Quantitative Measurements of the Strength of Adhesion of Human

(b) Distribution of the surface stress across the channel, τs(y), along the line ...... Amir Manbachi , Shamit Shrivastava , Margherita Cioffi , Bong...
0 downloads 0 Views 496KB Size
Download PDF
Anal. Chem. 2007, 79, 2249-2258

Quantitative Measurements of the Strength of Adhesion of Human Neutrophils to a Substratum in a Microfluidic Device Edgar Gutierrez and Alex Groisman*

Department of Physics, University of California, San Diego, 9500 Gilman Drive, La Jolla, California 92093

We describe a quantitative assay of the strength of adhesion of activated and nonactivated human neutrophils to a substratum, which is carried out in a custom-made microfluidic device. The strength of adhesion is quantified by the fraction of cells remaining adherent (ACF) after a given time of exposure to shear stress in a test microchannel. The microfluidic device is made of two layers of poly(dimethylsiloxane) with integrated membrane valves. This construction allows concurrent testing of two different populations of cells, as well as setting well-defined times of exposure of cells to stress and of their incubation prior to the exposure. The test microchannels have a tapered profile, exposing cells to nearly an order of magnitude range of shear stress. ACF is measured periodically by computer-controlled videomicroscopy scans of the device, with up to 60 000 individual cells identified within a 90 seconds scan. The high throughput of the scans allows reliable quantitative assessment of the ACF. Adhesion of untreated neutrophils and neutrophils activated with formyl-Met-Leu-Phe was tested concurrently in a series of experiments with a fibrinogen-coated glass substratum. At optimized testing conditions, the ACF of activated cells was consistently found to be three times higher than that of nonactivated cells. An adhesion assay could be completed within 11 min from the loading of cells into the device without any intervention by the operator. The proposed device and assay could be used to assess the state of activation of neutrophils in human blood with a potential application to diagnostics of inflammation. Adhesion of mammalian cells to other cells and connective tissue is an important physiological process that often determines the function and fate of cells. Adhesion is mediated by a host of specific receptor-ligand bonds between cells and extracellular matrixes. Cell adhesion plays an important role in the formation of organ tissues during embryonic development and of vascular networks during angiogenesis.1-3 Much cell adhesion activity in * To whom correspondence should be addressed. E-mail: [email protected]. (1) Forgacs, G.; Newman, S. A. Biological physics of the developing embryo; Cambridge University Press: New York, 2005. (2) Zubar, R. V. Trends in angiogenesis research; Nova Biomedical Books: New York, 2005. (3) Brooks, P. C.; Clark, R. A. F.; Cheresh, D. A. Science 1994, 264, 569-571. (4) Pearson, J. D. In Progress in inflammation research; Birkhauser: Basel, 1999. 10.1021/ac061703n CCC: $37.00 Published on Web 02/17/2007

© 2007 American Chemical Society

the human body occurs in blood vessels and involves platelets and leukocytes. Over two-thirds of leukocytes circulating in blood are neutrophils, which are the primary cells involved in inflammation. Stimulated by early inflammatory cues, neutrophils undergo a process of activation that involves changes in conformation of adhesion molecules on their surface.4 These changes allow flowing neutrophils to adhere to the blood vessel endothelium and subsequently cross the endothelial cell barrier and migrate into tissue. Inside the tissue, neutrophils generate an inflammatory response by secreting more inflammatory cues, which activate still more neutrophils.5 The propensity of neutrophils to adhere to a substratum and the strength of the adhesion are indicators of their state of activation.6,7 Therefore, testing the strength of adhesion of neutrophils to a substratum can provide information about the presence of inflammatory cues in blood and ongoing inflammation. The strength of adhesion of neutrophils to a substratum is often measured in multiwell plates.8-12 A suspension with a known concentration of cells is introduced into the wells; cells are incubated for 30-60 min and then exposed to a strong flow generated either with a micropipet or using centrifugation. Cells remaining attached to the bottom of the well are consecutively fixed and stained, and their number is evaluated under a microscope. Although this method enables rather reliable differentiation between activated and nonactivated neutrophils,11,12 the assays involve multiple consecutive steps and require extended time to complete. In most cases, the initial number of cells loaded into a well is not measured directly. Cells remaining attached to the substratum are usually counted under high-magnification microscopes,11,12 limiting the field of view and throughput. (5) Harlan, J. M., Liu, D. Y., Eds. Adhesion: its role in inflammatory disease; Breakthroughs in Molecular Biology; W.H. Freeman: New York, 1992. (6) Lawrence, M. B.; Smith, C. W.; Eskin, S. G.; McIntire, L. V. Blood 1990, 75, 227-237. (7) Brown, E. Semin. Hematol. 1997, 34, 319-326. (8) Dejana, E. C. M. In Methods in cell-matrix adhesion; Adams, J. C., Ed.; Methods in Cell Biology 69; Academic Press: San Diego, 2002. (9) Vaporciyan, A. A.; Jones, M. L.; Ward, P. A. J. Immunol. Methods 1993, 159, 93-100. (10) Stjohn, J. J.; Schroen, D. J.; Cheung, H. T. J. Immunol. Methods 1994, 170, 159-166. (11) Kuijper, P. H. M.; Torres, H. I. G.; vanderLinden, J. A. M.; Lammers, J. W. J.; Sixma, J. J.; Zwaginga, J. J.; Koenderman, L. Blood 1997, 89, 21312138. (12) Barbosa, J. N.; Barbosa, M. A.; Aguas, A. P. J. Biomed. Mater. Res., Part A 2003, 65A, 429-434.

Analytical Chemistry, Vol. 79, No. 6, March 15, 2007 2249

Moreover, in some cases, counting of cell is substituted by indirect evaluation of their number by measurements of bulk optical density10 or cumulative fluorescence.13 In addition, the flow and shear stress applied to cells in multiwell plates are usually not well controlled and not exactly reproducible. Quantitative tests with well-controlled levels of hydrodynamic stress applied to cells can be performed in flow chambers.14,15 For example, flow chambers have been used to assess the strength of adhesion of fibroblasts, endothelial cells, and chondrocytes to different substrata.16-19 Flow chambers have also been employed in experiments on neutrophils, but those experiments focused on the dynamics of attachment of cells to substratum rather than measurements of overall strength of adhesion.20-22 Flow chambers for experiments with adherent mammalian cells normally have a length and width on the scale of centimeters, depth of 0.1-0.3 mm, and uniform cross section. Consequently, all cells in a chamber are subjected to nearly uniform stress, which is proportional to the flow rate through the chamber. By varying the flow rate, the strength of cellular adhesion can be tested under different hydrodynamic stresses. However, interpreting the results of an assay with a varying flow rate can be difficult, since the strength of adhesion of cells to the substratum can depend on the history of the stresses they have experienced.23 This limitation is lifted by the design introduced by Usami et al.,24 in which the width of the flow chambers varies such that the stress at the surface changes linearly with the distance between the inlet and outlet of the chamber. The flow chamber of ref 24 has been used for experiments on attachment of platelets,24 erythrocytes,25 and lymphocytes26 to substratum from flow. Commonly used flow chambers are made of bulky mechanical parts, have to be disassembled and cleaned after each assay, and require relatively large cell samples and large amounts of perfusion liquid per assay. As has been recently shown in experiments with fibroblasts, many drawbacks of flow chambers can be ameliorated by the use of microfludic devices.27 (13) Kartikasari, A. E. R.; Georgiou, N. A.; Visseren, F. L. J.; van Kats-Renaud, H.; van Asbeck, B. S.; Marx, J. J. M. FASEB J. 2005, 19, 353. (14) Vankooten, T. G.; Schakenraad, J. M.; Vandermei, H. C.; Busscher, H. J. J. Biomed. Mater. Res. 1992, 26, 725-738. (15) Bacabac, R. G.; Smit, T. H.; Cowin, S. C.; Van, Loon, J. J.; Nieuwstadt, F. T.; Heethaar, R.; Klein-Nulend, J. J. Biomech. 2005, 38, 159-167. (16) Vankooten, T. G.; Schakenraad, J. M.; Vandermei, H. C.; Busscher, H. J. Biomaterials 1992, 13, 897-904. (17) Xie, H.; Pallero, M. A.; Gupta, K.; Chang, P.; Ware, M. F.; Witke, W.; Kwiatkowski, D. J.; Lauffenburger, D. A.; Murphy-Ullrich, J. E.; Wells, A. J. Cell Sci. 1998, 111, 615-624. (18) Iuliano, D. J.; Saavedra, S. S.; Truskey, G. A. J. Biomed. Mater. Res. 1993, 27, 1103-1113. (19) Schinagl, R. M.; Kurtis, M. S.; Ellis, K. D.; Chien, S.; Sah, R. L. J. Orthop. Res. 1999, 17, 121-129. (20) Reinhardt, P. H.; Elliott, J. F.; Kubes, P. Blood 1997, 89, 3837-3846. (21) Yung, L. Y. L.; Lim, F.; Khan, M. M. H.; Kunapuli, S. P.; Rick, L.; Colman, R. W.; Cooper, S. L. Immunopharmacology 1996, 32, 19-23. (22) Schmidtke, D. W.; Diamond, S. L. J. Cell Biol. 2000, 149, 719-729. (23) Marshall, B. T.; Sarangapani, K. K.; Lou, J.; McEver, R. P.; Zhu, C. Biophys. J. 2005, 88, 1458-1466. (24) Usami, S.; Chen, H. H.; Zhao, Y. H.; Chien, S.; Skalak, R. Ann. Biomed. Eng. 1993, 21, 77-83. (25) Wagner, M. C.; Eckman, J. R.; Wick, T. M. J. Lab. Clin. Med. 2004, 144, 260-267; discussion 227-268. (26) Murthy, S. K.; Sin, A.; Tompkins, R. G.; Toner, M. Langmuir 2004, 20, 11649-11655. (27) Lu, H.; Koo, L. Y.; Wang, W. C. M.; Lauffenburger, D. A.; Griffith, L. G.; Jensen, K. F. Anal. Chem. 2004, 76, 5257-5264.

2250 Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

Figure 1. Microfluidic device. (a) Drawing of the microchannel network. Rectangular and rounded segments of microchannels in the flow layer are shown in black and gray, respectively; channels in the control layer are in blue. Pressure-actuated membrane valves are blue rectangles on top of gray channels. (b) Photograph of the device. Bright rounded spots are the inlets and outlets of the device, shown by circles in (a).

Here we describe the design and operation of a highthroughput microfluidic device (Figure 1), in which the strength of adhesion of cells to a substratum is quantitatively assessed by applying controlled shear stress and recording the dynamics of detachment of cells from the substratum.27 The microfluidic device is made of a transparent silicon elastomer poly(dimethylsiloxane) (PDMS) bonded to a glass slide. The device is advantageous over commonly used flow chambers in two respects: the substratum area and the volume of the microchannels are small, requiring small samples of cells and minimal amounts of reagents for perfusion and for coating the substratum; the device is disposable, which eliminates the need for cleaning and reassembly between assays. Moreover, similar to the flow chamber in ref 24, cells on the substratum are exposed to an order of magnitude wide, continuous range of shear stress and their rate of detachment from the substratum can be concurrently measured in this whole range of shear stress. This last feature is an improvement compared with the previous microfluidic design,27 where only a discrete set of stresses could be tested. In addition, the proposed microfluidic device has integrated membrane valves, which partition it into two compartments with identical sets of microchannels. Two different cell populations are concurrently loaded into the two compartments of the device and are tested in parallel under identical flow conditions, so the adhesion strengths of the two populations can be immediately compared. The valves are also used to seal the microchannels after they are loaded with suspensions of cells, so cells sediment and adhere to the substratum without any perturbation by flow. By setting

an appropriate difference in hydrostatic pressure between an inlet and outlet of the device and by subsequent opening of valves, adhered cells are almost instantaneously exposed to a preset level of shear stress. Thus, repeated experiments can be run in the device, in which cell incubation time and shear stress stimulus are precisely reproduced, and the stimulus has a particularly simple temporal shape of a step function. The device was used to measure the strength of adhesion of human neutrophils to a substratum.28,29 The measurements were performed concurrently for neutrophils activated with formyl-MetLeu-Phe (fMLP) and nonactivated neutrophils that were loaded in the two isolated compartments of the device. The strength of adhesion was quantified by the fraction of cells remaining adhered to the substratum after exposure to a given shear stress for a given time.27 The adherent cell fraction (ACF) was evaluated by direct counting of cells in different areas of the test channels before their exposure to flow and after different durations of flow exposure. Cells were counted using a custom-made image recognition code and a computer-controlled videomicroscopy system. The videomicroscopy system performed automated scans of the entire area of the test channels within 90 s, with a resolution sufficient to capture images of ∼6 × 104 individual cells initially loaded into the device (∼3 × 104 from each group). The experiments did not require any intervention by the operator from the moment of loading of cells into the device. The conditions of the neutrophil adhesion assays were chosen to achieve the maximal difference in the ACF between the activated and nonactivated neutrophils. At optimized testing conditions, the difference in the ACF between the two groups of cells reached a factor of 3. In its basic version, an adhesion assay could be finished within ∼11 min from the moment of loading of cells into the device. The capacity of the device to detect differences between populations of activated and nonactivated neutrophils could be used to evaluate the level of activation of neutrophils extracted from patient blood, with a potential application to diagnostics of inflammation. The use of the device could also be extended to explore differences in adhesion of a homogeneous population of cells to substrata coated with different substances. EXPERIMENTAL SECTION Device Rationale. The microfluidic device (Figure 1) has two layers of channels, the flow layer and the control layer, which are separated by thin membranes in the areas where they overlap. The flow layer has three inlets, three outlets, and eight identical long tapered channels (test channels), in which the strength of adhesion of cells to the substratum is measured. Because of the tapered profile, the mean flow velocity and shear stress in the test channels grow along the flow direction (positive x-direction). The microfluidic device has eight integrated membrane valves (blue rectangles in Figure 1a),30 which are connected through channels in the control layer to four separate control inlets. Application of high pressure to a control inlet closes valves (28) Diamond, M. S.; Springer, T. A. J. Cell Biol. 1993, 120, 545-556. (29) Loike, J. D.; Sodeik, B.; Cao, L.; Leucona, S.; Weitz, J. I.; Detmers, P. A.; Wright, S. D.; Silverstein, S. C. Proc. Natl. Acad. Sci. U. S. A. 1991, 88, 1044-1048. (30) Unger, M. A.; Chou, H. P.; Thorsen, T.; Scherer, A.; Quake, S. R. Science 2000, 288, 113-116.

connected to the inlet and seals flow layer channels under the valves. When control inlets 2 and 3 are pressurized, the device is divided into two mirror-symmetric compartments, each containing a set of four identical test channels, an inlet, and an outlet (cell-A inlet and cell-A outlet in one compartment and cell-B inlet and cell-B outlet in the other compartment; Figure 1a). This partitioning of the device enables independent loading of different cell populations into the two compartments. The presence of four identical test channels in each compartment allows analysis of a large number of cells from each population and collection of large data samples in a single test. The compartmentalization of the device could also be used for in situ coating of the substrata in the two sets of test channels with different molecules. Closing all valves (by pressurizing all control inlets) hermetically seals both sets of test channels, preventing any flow through them. The absence of flow provides a favorable environment for adhesion of cells to the substratum. The flow through the test channels is driven by a positive difference in pressures between the perfusion inlet and perfusion outlet (Figure 1). The pressures are preset when all valves are closed. Simultaneous opening of valves connected to the control inlets 2 and 3 causes an almost instantaneous exposure of cells in both sets of test channels to a steady flow and hydrodynamic stress. The device has a symmetric layout with respect to the two sets of test channels, and the resistances of the lines connecting individual test channels with the perfusion inlet and perfusion outlet are all equal. Therefore, volumetric flow rates through all test channels are equal. Furthermore, flow conditions at equal x-coordinates are identical in all test channels, allowing straightforward comparison of the adhesion strength between two cell populations and evaluation of its variability within each population, both at any shear stress found in the test channels. The segments of the flow layer channels lying under the membrane valves have rounded profiles that are essential for tight sealing of the channels by the valves.30 In contrast, the profiles of the test channels are rectangular for better cross-channel (ydirection) uniformity of stress at the substratum. The test channels have a length L ) 12 mm, a uniform depth h ) 35 µm, and a width, w, that varies linearly from w1 ) 1800 µm at the beginning to w2 ) 225 µm at the end as w ) w1 - (w1 - w2)(x/L), where x is the distance from the beginning of the channel (Figure 2a). The Reynolds number in the test channels can be calculated as Re ) Fvjh/η, where vj is the mean flow velocity, and F ) 1 g/cm3 and η ) 10-3 Pa‚s are the density and viscosity of water, respectively. Under typical experimental flow conditions, the maximal value of Re in a test channel (measured at w ) 225 µm) was ∼0.1, suggesting that the flow was laminar, and nonlinear effects in the flow were negligible. We computed the three-dimensional field of flow velocity in a test channel, b v(x,y,z), at typical experimental conditions using FemLab (Comsol Inc., Los Angeles, CA). We used the computed velocity field to numerically calculate the distribution of shear stress at the substratum, τs(x,y) ) η((∂vx/∂z)2 + (∂vy/∂z)2)1/2, where the partial derivatives ∂vx/∂z and ∂vy/∂z are taken at the bottom of the channel (Figure 2a). At all positions along the channel (x-coordinates), the distributions of the surface stress across the channel, τs(y), are nearly uniform over most of the channel width. In particular, near the end of the channel (P2 in Figure 2a), τs(y) Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

2251

Figure 2. Shear stress at the bottom of a test channel (near the substratum), τs(x, y), at typical experimental conditions (volumetric flow rate through the channel Q ) 1.60 µL/min) calculated with FemLab. (a) Distribution of τs(x, y) in the channel (linear scale; color coded). (b) Distribution of the surface stress across the channel, τs(y), along the line marked P1 in (a) (0.4 mm from the beginning of the channel). (c) τs(y) along the line marked P2 in (a) (0.2 mm from the end of the channel). The channel boundaries are y ) 0 and 1.75 mm in (b) and y ) 0 and 0.25 mm in (c). (d) Dependence of the surface stress in the middle of the channel [along the dashed line in (a)], τmid, on the position along the channel, x (blue circles). Continuous line is a curve τmid ∝ 1/w(x) shown to guide the eye; w(x) is the channel width at the position x.

is within 10% of its maximal value over 75% of the channel width (Figure 2c); near the beginning of the channel (P1 in Figure 2a), τs(y) is within 10% of the maximum over 92% of the channel width (Figure 2b). (The fraction of the channel width, where τs(y) is nearly uniform, grows monotonically as w increases from 225 to 1800 µm.) The dependence of the shear stress at the substratum in the middle of the channel (along the dashed line in Figure 2a), τmid, on the position along the x-axis is shown in Figure 2d. By conservation of mass, the mean flow velocity in the channel, vj, is inversely proportional to the channel width, vj ∝ 1/w, whereas τmid ≈ 6ηvj/h. (The equality τmid ) 6ηvj/h is exact in the case of a rectilinear channel with an infinitely high aspect ratio, w/h f ∞.) Therefore, τmid(x) is close to a curve proportional to 1/w(x) (Figure 2d), and the characteristic shear stress at the substratum increases ∼8-fold between the beginning and end of the channel (where 2252

Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

w ) 1800 and 225 µm, respectively). Therefore, the test channels allow testing the strength of cellular adhesion to the substratum in almost an order of magnitude wide, continuous range of shear stress, with shear stress being nearly uniform over at least 75% of the channel width. Device Fabrication. The microfluidic device is made of two layers of PDMS created using two different master molds. Both molds are fabricated lithographically with 8000 dpi photomasks. The master mold for the first layer of PDMS with flow channels is made in a two-step process. First, a 35-µm layer of UV-curable epoxy (SU8-2015 by MicroChem) is spin-coated onto a 4-in. silicon wafer, exposed to UV light through a photomask and developed, producing the desired rectangular relief pattern. Next, the wafer is coated with a 35-µm layer of a positive photoresist (SPR220-7.0 by Shipley) and exposed through another photomask, aligned with the SU8 pattern. After development, the SPR220 relief is rounded

by baking for 15 min on a 115 °C hot plate, which completes the first master mold. To make the master mold for the second layer of PDMS with control channels, a 35-µm layer of SU8-2015 is spincoated on another silicon wafer, exposed to UV light through a third photomask, and developed. A PDMS prepolymer (5:1 mixture of base and curing agent of RTV615 by General Electric) is poured onto the control layer mold to a depth of ∼5 mm and partially cured by baking at 85 °C for 30 min. The PDMS cast is peeled off the wafer, cut into individual chips, and punched with a gauge 20 Luer stub to make holes for the control layer ports. In parallel, the flow layer mold is spincoated with a 75-µm layer of another PDMS prepolymer (20:1 mixture of the base and curing agent). The spin-coating is performed at 1300 rpm for 30 s. Then the wafer is placed on a horizontal substrate for 30 min to let the prepolymer reflow and to improve the flatness of its surface. The PDMS is partially cured for 30 min at 85 °C, and the 5-mm-thick chips are placed on top of it in correct alignment with the photoresist relief on the mold. After 2 h of baking at 85 °C, a monolith of completely cured PDMS with integrated membrane valves is formed.30 The chips are peeled off from the wafer, and holes are punched in the chips with a gauge 20 Luer stub to make ports for the flow layer channels. The chips are soaked in 1 mM HCl at 90 °C for 1 h and then reversibly bonded to 1-mm-thick glass slides (FisherBrand, 50 × 75 mm; rinsed with methanol before bonding) by baking at 85 °C for at least 2 h. Flow Control. The perfusion buffer solutions fed to the perfusion inlet and drawn off from the perfusion outlet were kept in two 10-mL plastic syringes, which were held upright with the Luer connectors at the bottom and were open to the atmosphere at the top. The syringes were connected to the microfluidic device through Tygon tubing with an inner diameter of 0.5 mm. The syringes were attached to stages, which slid on a vertical rail with a precise ruler. The elevation of the liquid in the syringe connected to the perfusion inlet above the liquid level in the syringe connected to the perfusion outlet created a differential hydrostatic pressure that drove the flow. The control inlets were connected through four separate three-way solenoid valves to a source of compressed air with a pressure of 10 psi. The valves were controlled by a homemade driver interfaced with a computer through a NI 6503 card, which was programmed with LabView7.1 (National Instruments). When a solenoid valve was powered, the corresponding control inlet (normally vented to atmosphere) was pressurized and the membrane valves connected to the inlet were closed. Human Neutrophil Isolation. Human blood from healthy volunteers, three males and two females, was collected in 6-mL sodium heparin tubes (Vacutainer by Becton Dickinson) and used the same day in accordance with university Human Subjects Protocol 050306. Neutrophil isolation was achieved by densitygradient centrifugation. A 5-mL volume of whole blood was carefully layered over 5 mL of a density-gradient solution (Polymorphprep by Axis-Shield, Dundee, UK) in a 15-mL conical centrifuge tube. The sample was centrifuged at 500g for 35 min in a swing-bucket rotor at room temperature. After centrifugation, the lower of the two leukocyte bands containing neutrophils was carefully aspirated, transferred to a separate tube, and diluted with an equal volume of 0.5× phosphate buffer saline (PBS) to restore

osmolarity. Neutrophils were pelleted by centrifugation at 400g for 10 min. In parallel, the plasma on top of the original blood sample was aspirated, transferred to another tube, and centrifuged at 1200g for 10 min to remove platelets and to obtain platelet-free plasma (PFP); 0.25 mL of PFP was used to resuspend the pelleted neutrophils. To remove residual erythrocytes, resuspended neutrophils were layered over 0.5 mL of Polymorphprep in a 1.5-mL Eppendorf tube and spun at 400g in a microcentrifuge for 5 min. The band of cells formed at the interface was collected; the cells were washed in 0.25 mL of 1:1 mixture of PBS and PFP and resuspended in 50 µL of the PBS-PFP mixture to make a stock cell suspension. Cell viability was assessed by Trypan blue exclusion. Immediately before loading cells into the microfluidic device, two ∼5-µL aliquots were taken from the stock cell suspension to make the activated and untreated cell groups. Each aliquot was gently mixed with 2 volumes of PFP diluted with 50% Hanks Balanced Saline (containing calcium, magnesium, and 10 mM HEPES) and 0.5 µmol of fMLP was added to one aliquot to activate cells. The volumes of the two resulting cell suspensions were ∼15 µL each, and the neutrophil densities were (10-15) × 106 cells/mL in both of them. Microscopy. The microfluidic device was mounted on a motorized stage (OptiScan by Prior Scientific) on a Nikon TE2000 inverted microscope equipped with a 10×/0.45 objective lens, a 0.5× video adapter, and a Sony X900 IEEE1394 camera with a 1/ -in., 1280 × 960 pixels CCD array. Motions of the stage and 2 image acquisition were controlled through RS232 and IEEE1394 interfaces, respectively, using a code in LabView7.1. (The same code also controlled closing and opening of the device membrane valves.) In one scanning loop that took ∼90 s, the stage was programmed to move through 184 stations covering the whole working area of the test channels, with one bright-field image taken at every station. Each test channel was divided into 14 evenly spaced consecutive imaging sections with the same extension along the x-axis (∼0.86 mm) that was ∼10% smaller that the width of the field of view of the camera. To cover the whole area of a test channel, two separate images per section were taken close to the beginning of the channel (where the channel was wide) and a single image per section was taken near the end of the channel (the total of 23 images/channel). Adhesion Assay. Shortly before an experiment, the device was treated for 1 min with air plasma in a Plasma-Preen II apparatus to make PDMS and glass more hydrophilic. The flow layer channels were filled with 70% ethanol in water and subsequently rinsed with excess volumes of PBS. Next, the channels were filled with a 0.5 mg/mL solution of human fibrinogen (SigmaAldrich, St. Louis, MO) in PBS, and the device was left at room temperature for 30 min to allow adsorption of fibrinogen onto the glass substratum. The high concentration of fibrinogen ensured saturated protein adsorption to the glass surface to prevent undesirable adsorption of serum proteins during the cell incubation. Subsequently, the channels were rinsed with excess volumes of PBS and remained filled with PBS until cell loading. For independent loading of the two cell suspensions into the device without cross-contamination, control inlets 2 and 3 were simultaneously pressurized, separating the flow channel network into two compartments (Figure 1). Each of the two cell suspenAnalytical Chemistry, Vol. 79, No. 6, March 15, 2007

2253

sions was aspirated with a separate 25-µL Hamilton syringe with a blunt Luer needle. The suspensions of fMLP-activated and untreated cells were loaded into the device through the cell-A and cell-B inlets, respectively. To avoid unintended activation by hydrodynamic stresses during the aspiration and loading cells into the device, care was taken to minimize agitation and to carry out the aspiration and loading slowly. The interval between the loadings of the two suspensions was typically ∼30 s. After both suspensions were loaded, control inlets 1 and 4 were pressurized and remained so through the rest of experiment. Cells were allowed either 7 or 12 min to sediment and attach to the substratum at room temperature in the absence of flow. The pressure difference between the perfusion inlet and perfusion outlet was set to ∆P ) 1.25 kPa, and the flow was initiated by synchronously depressurizing control inlets 2 and 3. To record the distribution of cells and quantify the initial cell density, a channel videomicroscopy scanning loop was performed immediately before the flow through the channels was started. The perfusion buffer was PBS containing 0.9 mM calcium, 0.5 mM magnesium, and 10 mM HEPES. (The divalent cations were added to promote integrin-ligand binding.) Cells were exposed to flow for a total of 10 min. After every 2 min of continuous flow, the flow was stopped to perform a channel scanning loop and was then started again. Data Analysis. Cell detachment dynamics was analyzed postassay. The number of cells in a micrograph was evaluated automatically using a code in LabView Vision (National Instruments). The code implemented consistent and stringent rules for recognition and counting of cells attached to the substratum. The code was adaptively tuned using a series of micrographs of neutrophils on the substratum obtained in pilot experiments to optimize the reliability of cell recognition and counting. Adjusted parameters included intensity threshold, cell area, and cell circularity. The largest source of error in counting was residual erythrocytes, which were sometimes misidentified as neutrophils. The ACF was determined as the number of cells remaining on the substratum after exposure to shear flow, divided by the number of cells present before the flow was applied. Cells from the marginal areas of the channels, where the shear stress was expected to be more than 10% below the stress in the middle, τmid, were excluded from the analysis. The ACF was evaluated separately for each of the 14 sections of the test channels, and the results were averaged between the corresponding sections from the four channels in the set (with cells from the same group). To estimate the uncertainty of the ACF, we calculated the standard deviation of the four ACF values obtained from the different channels. The values lying beyond 1.28 standard deviations from the mean ACF (80% confidence interval) were disregarded (not more than one value per each set of four) and the mean ACF was subsequently recalculated. RESULTS AND DISCUSSION The flow rates at various locations along the test channels were evaluated by measuring the maximal flow velocity, vmax , without cells in the channels. The measurements were performed by seeding the perfusion buffer with 1.9-µm fluorescent latex beads (by Bangs Labs, Fishers, IN), imaging the flow near the midplane of a test channel under fluorescence illumination, and evaluating 2254

Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

the maximal length of streak lines produced by the beads. The distribution of vmax along the channel was well fitted by the results of numerical simulations for a pressure difference of 124 Pa across the computation domain (Supporting Information, Figure S1). We used the results of the simulations to compute the volumetric flow rate through the test channel, Q. It was Q ) 1.60 µL/min. The range of the characteristic shear stress at the substratum, τmid, was from 0.72 to 6.2 dyn/cm2 (Figure 2d), close to the physiological range of shear stress in postcapillary venules, where adhesion of neutrophils to endothelium usually occurs. We also measured the maximal flow velocity, vmax, at x ) 3 mm and x ) 9 mm (Figure 2a) in all eight test channels of the device. At both locations, the values of vmax in individual channels deviated from the value of vmax averaged over the eight channels by not more than 3%, indicating that the volumetric flow rates were practically identical in all test channels, as expected (Supporting Information, Table S1). For unbiased comparison of the ACF between different regions of a channel and between different channels, it is important that the initial number of cells per unit area (cell density) is uniform throughout the channels. Therefore, we measured the distribution of cell densities along the test channels (Supporting Information, Figure S2). The overall mean density was ∼650 cells/mm2, corresponding to a mean distance of ∼40 µm between adjacent cells. The mean densities of untreated and fMLP-activated cells differed by only 5%, and the distribution of cells along the test channels was practically uniform. While there was a measurable decline in cell density from the beginnings to the ends of the channels (∼10 and ∼20% for untreated and fMLP-activated cells, respectively), it was comparable with the density variations between identical sections in nearby channels and between adjacent sections of individual channels (both ∼8%). Representative time series of images of test channels with cells on the substratum are presented in Figure 3. When cells are exposed to the shear stress, there is a notable reduction in the number of fMLP-activated cells and a major reduction in the number of untreated cells. The results on dependence of the ACF on the shear stress and on the time of exposure to the stress for incubation times tinc ) 7 and 12 min are summarized in Figure 4, in which the mean values of ACF are plotted in 12 separate regions along the test channels versus the mean values of τmid in the regions. The data on the location of the individual regions, the ranges of τmid in them, and the average number of cells in the regions are presented in Table 1. The presence of neutrophils on the substratum leads to a reduction of the channel depth available for the flow and to a proportional increase of the shear rate experienced by cells compared with the values of shear rate calculated for the channel without cells (Table 1).31 For an estimated height of the neutrophils of ∼2.5-3 µm and the channel depth h ) 35 µm, the actual shear rate experienced by neutrophils is estimated to be 7-9% higher than in Table 1. (The shear rate would be unchanged by the presence of cells in a channel of an infinitely large depth.) The experiments had a high throughput, with ∼3 × 104 cells from each group analyzed in each experiment. The total number of cells in identical regions of four test channels (with a homogeneous cell population) varied from ∼8000 at the beginning (31) Khismatullin, D. B.; Truskey, G. A. Microvasc. Res. 2004, 68, 188-202.

Figure 3. Micrographs of fragments of two test channels with untreated and fMLP- activated neutrophils attached to the substratum, taken in the middle to the test channels at x ≈ 5 mm from the beginning of the channels at τmid ) 1.2 dyn/cm2. (a)-(c) untreated neutrophils before the flow is started, after 2 min of flow, and after 10 min on flow, respectively. (d)-(f) fMLP-activated neutrophils before the flow is started, after 2 min of flow, and after 10 min of flow, respectively. Scale bar 100 µm.

of the channels to ∼550 at the end of the channels (cf. Table 1). We performed five experiments with tinc ) 7 min and four experiments with tinc ) 12 min. For both incubation times and both cell groups, the rate of reduction of ACF after the first 10 min of the exposure to flow was found to be minimal, and thus, no data were acquired past that time point. In all experiments, ∼5% of the reduction of ACF after the first 2 min was due to residual erythrocytes, which were initially present on the substratum. A characteristic time for a cell detached from the substratum to be washed from the test channel was estimated as ∼7 s (see Supporting Information for details of the estimation). Therefore, cells detached during the last 7 s of the 2-min flow intervals were likely counted as cells attached to the substratum in downstream sections of the channel. We believe that the experimental error due to these cells was minimal, because 7 s constituted only 6% of a flow interval and because the rate of cell detachment was quickly decreasing with the time of their integral exposure to the flow (especially for untreated cells; cf. ACF at 2 and 4 min in Figure 4a, b and the discussion below).

Figure 4. Results of adhesion assays from five experiments with the cell incubation time tinc ) 7 min (a) and four experiments with tinc ) 12 min (b). Plots show fractions of cells remaining adherent, ACF, in 12 different sections of the test channels, as functions of the mean surface stress in the middle the channel, τmid, in the sections. (See Table 1 for details on individual sections.) ACF was evaluated every 2 min. Values of ACF measured at consecutive time points are shown by different symbols and plotted with small positive shifts along the x-axis for clarity of presentation. Filled symbols are untreated cells; open symbols are fMLP-activated cells. Error bars (shown for stress exposure times tstr ) 2 and 10 min only) are standard deviations of different individual tests.

The most striking result of the experiments with tinc ) 7 min (Figure 4a) is a major difference in the strength of adhesion between cells activated with fMLP and untreated cells. For example, after 2 min of exposure, only ∼15-20% of the fMLPactivated cells detached from the substratum, whereas the fraction of untreated cells that detached was ∼70-75%. The difference in Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

2255

Table 1. Specifications for 12 Regions of the Test Channels Used in Analysis of the ACF: the Mean Value of the Surface Stress in the Middle of the Channel, τmid; the Range of τmid; Position of the Region along the x-Axis; Surface Area of the Region; and Mean Number of Cells in the Region in the Beginning of an Experimenta 1 mean stress (dyn/cm2) stress range (dyn/cm2) position (mm) area (mm2) no. of cells a

2

3

4

5

6

7

8

9

10

11

12

0.77

0.89

1.01

1.11

1.22

1.37

1.55

1.79

2.2

2.7

3.5

5.1

0.72-0.82

0.82-0.96

0.96-1.05

1.05-1.16

1.16-1.28

1.28-1.45

1.45-1.64

1.64-1.94

1.9-2.4

2.4-3.0

3.0-4.0

4.0-6.2

0-1.7

1.7-3.4

3.4-4.3

4.3-5.1

5.1-6

6-6.9

6.9-7.7

7.7-8.6

8.6-9.4

9.4-10.3

10.3-11.1

11.1-12

2.89

2.51

1.11

1.01

0.92

0.82

0.72

0.63

0.53

0.43

0.34

0.24

2000

1800

770

720

640

550

480

400

340

270

200

135

Each of the regions 3-12 corresponds to one imaging section, whereas the regions 1 and 2 correspond to two imaging sections.

ACF between the two groups of cells after 2 min was 3-fold on average. The rate of reduction of ACF substantially decreased with the time of exposure to stress, tstr. So, the last 6 min of the exposure (from tstr ) 4 to 10 min) accounted for only 19 and 3% of the ultimate reduction in ACF for the fMLP-activated and untreated cells, respectively (an average between all channel sections). For the fMLP-activated cells, the ACF was practically independent of the stress at tstr ) 2 and 4 min but became increasingly dependent on it at extended exposure times, decreasing by as much as ∼10% between τmid ) 0.72 and 6.2 dyn/cm2 at tstr ) 8 and 10 min (Figure 4a; the reduction of the ACF was estimated by the slope of straight lines fitted to the data points). In contrast, the ACF of the untreated cells consistently exhibited substantial dependence on stress, decreasing by ∼30% between τmid ) 0.72 and 6.2 dyn/cm2 at all time points. The dependence of ACF on the stress and the time of exposure to it with tinc ) 12 min (Figure 4b) had the same general trends as with tinc ) 7 min: the ACF was higher for the fMLP-activated than for untreated cells, and it decreased with the magnitude of stress and the time of exposure to it. Remarkably, however, the ACF of the untreated cells was consistently about twice higher than with tinc ) 7 min at the corresponding tstr, substantially reducing the disparity between the untreated and fMLP-activated cells. Further, the reduction in the ACF during the last 6 min of exposure to stress accounted for a larger portion of the total reduction as compared with tinc ) 7 min; it was 30 and 13% of the total reduction, for the fMLP-activated and untreated cells, respectively. In addition, the dependence of the ACF of the untreated cells on stress at tinc ) 12 min substantially diminished compared with tinc ) 7 min, and the ACF decreased by an average of only ∼8% between τmid ) 0.72 and 6.2 dyn/cm2. In contrast, for the fMLP-activated cells, the dependence of the ACF on τmid became stronger, with the ACF decreasing by ∼15% on average between τmid ) 0.72 and 6.2 dyn/cm2. Increasing the incubation time from 7 to 12 min does not seem to change the mean strength of adhesion of fMLP-activated cells to the substratum, because the mean ACF stays largely the same (Figure 4a and b). However, it appears to subtly modify the character of the adhesive bonds, because the ACF is substantially more stress dependent at tinc ) 12 min than at tinc ) 7 min. This modification could be due to gradual fading of the initial activation response to fMLP (which is applied to the cells ∼1 min before 2256

Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

their loading into the device). Surface-bound fibrinogen is known to slowly activate neutrophils32 that could be an important factor in the dramatic enhancement of the adhesion strength of the untreated cells with the extended incubation time, as evidenced by the doubling of the ACF at tinc ) 12 min compared with tinc ) 7 min. Interestingly, the character of modification of the adhesive bonds by the extended incubation appeared to be different for the untreated cells than for the fMLP-activated cells, with the dependence of the ACF on stress weakening rather than growing. With the extended incubation time, neutrophils could also bind to fibrinogen through receptors that do not require activation. Fibrinogen is known to have at least two binding domains that are not exclusively specific to β2 integrin (the integrin activated by fMLP) and can bind to β1 and LR1 receptors on neutrophils.33 The modification of adhesive bonds (especially for the untreated cells) with incubation time could be reduced by substituting fibrinogen with endothelial cell adhesion molecules that provide more homogeneous receptor-ligand bonds and do not cause neutrophil activation. The adhesion strength of the untreated cells (as measured by the ACF) further increased as tinc was extended beyond 12 min, reducing the difference between them and the fMLP-activated cells to a minimum (data not shown). On the other hand, a tinc less than 5 min led to weak adhesion and low ACF for all cells, which prevented collecting sufficient data to obtain reliable quantitative results. Therefore, to obtain both a large amount of data and a large difference between activated and untreated cells, it was important to have tinc between 7 and 12 min. Because of the dependence of ACF on the incubation time and the time of exposure to fMLP, it was essential to load the two groups of cells into the device with a minimal time interval (∼30 s) and to expose cells to fMLP immediately prior to loading into the device (∼1 min in the protocol we used). Another important condition of the assay was that cells are kept at room temperature at all times. We found that warming the device up to 37 °C and incubating cells at this more physiological temperature led to a substantial increase in the strength of adhesion of the untreated cells. As a result, the difference in the ACF between untreated and fMLP(32) Rubel, C.; Fernandez, G. C.; Dran, G.; Bompadre, M. B.; Isturiz, M. A.; Palermo, M. S. J. Immunol. 2001, 166, 2002-2010. (33) Gresham, H. D.; Adams, S. P.; Brown, E. J. J. Biol. Chem. 1992, 267, 1389513902.

activated cells notably decreased, and the capacity of the assay to distinguish between these two populations of cells significantly deteriorated. The choice to drive the flow with differential pressure ∆P ) 1.25 kPa was made after a series of preliminary tests. We tried to find a value of ∆P that maximized the difference in ACF between activated and untreated cells. In addition, we aimed at ACF values close to 0.5 for both activated and untreated cells, so the numbers of detached cells and of cells remaining adherent were both large. The adhesion assay proved to be highly sensitive to the purity of cell samples. In our preliminary experiments, the densitygradient centrifugation was done in 1.5-mL test tubes, where the separation between neutrophil and lymphocyte bands was small, leading to some contamination of neutrophil samples during their aspiration. (The contamination resulted in increased strength of adhesion of untreated cells compared with the experiments reported in this study.) The problem was resolved by using larger blood samples, 15-mL test tubes, and a swing-bucket rotor for the density-gradient centrifugation. Presumably, lymphocytes activated neutrophils by interacting with them in the high-density suspensions (>106/mL) before cells were loaded into the device. This type of neutrophil activation has been reported in the literature.13,14 The proposed device shares some of the common advantages of microfluidic devices, including small samples of cells (∼5 µL of the stock suspension for each group; total of 20% of neutrophils from 5 mL of whole blood) and small amounts of the perfusion buffer (∼13 µL/min; 8 times the volumetric flow rate through individual test channels) and of the coating substance (fibrinogen) required for assays. Fabricating the device out of PDMS made it disposable and, most importantly, enabled its two-layer construction with integrated membrane valves. Valves were essential to test and compare two different cell populations in a single assay. Valves also made it possible to perform repeated assays at conditions that were highly reproducible in terms of both the time of exposure of cells to stress and the incubation time before the exposure. In addition, closing the valves and stopping the perfusion flow before performing the videomicroscopy scans permitted unbiased comparison of the ACF between all areas of the device at any given time. Tapered shapes of the test channels enabled concurrent exposure of cells to almost an order of magnitude range of shear stresses. The wide range of stresses applied to cells allowed exploring the dependence of ACF on stress and was also important for identification of flow conditions that maximized the disparity in ACF between activated and nonactivated neutrophils. The shape of the test channels was different from that of the flow chamber in ref 24, where the width varied inversely with the distance from an edge of the channel, w ∝ 1/x. A test channel with w ∝ 1/x would have a shape w(x) ) w1/[(x/L)(w1/w2 - 1) + 1] (where w1 and w2 are the widths at the beginning and at the end). Since vj ∝ 1/w, channels with w ∝ 1/x have a benefit of linear variation of the mean flow velocity and characteristic surface stress along the channel, τs ∝ x. Flow chambers with linear variation of τs are generally advantageous for studying the surface stress dependences of various processes, including the dynamics of changes of the ACF with time. Nevertheless, a channel with w ∝ 1/x and the same w1, w2, and L as in this study, would have an almost twice smaller surface area of the substratum. This reduction of

the surface area would lead to twice smaller number of cells interrogated in each experiment that would be disadvantageous for reliable discrimination between different cell populations. Finally, we note that increasing the ratio w1/w2 above 1 order of magnitude would not be very practical, because of the large variation of surface areas corresponding to different τs. An appropriate strategy for covering a large continuous range of τs would be to use a set of identical tapered test channels with different volumetric flow rates, Q, through them (e.g., test channels with w1/w2 ≈ 10 and Q varying by a factor of 3-5). An important component of the experimental system was the computer-controlled videomicroscopy setup. The duration of a scan covering the whole area of the test channels (90 s) was compatible with incubation times as short as 2 min and was smaller than the duration of the individual intervals of continuous perfusion flow (2 min). The resolution of the images acquired during a scan was sufficient to individually identify as many as 6 × 104 cells. This high throughput of the system was important for quantitative measurements of ACF and its stress dependence and for clear differentiation between activated and nonactivated populations of neutrophils. The proposed neutrophil adhesion assay for differentiation between activated and nonactivated neutrophils based on their strength of adhesion to a substratum has several advantages over the commonly used cell adhesion assays.8-12 In the proposed assay, the hydrodynamic stress applied to cells is precisely controlled and exactly reproducible. The strength of cellular adhesion is quantified by a straightforward parameter, ACF, the fraction of cells remaining adhered to the substratum after exposure to a given shear stress for a given time, which is calculated by direct counting of cells before and after the stress exposure. The comparison between the two groups of cells is based on the analysis of a large number of cells (3 × 104 from each group), and all cells loaded into the device are individually accounted for. The 3-fold difference in the ACF between the fMLPactivated and untreated neutrophils is detected after 7 min of incubation, 2 min of exposure to the flow, and 1.5 min of the videomicroscopy scanning (2-min curves in Figure 4a). This minimal version of the assay, which provides quantitative results on the strength of adhesion of both activated and nonactivated cells and a clear differentiation between them, is finished in ∼11 min from the moment of loading of cells into the device without any intervention by the operator. Existing clinical tests of the inflammatory response usually rely on measurements of the concentrations of soluble signaling molecules, such as C-reactive protein (CRP) and to lesser extend interleukins 1 and 6 (IL1, IL6),34 which are usually performed with ELISA-type assays. The difference in concentrations between a normal and an acute inflammatory state can be very small. For example, for CRP, it corresponds to the levels 3 mg/L, respectively,35 whereas the dynamic range of physiological concentrations for CRP is from 0.01 to 250 mg/L. Therefore, a reliable test based on CRP may require multiple assays over a period of time. In the proposed neutrophil adhesion assay, the physiological response of neutrophils to an inflammatory cue (fMLP) is (34) Schmid-Schonbein, G. W. Annu. Rev. Biomed. Eng. 2006, 8, 93-131. (35) Danesh, J.; Wheeler, J. G.; Hirschfield, G. M.; Eda, S.; Eiriksdottir, G.; Rumley, A.; Lowe, G. D.; Pepys, M. B.; Gudnason, V. N. Engl. J. Med. 2004, 350, 1387-1397.

Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

2257

measured before their programmed respiratory burst, which is the root cause of tissue damage and the ensuing inflammation cycle. Thus, we speculate that the proposed assay can potentially be used for alternative diagnostic tests and allow earlier detection of gradual changes in the state of inflammation (especially lowgrade inflammation). CONCLUSIONS We presented quantitative measurements of the difference in the strength of adhesion to a substratum between activated and nonactivated human neutrophils. The difference between neutrophils activated with fMLP and untreated neutrophils reached a factor of 3 in terms of fraction of cells remaining adherent to substratum. Measurements were carried out using a novel microfluidic device with integrated membrane valves and an array of tapered test channels. Adhesion strengths of the two groups of cells were compared directly, and the dependence of the ACF on the magnitude and duration of the stress applied to the cells and on the incubation time was explored. Large samples of data on the dynamics of cell detachment under stress were collected using a computer-controlled scanning videomicroscopy system. The proposed device could be used to measure differences in the

2258

Analytical Chemistry, Vol. 79, No. 6, March 15, 2007

strength of adhesion of cells to different substrata and could be modified for concurrent testing of a larger number of different cell populations. The device could be applied for testing the physiological state of inflammation in a subject’s blood by measuring the strength of adhesion of neutrophils to a substratum. ACKNOWLEDGMENT This work was partially supported by an NSF NIRT grant. E.G. acknowledges the support by UCSD/SDSU Institutional Research and Academic Career Award (NIH GM 68524). The authors thank Geert Schmid-Schonbein, Amy Sung, Robijn Bruinsma, and Virginia VanDelinder for helpful discussions. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review September 8, 2006. Accepted January 10, 2007. AC061703N