Continuous-Flow Biomolecule and Cell Concentrator by Ion

We present a novel continuous-flow nanofluidic biomolecule/cell concentrator, utilizing the ion concentration polarization (ICP) phenomenon. The devic...
0 downloads 0 Views 4MB Size
ARTICLE pubs.acs.org/ac

Continuous-Flow Biomolecule and Cell Concentrator by Ion Concentration Polarization Rhokyun Kwak,† Sung Jae Kim,‡ and Jongyoon Han*,‡,§ †

Department of Mechanical Engineering, ‡Department of Electrical Engineering and Computer Science, and §Department of Biological Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States

bS Supporting Information ABSTRACT: We present a novel continuous-flow nanofluidic biomolecule/ cell concentrator, utilizing the ion concentration polarization (ICP) phenomenon. The device has one main microchannel which bifurcates into two channels, one for a narrow, concentrated stream and the other for a wider but target-free stream. A nanojunction [cation-selective material (Nafion)] is patterned along the tilted concentrated channel. Application of an electric field generates the ICP zone near the nanojunction so that biomolecules and cells are guided into the narrow, concentrated channel by hydrodynamic force. Once biomolecules from the main channel are continuously streamed out to the concentrated channel, one can achieve a continuous flow of the same sample solution but with higher concentrations up to 100-fold. By controlling hydrodynamic resistance of the main and concentrated channel, the concentration factors can be adjusted. We demonstrated the continuous-flow concentration with various targets, such as bacteria [fluorescein sodium salt, recombinant green fluorescence protein (rGFP), red blood cells (RBCs), and Escherichia coli (E. coli)]. Specially, fluorescein isothiocyanate (FITC)conjugated lectin from Lens culinaris (lentil) (FITClectin) was tested on the different buffer conditions to clarify the effect of polarities of the target sample. This system is ideally suited for a generic concentration front-end for a wide variety of biosensors, with minimal integration-related complications.

M

icrofluidic devices for molecular detection have been extensively pursued, due to many well-documented advantages of such systems: rapid analyses, less consumption of samples and reagents, and potential for massive parallelization and automation.1,2 However, efficient world-to-chip interfacing, sample preparation, and concentration of low-abundance analytes remain as challenges, especially for non-nucleotide targets. To achieve more sensitive detection for any given sensor platform, various sample preconcentration approaches have been developed, including isotachophoresis,35 electrokinetic trapping,6,7 membrane filtration,8,9 and ion concentration polarization.1013 These methods could potentially enhance the sensitivity of biological assays such as immunoassays14 and enzyme activity assays.15 So far, most existing biomolecule concentration devices operate by collecting or trapping low-abundance biomolecules within a small-volume sample plug. Although this mode of concentration is efficient in increasing the local concentration,10 it is often limited in the maximum flow rate (to maintain the local plug) and/or sample volume one can process, and the integration with downstream detection steps is commonly challenging. One could avoid these problems by carrying out the detection within the plug during the ongoing concentration process,1418 but different electric, fluidic, pH, or other conditions19 within (or near) the concentrated plug could render such in situ detection less desirable. r 2011 American Chemical Society

Microfluidic systems with continuous-flow operation are much better suited for such “generic” sample preparation interfaces, to overcome many difficulties in microfluidic integration. For this reason, various continuous-flow biomolecule separators2024 have been developed for sample preparation, which requires seamless integration with downstream biosensors. However, a continuous-flow concentrator for sample preconcentration has not yet been reported. Previously, the concept of a continuousflow microfluidic demixer was published, by utilizing induced charge electrokinetics in structured electrode arrays, but the concentration factor was limited (up to 2-fold).25 In this paper, we have developed a continuous-flow concentrator for both biomolecules and cells, based on the ion concentration polarization (ICP) initiated within a microfluidic channel coupled with an ion-selective membrane. ICP is a fundamental transport phenomenon that occurs near ion-selective membranes and is often called an ion-depletion and ion-enrichment process. 26 Any charged molecule can be accumulated at the ion-depletion boundary with an appropriate tangential force field, either by an electric field or hydrodynamic pressure. 27 This system has been previously employed for enhancing immunoassays14,28,29 Received: May 17, 2011 Accepted: August 21, 2011 Published: August 21, 2011 7348

dx.doi.org/10.1021/ac2012619 | Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry

Figure 1. Schematic diagram of the continuous-flow concentrator (lower) and optical image of the device (upper). On application of an electric field (V+) and pressure field (ΔP), the ICP zone was developed based on force balance between the depletion force and hydrodynamic force. In the schematic figure, each arrow indicates the direction of hydrodynamic force (white), depletion force (red), and net force (yellow) acting on charged ions or particles. Depletion force indicates force acting on charged species as a result of ICP (ion depletion), which is electrostatic in nature. In the concentrator, the ICP boundary was generated along the Nafion with a specific distance, and charged species were accumulated on this boundary. Electrical voltage was applied at the inlet of the main channel (0150 V) to the buffer channel (electrical ground) across the Nafion junction. Flow rate (05 μL/min) was set by a syringe pump. L1 and w1 indicate a length and a width of the concentrated flow channel, and L2 and w2 indicate a length and a width of the filtered flow channel. R1 and R2 indicate hydrodynamic resistances of concentrated flow and filtered flow, respectively. For experiments, the devices were fabricated with various width ratios w0 = (w1 + w2)/w1 (w0 = 10, 20, 50, 100) and length ratios l0 = L1/L2 (l0 = 0.3, 0.7, 2).

and enzyme activity assays. 15,30 Recently, we demonstrated that the ion-depletion boundary could be controlled by changing the applied electric field and hydrodynamic pressure field along the microchannel so that any charged species could flow along a specific portion of the microchannel in the longitudinal direction, 19 which was later used to desalinate/disinfect seawater in a microfluidic device. 31 Otherwise, cells could be concentrated by an inertia force-based microfluidic device, 32,33 but they are not suitable for a biomolecule concentrator because the size of the molecules is too small to expect an inertial effect. In this paper, we demonstrate a continuous concentration of sample analyte by controlling the position of the concentrated plug and depletion zone within a microchannel. The concentration factors up to 100-fold are achievable with biomolecules [fluorescein sodium salt, FITClectin (positively and negatively charged), rGFP] and cells [RBCs and Escherichia coli (E. coli)]. The sample flow rate is reasonably high (approximately a few microliters per minute) without significant limitation in downstream integration.

’ THEORY Ion concentration polarization, which is a fundamental electrochemical transport phenomenon, occurs when an ion current is passed through an ion-selective membrane.34 Caused by the

ARTICLE

mismatch of charged carriers between electrolyte and membrane, it results in significant, dynamic perturbation in ion concentrations near the membrane (a.k.a. ion depletion or enrichment26), along with the generation of strong electrokinetic fluid flow.3537 The ion-selective material we used is Nafion, which has strong cation selectivity due to many sulfonate groups decorating the hydrophobic, Teflon-like polymer backbone, forming a nanometer size ion-selective current pathway within the material.38 Once voltage is applied across this cation-selective membrane, ICP is triggered near the membrane. Typical ionic behavior is that anions (negative ions) on the anodic side are drawn to the anode and the membrane forms a barrier preventing anion transport from the cathodic side to the anodic side of the membrane. The net outcome, in order to satisfy electroneutrality in the bulk solution nearby, is that both anions and cations are depleted away from the anionic side of the membrane (iondepletion) and enriched on the cathodic side of membrane (ion-enrichment). Under a static dc voltage/current bias across the membrane, the ion-depletion zone typically expands continuously,35,39 eventually leading to a nonsteady concentration/ flow profile in the system. However, it was shown that, by combining additional forced flow (either caused by a tangential electric field or pressure gradient) with the ion-depletion zone formed by ICP, one can limit and control the boundary of the ion-depletion zone.19,31 At lower flow rates, one can achieve a situation where the depletion zone is “blocking” the entire microchannel, leading to accumulation of charged species at the boundary (stationary batch-type biomolecule concentrator). At higher flow rates, however, one can reach a condition where this boundary is partially “burst”, leading to a steady-state establishment of the depletion zone and flow-dominated zone coexisting.19 This process will essentially create two parallel streams (desalted and concentrated streams) within the microfluidic channel, which was later utilized for continuous seawater desalination.31 It came to our attention that, in the same system, the concentrated stream can also be viewed as a continuous-flow concentration system, which is the subject of this paper. To realize a continuous-flow concentrator, a Nafion (cationselective material) line was patterned with an angle to the main channel, as shown in Figure 1. Application of an electric field (V+) with Poiseuille flow (ΔP) would generate a controlled depletion zone near the Nafion pattern, which was used to block the stream of charged biomolecules and cells and create a concentrated plug in the sample fluid. These concentrated charged species were guided (by the external flow) into a narrower concentrated channel continuously, achieving a continuous flow of the same sample solution but with higher concentration. It is important to control the position of the concentrated plug and ion-depletion zone stably in this system. The schematic of the device is similar to the desalination microfluidic system reported previously,31 with two modifications. First, outlets of the concentrated channel and filtered channel were electrically floated. Although an electrical voltage has been applied to one inlet and two outlets previously,31 current flows through outlets were unnecessary in this work. Second, the Nafion was patterned in a straight line parallel to the concentrated channel. Then, we were able to obtain a straight ICP boundary along the Nafion pattern, which was generated instead of a curved ICP boundary. This facilitated us to predict the shape and the position of the ICP boundary and concentrated plug. With the ideal case assumption that no charged molecules can overcome the energy barrier of the depletion zone boundary, the 7349

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry

ARTICLE

concentration of charged molecules in the lower filtered channel would be zero. Actual electrolyte (i.e., majority carrier) concentration inside the depletion zone was nonzero, but minority carrier (biomolecules and cells in this study) was completely rejected by ICP so that the zero concentration assumption inside the depletion zone is valid.31 Then, the concentration factor of the concentrated flow can be expressed by evaluating the ratio of flow rates of two (concentrated flow and sample flow) channels. In the case of a rectangular channel with length L, width w, and height h, hydrodynamic resistant Rh and flow rate Q are as the following.40 "  !#1 12μLi h 192 ∞ 1 nπwi Rh, i ¼ 1 tanh wi π5 n ¼ 1, 3, 5 n5 wi h3 2h



ð1Þ Qi ¼

ΔP Rh, i

ð2Þ

Here, μ is the viscosity of the solution. Flow rate Q in a channel is proportional to the applied pressure drop ΔP, and for the index i represents different channels, i = 1 for the concentrated channel and 2 for the filtered channel. Equation 1 could be linearized when a channel aspect ratio is much larger than 1 as Rh,i = 12μLi/wih3. However, the additional viscous drag causes fluid retardation near the edges that becomes considerable with the aspect ratio around 1, yielding the nonlinear eq 1.41 By using eqs 1 and 2, the concentration factor r of the concentrated flow can be expressed as  1 Q1 R1 ¼1 þ ð3Þ r ¼ Q1 þ Q 2 R2 We assume there is no other factor to generate additional resistance such as surface roughness, diffusion potentials, turbulences, etc. In eqs 1 and 2, ΔP, μ, and h are of constant value. Therefore, the concentration factor can be adjusted by modifying the two channel’s length L and width w. In a uniform height of the system, flow rate is approximately proportional to the width of the channel and inversely proportional to the length of the channel.

’ EXPERIMENTAL METHODS Device Fabrication. The systems under the study were composed of two parallel microchannels connected by Nafion film (nanojunction), which has high ion selectivity due to the cluster network of sulfonate groups (Figure 1). We fabricated poly(dimethylsiloxane) (PDMS) microfluidic chips with this ion-selective material using previously published methods.42 Nafion was patterned with PDMS line mold (width, 100 μm; height, 50 μm) on a clean glass substrate. Nafion perfluorinated ion-exchange resin (Sigma-Aldrich, St. Louis, MO) with 20 wt % solution was used. The Nafion-patterned glass substrate was then bonded with the PDMS microchannel by means of plasma treatment. For most of the test specimens (for the preconcentration test of fluorescein sodium salt, rGFP, RBCs, and E. coli), the main channel and buffer channel had dimensions of 200, 400, 1000, 2000 μm for width, 15 μm for height, and 1 cm for inlet length (from inlet to bifurcated point). The width of the concentrated channel has been fixed at 20 μm so that the width ratios of the concentrated channel to filtered channel are 1:10,

1:20, 1:50, and 1:100. For the FITClectin test, in addition, the devices which had dimensions of 10 μm for the width of the concentrated channel, 15 μm for height also used to control the velocity of concentrated flow. The main channel had dimensions of 200 and 500 μm for width. The lengths of the concentrated channel and filtered channel were controlled by the location of the outlet reservoirs, ranging from 5 to 10 mm. PDMS molds for Nafion nanojunction patterning and the concentration microchannels were fabricated by standard photolithography.43 Device Operation. The hydrodynamic pressure was generated by a syringe pump (Harvard apparatus, PHD 2200). All the experiments were imaged using an inverted epifluorescence microscope (Olympus, IX-71) with a thermoelectrically cooled charge-coupled device (CCD) camera (Hamamatsu Co., Japan). Sequences of images were analyzed by Image Pro Plus 5.0 (Media Cybernetics). The dc voltage across the ion-selective membrane was applied with a Keithley 236 currentvoltage source measurement unit (Keithley Instruments, Inc.) and highvoltage dc power supplier (Stanford Research Systems, Inc., model P350) with Ag/AgCl wires (A-M Systems, Inc.). Operational voltage and flow rate values were chosen between 0 and 150 V and 0 and 5 μL/min, respectively, which were adjusted to balance the depletion force (by applied voltage) and hydrodynamic force (by flow rate). The channel and interface of Nafion and PDMS would break down if the applied voltage (or electric field) is too high. In addition, we observed stronger instability under higher voltage bias. Therefore, we set the maximum voltage as 150 V (150 V/cm) which is ∼1000 times less than the junction gap breakdown voltage of the PDMS substrate (∼25 V/μm).44 With this applied voltage level between 0 and 150 V, we optimized the experimental parameters that could provide the maximum flow rate (up to ∼5 μL/min flow) and the maximum concentration factor with various width ratios w0 = (w1 + w2)/w1 (w0 = 10, 20, 50, 100) and length ratios l0 = L1/L2 (l0 = 0.3, 0.7, 2). Materials. To demonstrate the continuous-flow concentrator, 3 pM fluorescein sodium salt (Sigma-Aldrich, St. Louis, MO) and 3 pM recombinant green fluorescence protein (rGFP, BD bioscience, Palo Alto, CA) were prepared in the main buffer solution, 1 mM phosphate (dibasic sodium phosphate) at pH = 8.7. We also prepared three different concentrations of phosphate, 1, 10, and 100 mM, to measure the concentration factors with various ionic strength conditions. An amount of 5 μg/mL fluorescein isothiocyanate (FITC)conjugated lectin from Lens culinaris (lentil) (FITClectin) (Sigma-Aldrich, St. Louis, MO) was used to show the effect of different surface charge polarities of the target sample. Under 1 mM phosphate (dibasic sodium phosphate) at pH = 8.7 and 1 mM phosphate at pH = 7.4, FITClectin [MW ∼ 49 kDa, isoelectric point pI (characteristic pH value at which proteins exhibit zero net charge) ∼8.08.8]23,45 has different surface charge polarities. Average ζ-potentials (electric motilities) of FITClectin were 18.8 mV (1.464  108 m2/V 3 s) at pH 8.7 and 3.35 mV (0.263  108 m2/V 3 s) at pH 7.4, respectively (measured by a Zetasizer nano ZS, Malvern). In experiments, 0.1% bovine serum albumin (BSA) (Sigma-Aldrich, St. Louis, MO) was added into 1 mM phosphate at pH 7.4 for preventing nonspecific binding. Human whole blood with red blood cells (RBCs) (Hoechst, 2000 dilution, Innovative Research) and E. coli ER2738 (New England BioLabs Inc.) were used to show the potential of the application for various biosamples. Human whole blood was diluted 200500 times again using 1 phosphate-buffered saline 7350

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry

ARTICLE

Figure 2. A 62-fold continuous-flow concentrator (w1 = 20 μm, w2 = 980 μm, L1 = 0.6 mm, L2 = 0.9 mm, and rtheoretical = 62) with a 30° inclined angle of the Nafion junction. Flow rate and applied voltage were 5 μL/min and 80 V, respectively. (a) The comparison images of the concentrator when voltage was applied (lower) and when it was released (upper) and (b) the fluorescence intensity curve. Generation of concentrated flow (blue line, B), concentrated plug (the peak of the red line, C), filtered flow (decreased florescence intensity of red line C than inlet sample flow), and sample flow (green line, A) are well-defined. Fluorescence intensity of the concentrated flow jumped from 300 to 800 (corresponding 3 and 150 pM, respectively), which indicates 50-fold jumping of the concentration of fluorescein sodium salt. The wide peak from 200 to 300 pixels was due to the autofluorescence of the patterned Nafion. (c) Integrated image of the concentrator from the bifurcated point to the outlet of the channel. Yellow dotted lines indicate the geometry of the channel; 1 mM dibasic phosphate buffer solution (pH 8.7) and 3 pM fluorescein sodium salt for tracking ions were used.

(PBS, ∼150 mM) solution. Pluronic F108 (BASF, Cumberland, RI) at 0.2% was added to reduce osmosis hemolysis, nonspecific binding, and aggregation. White blood cells (WBCs) and other blood cells were not removed from the whole blood, because the density of WBCs is much lower than the density of RBCs, and their presence did not affect the concentration behavior. Cell Tracker Orange CMRA (Molecular Probes, Eugene, OR) at 3 μM was added for tracking RBCs in the Supporting Information video. A single colony of E. coli ER2738 was cultured in LBTet medium (LB medium 50 mL plus tetracycline (Tet) 50 μL) at 37 °C in 24 h. LBTet medium contains 10 g/L NaCl (∼171 mM). A mixture of RBCs and E. coli ER2738 was prepared by mixing RBCs in 1 PBS and E. coli ER2738 in LBTet medium with 1:10 volume ratio (∼168 mM). Average ζ-potentials of red blood cells and E. coli were 13.3 and 15.5 mV under the experimental conditions (measured by a Zetasizer nano ZS, Malvern). Since PDMS and both microorganisms have the same polarity, they are electrostatically not favorable to bind together. Nonspecific binding of E. coli cells to PDMS surfaces was not

severe to interfere with the experiment, even after the preconcentration step in this device. Quantification of Molecular Concentration. A highly concentrated plug of fluorescent molecules often saturated the CCD array, which will lead to errors in quantification. The exposure time and contrast were controlled to map concentrated peak intensities between 10 and 100 nM fluorescein sodium salt and rGFP. Then, we measured the fluorescence intensities of the samples with known concentrations (0.03, 0.3, 3, and 30 pM and 0.3, 3, and 30 nM) and used them as internal references for calculating the concentration of fluorescein sodium salt and rGFP. Similarly, the fluorescence intensities of FITClectin with 0.5, 5, 50, 100, and 500 μg/mL were measured and used as internal references. For evaluating the concentration of RBCs and E. coli, we used two methods. First, the density of the cells of sample flow and filtered flow were measured by using a disposable hemocytometer (SKC Inc., Covington, GA). The hemocytometer is a device designed for counting cells or particles. However, we could not use this device to measure the cell density of the 7351

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry

Figure 3. Fluorescein sodium salt, rGFP, RBCs, and E. coli test results with various width ratios w0 = (w1 + w2)/w1 (w0 = 10, 20, 50, 100), in log scale, and length ratios l0 = L1/L2 (l0 = 0.3, 0.7, 2) with a 30° inclined angle of the concentrated channel and Nafion. Theoretical concentration ratio was calculated by eq 3, and the real concentration ratio was measured by comparing fluorescence intensity of target samples. Flow rate was fixed as 2 μL/min, and applied voltage was adjusted to generate the stable continuous concentrated flow between 0 and 150 V. The change of instability condition was also tested with 1, 10, and 100 mM dibasic phosphate buffer solutions for fluorescein sodium salt.

concentrated flow due to low output flow rate (in the case of Figure 5, flow rate of the concentrated flow was about 0.05 μL/min). Then, we were not able to obtain enough volume (∼10 μL) to use the hemocytometer in the short time period (it takes 200 min to collect 10 at 0.05 μL/min). Therefore, second, image analysis was done to measure the density of cells in the concentrated flow. With phase contrast images (Figures 5 and 6), cells were distinguished from background by the difference of their brightness so that one can count the number of cells within a defined volume by the ImageJ program.

’ RESULTS AND DISCUSSION Continuous Preconcentration Demonstration. Figure 2 shows a representative result of the continuous-flow concentrator. The device has the width ratio w0 = 50 and the length ratio l0 = 0.7 (w1 = 20 μm, w2 = 980 μm, L1 = 6 mm, L2 = 9 mm). With this geometry, the concentration along the concentrated flow (blue line) would be enhanced by 50-fold in nonleaking operation (the theoretical concentration factor given by eqs 13 was 62). With the application of 80 V of applied voltage and 5 μL/min flow rate, the concentrated plug was generated on the front of the slanted line-patterned Nafion, and concentrated sample flew along the concentrated channel successfully. Although fluorescence intensity of flow in the concentrated channel was the same with the sample flow (∼300 AU) with no applied voltage, the intensity increased by 50-fold compared to the sample flow with 80 V of applied voltage. In contrast, fluorescence intensity of the filtered flow in the main channel was dropped (∼260 AU) right after the Nafion membrane (Figure 2, parts a and b). This fluorescence intensity level is below the noise level compared to background signal, which means the concentration of fluorescein sodium salt was very low, if any. As shown in Figure 2c, concentrated flow, filtered flow, and the concentrated plug and the boundary of the

ARTICLE

ion-depletion zone were stably maintained across a 1 mm wide channel. The device was operated for more than 30 min without compromising the ICP boundary. We could observe the concentrated plug and measure its peak in the fluorescence intensity (Figure 2b, from 50 to 150 pixels), indicating 200-fold concentration enhancement. Although this value is lower than the concentration factor of the previously reported batch-type preconcentrator,10 steady-state concentrated flow of target molecules achieved in this system would be much more amenable to downstream sensor integration. Even though the batch-type concentrator can achieve very high local concentration within the plug during the preconcentration process, it is difficult to maintain such a highly concentrated sample plug as soon as the preconcentration process is stopped or to deliver toward downstream integration, due to diffusive dispersion. This system does not suffer from this issue, and the concentration factor achieved could be maintained in the downstream channel or reservoir. To characterize device performance and limitation, various continuous-flow concentrators (with different concentration factors) were fabricated and tested (Figure 3) under different buffer ionic strengths (1, 10, 100 mM dibasic phosphate buffer solution) in comparison with theoretical concentration factors from eq 3. The basic conditions for experiments was 3 pM fluorescein sodium salt, the width of the concentrated channel w1 = 20 μm, the height of the device h = 15 μm, and the sample flow rate 2 μL/min. Under these experimental parameters, the concentration factors for the device were designed with various width ratios w0 = (w1 + w2)/w1 (w0 = 10, 20, 50, 100) and length ratios l0 = L1/L2 (l0 = 0.3, 0.7, 2). At the main channel, overall input flow rate was regulated by the syringe pump, which was divided into the concentrated and filtered channel. Each flow rate at the filtered and concentrated channels was adjusted by their ratio of the widths (w0 ) and ratio of the lengths (l0 ), introducing different concentration factors. As shown in Figure 3, experimental concentration factors were proportional to w0 and l0 as we expected. However, these observed concentration factors were always lower than theoretical concentration factors given by eq 3. rGFP, E. coli, and RBCs concentration were also tested with w0 = (w1 + w2)/w1 (w0 = 10, 20, 30, 50) and length ratios l0 = L1/L2 (l0 = 0.7), resulting in a similar tendency. These differences could presumably be caused by the two following reasons: (1) the loss of charged species, which either leak thorough the ion-depletion zone or nonspecifically bind to the wall; (2) the change of flow rate ratio of the filtered/ concentrated channel by ICP. The former should be a minor factor, since the measured leakage rate of molecules and cells under microscopic observation was negligible in our experiment. Also, no significant nonspecific binding on the wall and Nafion was observed (Figure 2). Therefore, leakage and nonspecific binding alone could not explain the large discrepancies in the experiments using high width and length ratios (e.g., l0 = 2 and w0 = 50). Instead, additional hydrodynamic resistance by ICP may explain the large discrepancies. According to the experimental observation of a polystyrene (PS) bead test and the previous experiment,19 the fluid velocity of concentrated flow was boosted adjacent to the ion-depletion zone (see Supporting Information Figure 1). In addition, the fluid velocity of filtered flow was slightly reduced. This change of the fluid velocities was able to decrease the flow rate ratio (from total flow rate divided by concentrated flow rate), resulting a drop in the concentration factor. The adjustment of fluid velocities of the two channels is 7352

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry

Figure 4. FITClectin concentration test results with three different theoretical concentration factors/concentrated flow velocities: 12-fold (w0 = 10, l0 = 0.7, w1 = 20 μm), 46-fold (w0 = 20, l0 = 0.7, w1 = 10 μm), and 130-fold (w0 = 50, l0 = 0.7, w1 = 10 μm). Inset images show the fluorescence image of the operation of the concentrator with (i) 130-fold at pH 8.5, (ii) 130-fold at pH 7.4, (iii) 12-fold at pH 8.5, and (iv) 12-fold at pH 7.4. The scale bar indicates 500 μm for all inset images. The higher designed concentration factor gave larger loss of the concentration factor due to the underestimation of concentrated flow rate enhancement by second electroosmotic components. In addition, the additional loss of concentration factor in positively charged FITClectin (pH 7.4) compared to negatively charged FITClectin (pH 8.5) also became larger because lower concentrated flow velocity exacerbated nonspecific, electrostatic binding.

due to fast fluid vortices near the ion-selective membrane,35,37 which were routinely observed in this system. Counter-rotating vortex pairs can have asymmetric sizes (smaller vortex and larger vortex existed in series) with an external pressure field, whereas they have the same size of rotation with symmetric conditions (Figure 3 in ref 35). In such case, the larger vortices would be dominant and can induce higher flow speed in the concentrated stream (see Supporting Information Figure 1). It is noted that the (apparent) loss of concentration factor in the experiment is not a real loss of target molecules. Rather, it is due to the underestimation of flow rate in the concentrated stream, caused by ignoring additional convection which is the essential part of the second electroosmotic components.35,46 Accurate numerical modeling of ICP zone flow is still not yet realized, which would be needed for quantifying the concentration factor exactly. A critical limitation to achieve a higher concentration factor in this system was the destabilization of the ICP boundary (gray regime in Figure 3). The magnitude of fluctuation increased with an electric field and a concentration gradient across the concentrated plug (see Supporting Information Figure 2). Small and steady fluctuation of the ICP boundary was acceptable to the operation of this system because it did not disrupt the continuous concentration process. However, for a higher concentration factor, the ICP boundary fluctuation worsened and the system could not maintain continuous concentrated flow anymore. The fluctuations tend to occur more frequently when ionic strength of the buffer solution becomes higher. It is because a higher electric field is needed for generating the ICP boundary with high molar concentrations. As a result, the instability regime expanded in higher ionic strength conditions, designated as the light-gray regime in Figure 3.

ARTICLE

Figure 5. A 20-fold continuous-flow concentrator for RBCs (w0 = 10, l0 = 1.2). (a) The operating images of the concentrator with voltage (90 V) and flow rate (5 μL/min). Comparison images of (b) the outlet of concentrated flow and (c) the outlet of filtered flow. The density of RBCs increased from 1.9  106 to 3.7  107 cells/mL, indicating 19fold. The red dotted line indicates the patterned Nafion.

Positively Charged Analyte. Next, continuous-flow concentrations of target samples with different polarities were tested. FITC-conjugated lectin from L. culinaris was used. FITClectin (pI ∼ 8.08.8) has a negatively charged surface under pH 8.5 and a positively charged surface under pH 7.4. Figure 4 shows FITClectin concentration test results with three different concentration factors: 12-fold (w0 = 10, l0 = 0.7, w1 = 20 μm), 46-fold (w0 = 20, l0 = 0.7, w1 = 10 μm), and 130-fold (w0 = 50, l0 = 0.7, w1 = 10 μm). Other conditions were fixed as 15 μm height, 1 μL/min sample flow rate, and 1 mM phosphate buffer. Applied voltage was adjusted from 0 to 150 V. Concentrated flow velocity vc was newly defined as the theoretically expected fluid velocity of concentrated flow under the intention for showing the effect of nonspecific binding (slow velocity gives a higher chance for nonspecific binding to occur). It was calculated by following equation.

vc ¼

Q1 Q1 þ Q2 ¼ w1 h1 rtheoretical w1 h1

ð4Þ

As shown in Figure 4, experimental concentration factors for FITClectin were also always lower than the theoretical concentration factor due to additional fluidic resistance by ICP. In addition, the measured concentration factors of positively charged FITClectin (pH 7.4) were lower than those of negatively charged FITClectin (pH 8.5). This additional drop of the concentration factor was due to nonspecific electrostatic binding to the wall and Nafion. Although 0.1% BSA was added to prevent nonspecific binding, the electrostatic binding would be significant at low flow rate because positively charged FITClectin had more chance to detect and stuck to the negatively charged wall and Nafion. Consequently, there was no obvious significant drop with relatively fast flow (12-fold case), but large discrepancies were presented at slow flow (130-fold case) (Figure 4). Cell Preconcentration Demonstration. For the demonstration of cell concentration, RBCs were tested with a 20-fold concentrator (w0 = 10, l0 = 1.2) with high flow rate (5 μL/min) and 1 PBS (∼150 mM) solution with 0.2% Pluronic F108 as shown in Figure 5. Comparison images of two outlets show the contrast of the densities of RBCs between the outlet of the concentrated 7353

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry

ARTICLE

Figure 6. A 20-fold continuous-flow concentrator for RBCs and E. coli (w0 = 10, l0 = 1.2). (a) The operating images of the concentrator with voltage (14 V) and flow rate (0.1 μL/min) and magnified image of the sample flow. Comparison images and magnified images of (b) the outlet of concentrated flow and (c) the outlet of filtered flow. The density of RBCs increased from 7.3  105 to 1.4  107 cells/mL, and the density of E. coli jumped from 3.2  107 to 6.5  108 cells/mL, indicating 19- and 20-fold, respectively. The red dotted line indicates the patterned Nafion.

flow (Figure 5b) and the outlet of the filtered flow (Figure 5c) obviously (see the Supporting Information video). The density of RBCs was jumped from 1.9  106 cells/mL in the sample flow to 3.7  107 cells/mL in the concentrated flow, indicating 19-fold increments. This value was almost the same with the theoretical concentration factor, 20-fold. Representing motions of blood cells were blocked and collected in the concentrated plug near the Nafion, climbed the Nafion to the bifurcated point, and flew through the concentrated channel. Concentration of the mixture of various biomolecules (RBCs and E. coli) was also tested with the mixed buffer solution of 1 PBS and LBTet medium (171 mM). The continuous concentrated flow of RBCs and E. coli was developed successfully as shown in Figure 6. In the comparison, images of two outlets show the contrast of the densities of RBCs and E. coli between the outlet of the concentrated flow (Figure 6b) and the outlet of the filtered flow (Figure 6c) obviously. The density of RBCs was jumped from 7.3  105 to 1.4  107 cells/mL, and the density of E. coli was jumped from 3.2  107 to 6.5  108 cells/mL, indicating 19- and 20-fold, respectively. This value was also almost the same with the concentration factor, 20-fold. With the above two tests, leakage rate to the filtered flow was also calculated by live imaging of the concentration operation, resulting in 0% during 120 s of operation. In addition, RBCs/E. coli concentration tests were performed with a 12- and 24-fold concentrator (w0 = 10, 20 and l0 = 0.7) (Figure 3). The concentration factors for RBCs and E. coli were almost the same as the factors for fluorescein sodium salt and rGFP. That is, when the geometry of the device is fixed, we can obtain the same concentration factor for any target. It is because the concentration factor is only governed by the flow rate ratio of the concentrated channel and filtered channel. Such a uniform concentration over many different targets would be a useful feature as a generic preconcentration interface. As the first continuous-flow concentration system, applied to both biomolecules and cells, this work has several significant contributions, although the maximum concentration factor is

lower than that of the previously reported batch-type preconcentrator (∼106-fold).10,11,15,16,18 The concentration factor reported here was uniform and controllable (by adjusting channel geometry) for various targets, including dye, protein (positively and negatively charged species), deformable cells (RBCs), and bacteria, because of the new continuous ICP scheme. Therefore, the continuous-flow preconcentrator can inherently eliminate unfavorable characteristics of a batch-type concentrator [low throughput, unstable system parameters (concentration and electric field, etc.) as a function of time] by isolating concentrated streams.

’ CONCLUSION Herein, we have demonstrated the continuous-flow concentrator based on the ICP phenomenon, allowing one to process biosamples at reasonable sample flow rate (approximately a few microliters per minute) successfully. Various geometries of the device, the width ratio w0 , and the length ratio l0 , were tested to control the concentration factor from 1 to 200 with diverse samples at different buffer electrolyte concentrations and pH’s (giving opposite surface charge polarities). A continuous-flow biomolecule and cell concentration device would be much more flexible for integration with various types of downstream devices, such as a continuous-flow separator, mass spectrometry of target samples. The quantity of the concentrated sample volume is also controllable from a few picoliters to microliters. In addition, the nanofluidic preconcentration system was successfully working for both biomolecules and cells with continuous-flow manner. We tested versatile samples including dye, protein, deformable cells (RBCs), and bacteria because this continuous-flow concentration system utilizes the novel ICP phenomena which can filter out any charged species regardless of the samples’ mechanical or electrical properties. Instead of specific demonstration of the preconcentration of rare samples (e.g., circulating tumor cells), we showed general applicability of our system by testing the preconcentration of samples with different polarities (positively/negatively charged species) and various sizes from dyes and proteins to cells. Combination of this 7354

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355

Analytical Chemistry wide range of working conditions, high sample throughput, and integration flexibility into other microfluidic devices would make this system useful as a front-end concentration interface for many biosensors and detection systems.

’ 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

*E-mail: [email protected]. Phone: +1-617-253-2290. Fax: +1-617258-5846.

’ ACKNOWLEDGMENT This work was mainly supported by the National Science Foundation (CBET-0854026), Defense Advanced Research Projects Agency (DARPA) under contract no. HR 0011-10-1-0075 through the DARPA CIPHER project (S. Rodgers), Singapore MIT-Aliance II CE programme, and an Innovation Grant from SMART Innovation Centre. The MIT Microsystems Technology Laboratories are acknowledged for support in fabrication. We greatly appreciate Professor Angela Belcher and Heechul Park for the support of the E. coli sample preparation and Professors Cullen R. Buie and Y. S. Joung for the support of the ζ-potentiometer (Zetasizer nano ZS, Malvern). ’ REFERENCES (1) Bradley, T. E.; Arthur, W. W.; Robert, W. S.; Farshid, G. J. Cell. Physiol. 2006, 209, 987–995. (2) Engler, A. J.; Sen, S.; Sweeney, H. L.; Discher, D. E. Cell 2006, 126, 677–689. (3) Jung, B.; Bharadwaj, R.; Santiago, J. G. Anal. Chem. 2006, 78, 2319–2327. (4) Janasek, D.; Schilling, M.; Franzke, J.; Manz, A. Anal. Chem. 2006, 78, 3815–3819. (5) Gebauer, P.; Bocek, P. Electrophoresis 2002, 23, 3858–3864. (6) Singh, A. K.; Throckmorton, D. J.; Kirby, B. J.; Thompson, A. P. In Micro Total Analysis Systems; Kluwer Academic: Nara, Japan, 2002; Vol. 1, pp 347349. (7) Foote, R. S.; Khandurina, J.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 2005, 77, 57–63. (8) Song, S.; Singh, A. K.; Kirby, B. J. Anal. Chem. 2004, 76, 4589–4592. (9) Khandurina, J.; Jacobson, S. C.; Waters, L. C.; Foote, R. S.; Ramsey, J. M. Anal. Chem. 1999, 71, 1815–1819. (10) Wang, Y. C.; Stevens, A. L.; Han, J. Y. Anal. Chem. 2005, 77, 4293–4299. (11) Kim, S. J.; Song, Y. A.; Han, J. Chem. Soc. Rev. 2010, 39, 912–922. (12) Sueyoshi, K.; Kitagawa, F.; Otsuka, K. J. Sep. Sci. 2008, 31, 2650–2666. (13) Skelley, A. M.; Kirak, O.; Suh, H.; Jaenisch, R.; Voldman, J. Nat. Methods 2009, 6, 147–152. (14) Wang, Y. C.; Han, J. Y. Lab Chip 2008, 8, 392–394. (15) Lee, J. H.; Cosgrove, B. D.; Lauffenburger, D. A.; Han, J. J. Am. Chem. Soc. 2009, 131, 10340–10341. (16) Cheow, L. F.; Ko, S. H.; Kim, S. J.; Kang, K. H.; Han, J. Anal. Chem. 2010, 82, 3383–3388.

ARTICLE

(17) Chen, C.-H.; Sarkar, A.; Song, Y.-A.; Miller, M. A.; Kim, S. J.; Griffith, L. G.; Lauffenburger, D. A.; Han, J. J. Am. Chem. Soc. 2011, 133, 10368–10371. (18) Ko, S. H.; Kim, S. J.; Cheow, L. F.; Li, L. D.; Kang, K. H.; Han, J. Lab Chip 2011, 11, 1351–1358. (19) Kim, S. J.; Li, L. D.; Han, J. Langmuir 2009, 25, 7759–7765. (20) Chou, C.-F.; Austin, R. H.; Bakajin, O.; Tegenfeldt, J. O.; Castelino, J. A.; Chan, S. S.; Cox, E. C.; Craighead, H. G.; Darnton, N.; Duke, T. A. J.; Han, J.; Turner, S. Electrophoresis 2000, 21, 81–90. (21) Han, J.; Fu, J.; Schoch, R. B. Lab Chip 2008, 8, 23–33. (22) Zeng, Y.; Harrison, D. J. Anal. Chem. 2007, 79, 2289–2295. (23) Fu, J.; Schoch, R. R.; Stevens, A. L.; Tannenbaum, S. R.; Han, J. Nat. Nanotechnol. 2006, 2, 121–128. (24) Dimitriadis, E. K.; Horkay, F.; Maresca, J.; Kachar, B.; Chadwick, R. S. Biophys. J. 2002, 82, 2798–2810. (25) Leinweber, F. C.; Eijkel, J. C. T.; Bower, J. G.; van den Berg, A. Anal. Chem. 2006, 78, 1425–1434. (26) Pu, Q.; Yun, J.; Temkin, H.; Liu, S. Nano Lett. 2004, 4, 1099–1103. (27) Xiao, Y.; Yu, X.-D.; Xu, J.-J.; Chen, H.-Y. Electrophoresis 2007, 28, 3302–3307. (28) Liu, V.; Song, Y. A.; Han, J. Lab Chip 2010, 10, 1485–1490. (29) Lee, J. H.; Han, J. Microfluid. Nanofluid. 2010, 9, 973–979. (30) Lee, J. H.; Song, Y.-A.; Tannenbaum, S. R.; Han, J. Anal. Chem. 2008, 80, 3198–3204. (31) Kim, S. J.; Ko, S. H.; Kang, K. H.; Han, J. Nat. Nanotechnol. 2010, 5, 297–301. (32) Hou, H. W.; Bhagat, A. A. S.; Chong, A. G. L.; Mao, P.; Tan, K. S. W.; Han, J. Y.; Lim, C. T. Lab Chip 2010, 10, 2605–2613. (33) Bhagat, A. A. S.; Hou, H. W.; Li, L. D.; Lim, C. T.; Han, J. Y. Lab Chip 2011, 11, 1870–1878. (34) Probstein, R. F. Physicochemical Hydrodynamics: An Introduction; John Wiley & Sons, Inc., Hoboken, NJ, USA, 1994. (35) Kim, S. J.; Wang, Y.-C.; Lee, J. H.; Jang, H.; Han, J. Phys. Rev. Lett. 2007, 99, 044501. (36) Eddington, D. T.; Puccinelli, J. P.; Beebe, D. J. Sens. Actuators, B 2006, B114, 170–172. (37) Rubinstein, S. M.; Manukyan, G.; Staicu, A.; Rubinstein, I.; Zaltzman, B.; Lammertink, R. G.; Mugele, F.; Wessling, M. Phys. Rev. Lett. 2008, 101, 236101. (38) Gao, L. T.; Seliskar, C. J. Chem. Mater. 1998, 10, 2481–2489. (39) Zangle, T. A.; Mani, A.; Santiago, J. G. Anal. Chem. 2010, 82, 3114–3117. (40) Beebe, D. J.; Mensing, G. A.; Walker, G. M. Annu. Rev. Biomed. Eng. 2002, 4, 261–286. (41) Giddings, J. C.; Schure, M. R. Chem. Eng. Sci. 1987, 42, 1471–1479. (42) Lee, J. H.; Song, Y. A.; Han, J. Y. Lab Chip 2008, 8, 596–601. (43) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974–4984. (44) Lee, J. H.; Chung, S.; Kim, S. J.; Han, J. Y. Anal. Chem. 2007, 79, 6868–6873. (45) Schoch, R. B.; Bertsch, A.; Renaud, P. Nano Lett. 2006, 6, 543–547. (46) Rubinstein, I.; Zaltzman, B. Phys. Rev. E 2000, 62, 2238–2251.

7355

dx.doi.org/10.1021/ac2012619 |Anal. Chem. 2011, 83, 7348–7355