Shear flow . Microfluidics
1 Introduction Cell adhesion is an important process that is associated with many aspects of cell behavior. The adhesive interactions between cell membrane receptors and extracellular matrix (ECM) ligands are known to affect cell growth (Huang and Ingber 1999), differentiation (Danen and Sonnenberg 2003), and migration (Lauffenburger and Horwitz 1996). In addition, physiological phenomena such as wound healing (McEver 2001) and pathological conditions such as cancer metastasis (Brakebusch et al. 2002) are characterized by unique cell-binding events. An understanding of cell adhesion is also essential in the development of implantable biomaterials and devices. The functionality of tissue engineering scaffolds (Griffith and Naughton 2002), implantable bioMEMS (Grayson et al. 2004), biosensors (Frost and Meyerhoff 2006), and artificial bone and teeth replacements (Anselme 2000) is highly dependent on adhesive interactions with surrounding cells and tissue. The process of cell attachment to surfaces in vitro can be described as a biphasic process that consists of an initial binding event followed by adhesion strengthening (Garcia and Gallant 2003). Immediately after it is seeded on a
444
substrate, the cell possesses a spherical conformation and is only weakly adhered. Over the course of many hours, the cell spreads and develops focal adhesions. Eventually, the cell will reach a steady state in which it is well-spread, stronglyadhered and fully functional. While significant progress has been made in characterizing the biomolecules and signaling pathways relevant to cell adhesion (Wiesner et al. 2005), the mechanical aspects of adhesion and the relationship between adhesion strength and cell behavior are not fully understood (Bershadsky et al. 2006). Analyses of cell migration and spreading can provide some information about adhesion, but they do not provide quantitative information about the strength of the cell-ECM interface. Thus, numerous techniques have been developed to measure adhesion strength (see recent reviews: Christ and Turner 2010; Garcia and Gallant 2003), and these assays generally fall into two categories: population methods and single-cell approaches. The simplest population method are wash assays that determine the fraction of cells that remain adhered after one or more washings. These assays provide basic qualitative adhesion data, but they are difficult to reproduce due to the poor control and characterization of the hydrodynamic forces involved. Centrifugation techniques employ standard laboratory centrifuges to apply forces to cells adhered to substrates and typically examine the fraction of cells that remain adhered after loading (Koo et al. 2002; Reyes and Garcia 2003). While this technique is widely used because it does not require specialized equipment, each experiment only allows a single force to be applied to the cells. In addition, common centrifuges cannot apply sufficiently large forces to detach well-adhered cells, and thus are often limited to short-term adhesion studies (cells adhered for ∼1 h or less) (Garcia and Gallant 2003). Hydrodynamic techniques, which are widely used, apply well-controlled shear stresses to adherent cells through fluid flow. A variety of methods have been developed which employ this concept, including the spinning disk apparatus (Gallant et al. 2005; Garcia et al. 1997), radial flow chambers (Goldstein and DiMilla 1998), and rectangular parallel plate flow chambers (Truskey and Pirone 1990; Truskey and Proulx 1993; van Kooten et al. 1992). Spinning disk experiments can generate high shear stresses and detach strongly-adhered cells, but are not compatible with real time microscopy. Fluid flow chambers allow the application of uniform stress fields to cells while enabling visualization of the cells via microscopy, but they typically are unable to generate stresses high enough to detach wellspread cells under laminar flow conditions. The transition from laminar to turbulent flow occurs at a fixed Reynolds number, which is a strong function of the height of channel. Most conventional parallel plate assays have a channel height greater than 200 μm, which is limited by the dimensions of the gaskets and spacers used in their
Biomed Microdevices (2010) 12:443–455
construction. A recent improvement to flow assays has been achieved via the use of devices where hydrodynamic loads are generated by flow within microfluidic channels (Gutierrez and Groisman 2007; Lu et al. 2004). While the microfluidic technique is similar to the classical parallel plate flow chamber, a key advantage is the small characteristic dimension of the flow channel (h>h):
2.1 Microfluidic device mechanics Flow in a microfluidic channel is used to apply hydrodynamic loading to the adhered cells in the present work. The characteristics of the flow and the shear stress applied to the cell were calculated using classical fluid mechanics. The following applies to a constant volumetric flow rate, Q, in a channel that has height, h, and width, w (Fig. 1). The Reynolds number, which is the ratio of inertial to viscous forces, is used to assess if flow is laminar and is defined as: Re ¼
t¼
3m2 Cfw Re: r h2
ð4Þ
As the transition from laminar to turbulent flow occurs at a fixed Reynolds number, this form of the shear stress equation demonstrates that reducing the height of the channel allows higher shear stresses to be achieved under laminar conditions. 2.2 Microfluidic device design and fabrication
r UDh ; m
ð1Þ
where ρ is the fluid density, μ is the viscosity, U is the average flow velocity ðU ¼ Q=whÞ, and Dh is the hydraulic diameter ðDh ¼ 2wh=ðw þ hÞÞ. Flow in micro- and largescale channels can generally be considered laminar up to 4.0 mm w = 1.0 mm
L = 22.9 mm cells
h = 57 µm
Fig. 1 Top and side view of the microfluidic device used. The nominal volume of the device 3.87 µL. The white circles represent the inlet and outlet ports
The central component of the microfluidic device is a rectangular microchannel 1.0 mm in width, 57 µm in height, and 22.9 mm in length (Fig. 1). At both ends, short tapered sections connect the main channel to square 4 mm×4 mm areas for inlet and outlet tube connections. A syringe pump, with a maximum flow rate of 1700 ml/hr and force up to 156 N, (NE-1000, New Era Pump Systems, Wantagh, NY) was used to supply a controlled flow. For the device in Fig. 1 and a flow rate of 950 ml/hr (the maximum flow rate used, see below), the shear stress generated is 507 Pa, Re=497, and Le =3 mm. The finite width correction factor in Eq. 3 is 1.0376 for this channel geometry. Using an analysis for a finite width channel (Hartnett and Kostic 1989), it was found that the flow across the width of the channel is quite uniform—the shear stress is within ± 1 Pa across the center 80%. This was confirmed by CFD simulations as well.
446
The devices were fabricated using PDMS replica molding (Xia and Whitesides 1998). The photomask was created using Adobe Illustrator CS 12.0.1 (Adobe Systems, Inc., San Jose, CA) and printed on a transparency film (Imagesetter, Inc., Madison, WI). A negative photoresist, SU-8 50 (Microchem, Inc., Newton, MA), was spun onto a 3-inch diameter silicon wafer at a rate of 2,350 rpm for 25 s (SCS G3-8 Spincoat, Specialty Coating Systems, Indianapolis, IN), and subsequently exposed to UV light (Series 1000 OmniCure, EXFO, Quebec, QC) through the photomask. After developing, raised SU-8 features that define the channel mold remained on the wafer surface. The heights of the SU-8 channel molds on the master were measured with a white-light interferometer (NewView 6300, Zygo, Middlefield, CT), and all channels were 57±1 µm high. PDMS (Sylgard 184, Dow Corning, Midland, MI) was mixed in a 10:1 base to curing-agent ratio, degassed, then poured over the master, which was set in a petri dish. The PDMS was cured on a hotplate for 4 h at 85°C. The thickness of the cast PDMS was measured to be approximately 2 mm. Tubing interconnects were created by punching holes into the PDMS replicas using a machined syringe needle (16G1, BD, Franklin Lakes, NJ). The PDMS replicas were subsequently bonded to glass coverslips (#1, 50×24 mm, Fisher). Prior to bonding, the PDMS replicas were cleaned with ethanol and the coverslips were cleaned with acid (3:1, 1 N H2SO4 : 1 N HNO3). Both the replicas and the coverslips were exposed to air plasma (PDC-001, Harrick Plasma, Ithaca, NY), brought into contact immediately thereafter, and then annealed at 135°C for 2 min on a hotplate. 2.3 Cell preparation in the microfluidic channels Unless otherwise stated, all reagents were from SigmaAldrich, St. Louis, MO. Following fabrication, the devices were manually flushed sequentially with ethanol and phosphate buffered saline (PBS), and then filled with a protein solution composed of either type I collagen (PureCol, Inamed Biomaterials, Fremont, CA) or fibronectin diluted in bicarbonate coating buffer (15 mM Na2CO3, 35 mM NaHCO3). The collagen (CL) or fibronectin (FN) was applied to substrates to achieve coating densities of 0.165 and 16.5 µg/cm2 for collagen and 0.057 and 5.7 µg/ cm2 for fibronectin. Following CL or FN adsorption at 4°C for 24 h, non-specific binding in the channels was blocked with a 3% bovine serum albumin (BSA, Fisher Scientific Research, Pittsburgh, PA) solution for 1 h prior to cell seeding. The cells, from an immortalized fibroblast cell line (NIH3T3, ATCC, Manassas, VA), were cultured in Dulbecco’s Modified Eagle’s Medium supplemented with 10% (v,v) FBS (Hyclone, Logan, UT), 2 mM L-glutamine, 100 µg/mL streptomycin, 1 U/ml penicillin, were dispersed
Biomed Microdevices (2010) 12:443–455
in growth medium at a density of 1,000,000 cells/ml and then delivered into the channels via syringe pump. The devices were placed in an incubator (37°C, 5% CO2) for 24 h prior to the adhesion strength experiments to allow the cells to adhere and spread. The channels were sparsely populated (5,700 cells/cm2) so that single cells, which adhere in random locations throughout the channels, could be analyzed. A separate set of control devices were incubated in bicarbonate buffer that did not contain an ECM protein and were then blocked and seeded as described above. 2.4 Microfluidic adhesion strength assay test procedure The syringe pump was used to provide the flow for the adhesion strength experiments. A 30 ml syringe (BD) was filled with PBS and attached to a blunted-end syringe needle (25G, Jensen Global, Inc., Santa Barbara, CA), which was, in turn, attached to tubing (0.020 in. ID, 0.060 in. OD, Tygon, Fisher). The inlet tubing (from the syringe) and the outlet tubing were attached to the ports of the device, which was secured to the stage of an inverted microscope (Axiovert 40 CFL, Carl Zeiss, Inc., Oberkochen, Germany). A 20× phase contrast objective and a digital camera (PL-A741, Pixelink, Ottawa, Canada) connected to a computer were used to monitor the cells in real time. The entire device was scanned to find cells suitable for study. Cells that were located in the center 80% of the channel width, at least 4 mm from the channel inlet, more than 2 mm from the end of the channel, and were not overlapping or touching were considered suitable. Most cells studied had an intercellular spacing of at least 50 µm. Three cells with sufficient intercellular spacing could typically be captured in one image (330 µm×413 µm). For the adhesion strength experiments, a Matlab script (The Mathworks, Inc., Natick, MA) was used to control the syringe pump and apply a specified flow rate ramp. The ramp began at 50 ml/hr and increased stepwise every second by 5 ml/hr up to 950 ml/hr over a total duration of 3 min. This corresponds to a change in shear stress of approximately 2 Pa every second. Images were captured at a rate of 1 Hz during the experiments. The time from when the device was removed from the incubator to the time of experiment completion was typically about 10–15 min per device, and each device was used only once. The flow rate, Q, at the time that a cell detached was determined by calculating the time difference between the first and final images of the cell. The nominal shear stress applied at cell detachment was then calculated from Eq. 3. Images captured before flow initiation were analyzed using ImageJ (NIH, Bethesda, MD) in order to determine the geometric properties of the cells, including the area, circularity, and angle of orientation. An outline of each cell
Biomed Microdevices (2010) 12:443–455
447
was traced and the area, A, perimeter, P, and circularity, C, (Eq. 5), were calculated. C¼
4pA P2
ð5Þ
The cell orientation in the channel was determined by fitting an ellipse to the cell outline and calculating the angle between the major axis of the ellipse and the long axis of the channel. Only cells that had an elongated shape were included in the orientation analysis. Based on the results of repeated tracings of an individual cell by multiple users, it was found that manually tracing the cell resulted in an uncertainty of about 3% in the computed cell area. 2.5 Finite element modeling of channel deformation and channel height correction PDMS microchannels can experience significant deformation due to the pressure generated by the flow through them (Gervais et al. 2006; Holden et al. 2003). In the devices used here, such deformations are potentially severe enough to alter the shear stress on the channel floor by as much as a factor of 3. Finite element analysis (FEA) was used to develop a method to account for the deformation. As the fluidic resistance, and hence the pressure drop, through the rectangular microchannel is much larger in comparison to the other sections of the device, it was independently modeled in the finite element program ABAQUS 6.4-1 (Dassault Systèmes, Suresnes, France). Symmetry along the channel length was employed, and the total height and width of the channel walls were determined based on measurements of the actual devices used in the experiments (2 mm and 3 mm, respectively). The length of the channel was 22.9 mm, the height was 57 µm, and the width was 1 mm. The channel was modeled with 9,100 20-node brick elements (C3D20), and the material properties were defined in terms of a Poisson’s ratio (0.499) and a Young’s modulus (1.9 MPa, a value that was measured independently by tensile testing small strips of PDMS and which is in agreement with published values (Gervais et al. 2006; Roca-Cusachs et al. 2005)). The pressure drop along the channel was calculated and varies linearly along the channel length. A corresponding surface pressure was applied to each element of the channel ceiling. Because of the low aspect ratio of the channel cross-section, pressure effects on the channel side walls were neglected (Gervais et al. 2006). A simple meshdensity study was completed to ensure convergence of the FE results. The shear stress on the channel floor changes significantly due to the channel deformation and is no longer constant along the channel length. As described in the “Appendix”, the mechanical deformation predicted using
the finite element analysis (FEA) was coupled to the fluid mechanics of the channel to determine an equation that approximates the height of the channel as a function of flow rate and position in the channel: hcorr ¼ 1:69 109 x 3:89 1011 Q2 þ 3:83 106 x þ 8:79 108 Q 1:05 104 x þ 5:94 105 ;
ð6Þ
where hcorr and x are in units of m, and Q is in units of ml/ hr. Eq. 6 is provided here as it is used to determine the height in Eq. 3 when calculating the shear stresses in the experiments. 2.6 Measurement of cell topography Glass coverslips were cleaned with acid and exposed to air plasma as described above. Afterwards, they were manually washed with ethanol, followed by rinsing in PBS. Collagen diluted in bicarbonate coating buffer was adsorbed onto the coverslip surfaces at a coating density of 16.5 µg/cm2 in slide chambers for 24 h at 4°C. The coverslips were then rinsed with PBS, and 3% BSA was applied for 1 h. Cells were seeded at a density of 50,000 cells/cm2, and allowed to adhere for 24 h. The samples were then prepared in a manner typical of those used in sample preparation for scanning electron microscopy in order to prevent significant alterations in cell morphology (King 1991). At 24 h following cell seeding, the cells were washed with PBS and fixed in 10% neutral buffered formalin. The samples were air-dried and then sputter-coated with an approximately 200 nm thick layer of gold. The Zygo interferometer was used to acquire topographical data on the cells, and the MetroPro 8.1.5 (Zygo) program was used to process the data. A low-pass filter was employed to reduce the noise. The data were then analyzed using a Matlab script to calculate the cell volume and height. The measurements were completed using a procedure similar to that described in (Revell et al. 2006). 2.7 CFD modeling of shear stress on cells In order to analyze how cell morphology influences the forces that fluid flow exerts on the cell, computational fluid dynamics simulations were carried out using the CFD package FLUENT 6.3.26 (ANSYS, Inc., Canonsburg, PA). The simulations also allowed verification of flow characteristics of the microchannel such as the entry length and the shear stress. In this analysis, cells were modeled as spherical caps inside of a rectangular channel with the same cross-section as the experimental channel (h=57 µm, w=1.0 mm). The channel and the cell, which was modeled
448
as a protuberance on the channel floor, were meshed with approximately 112,000 elements (8-node bricks and 6-node wedges). A channel length of 2 mm was long enough for the flow to fully develop at a flow rate of 100 ml/hr, and these two parameters were held constant throughout the simulations. The center of the cell base was located at 1.76 mm into the channel and centered across the width, and symmetry along the channel was exploited. The cell was approximated as a spherical cap. The volume of the cap was set equal to that of a typical cell, in this case 905 µm3. Based on experimental observations, the base of the cap was set equal to areas that ranged from 500 µm2 to 4,000 µm2. The cell height was determined from the areas and volume. The corresponding maximum cap heights were similar to the cell heights found by analyzing the cell topography as described above, rendering the cap geometry sufficient to serve as a basic model of the spread cell. Though PBS was used in the actual adhesion strength experiments, its fluid properties are close to those of water, so the properties of water at 20°C (ρ = 998 kg/m3, μ=1.002×10−3 Pa s) were used throughout. In all simulations, the average flow velocity was specified at the inlet, while an outlet pressure was specified at the end of the channel. Every simulation was solved in double precision using the pressure-based solver with second-order upwind momentum discretization, standard pressure discretization, and the SIMPLE pressure-velocity coupling method. Solution convergence was achieved when the residuals from the continuity and momentum equations reached 10−6.
Fig. 2 Distributions of measured adhesion strength for two different ECM proteins (CL and FN) and two different ECM concentrations
Biomed Microdevices (2010) 12:443–455
Typical simulations required fewer than 500 iterations and less than 1 h to converge. 2.8 Statistical analysis The results of the adhesion strength experiments were analyzed using one-way ANOVA with Tukey’s post-hoc means comparison and a t test for simple linear regression (Origin 7.5 SR6, OriginLab Corp., Northampton, MA). P values less than or equal to 0.05 were considered statistically significant. Data are presented as mean ± standard deviation.
3 Results 3.1 Single-cell microfluidic adhesion assay The results of the microfluidic-based single-cell adhesion strength experiments are presented in Fig. 2. Though a relatively wide distribution of adhesion strengths was measured, the higher concentrations of adsorbed protein were associated with significantly higher average adhesion strengths for both the FN and CL conditions (Table 1). The increase in FN coating density had a more pronounced effect on adhesion strength than the increase in CL density. Also shown in Table 1, cells on all FN- and CL-coated conditions were significantly more strongly adhered than those on uncoated control surfaces (n=71, average adhesion strength = 67±27 Pa).
Biomed Microdevices (2010) 12:443–455
449
Table 1 Statistical analysis of adhesion strength experiments Sample 1
Sample 2
CL, 0.165 µg/cm2 FN, 0.057 µg/cm2 Control Control Control Control
CL, FN, CL, CL, FN, FN,
16.5 µg/cm2 5.7 µg/cm2 0.165 µg/cm2 16.5 µg/cm2 0.057 µg/cm2 5.7 µg/cm2
P value 0.00712 0.00001 0.00031
