Nanofluidic Concentration of Selectively Extracted Biomolecule

tubules but allows ionic current and flow to pass through it. This device makes it possible to selectively extract target molecules such as streptavid...
0 downloads 0 Views 2MB Size
Download PDF
Anal. Chem. 2008, 80, 5383–5390

Nanofluidic Concentration of Selectively Extracted Biomolecule Analytes by Microtubules Taesung Kim*,†,§ and Edgar Meyho ¨ fer†,‡ Department of Mechanical Engineering, and Department of Biomedical Engineering, The University of Michigan, 2350 Hayward Street, Ann Arbor, Michigan 48109 We present a novel device for the selective extraction and high concentration of biomolecule analytes by integrating microtubules, one of cytoskeletal filaments, with nanofluidic technologies. Microtubules can be functionalized to provide numerous, nanoscopic binding sites for specific target biomolecules. The functionalized microtubules, with the target biomolecules bound to their surface, can be transported in the opposite direction to nontarget molecules using electrokinetic separation. Subsequently, the target molecule-bound microtubules are concentrated by a flat nanochannel structure, which filters the microtubules but allows ionic current and flow to pass through it. This device makes it possible to selectively extract target molecules such as streptavidin and bovine serum albumin and then highly concentrate them up to higher than 5 orders of magnitude from a complex mixture of analytes ranging from 1 nM to 10 fM. In addition, the device performs both extraction (separation) and concentration process simultaneously, which are typically performed in order in other devices, so that we significantly reduce analysis time and labor and even enable preconcentrated, identified target molecules to be available for postanalysis. Thus, we believe that the use of functionalized microtubules with nanofluidics will be a useful means to facilitate biochemical analysis systems. For the advancement of proteomic research and early disease diagnosis, many attempts have been made to develop a technique that enables sensitive detection, selective separation, and high concentration of proteins and biomarkers within a short period of time.1–4 Due to the advantages of miniaturized systems over traditional instrumentation,5–8 a wide variety of micro/nanofluidic biomolecule detection and preconcentration techniques have been * Corresponding author. E-mail: [email protected]. Phone: 734-846-8993. † Department of Mechanical Engineering. ‡ Department of Biomedical Engineering. § Present address: Department of Electrical Engineering and Computer Science, University of California at Berkeley, Berkeley, CA 94720-1770. (1) Nam, J. M.; Thaxton, C. S.; Mirkin, C. A. Science 2003, 301, 1884–1886. (2) Wang, Y. C.; Stevens, A. L.; Han, J. Y. Anal. Chem. 2005, 77, 4293–4299. (3) Wu, G. H.; Datar, R. H.; Hansen, K. M.; Thundat, T.; Cote, R. J.; Majumdar, A. Nat. Biotechnol. 2001, 19, 856–860. (4) Zheng, G. F.; Patolsky, F.; Cui, Y.; Wang, W. U.; Lieber, C. M. Nat. Biotechnol. 2005, 23, 1294–1301. (5) Auroux, P. A.; Iossifidis, D.; Reyes, D. R.; Manz, A. Anal. Chem. 2002, 74, 2637–2652. (6) Burns, M. A.; Johnson, B. N.; Brahmasandra, S. N.; Handique, K.; Webster, J. R.; Krishnan, M.; Sammarco, T. S.; Man, P. M.; Jones, D.; Heldsinger, D.; Mastrangelo, C. H.; Burke, D. T. Science 1998, 282, 484–487. 10.1021/ac8003874 CCC: $40.75  2008 American Chemical Society Published on Web 06/03/2008

developed by using nanocrystals as probes,9,10 the deflection of microcantilevers,3 the conductance change of nanowires,4,11 nanoparticle-based bar code technology,1,12,13 functionalized carbon nanotubes,14,15 electrokinetic trapping with silica membranes,16,17 and nanochannels.2 In particular, nanoscale biosensor techniques have shown extremely high sensitivity in detection (not in preconcentration), but some of them need additional, expensive instrumentation to amplify electrical or mechanical signals from the detected biomolecules.1,4,11,14,15 In addition, nanofluidic preconcentration techniques have increased the concentration of biomolecule analytes more than 106-fold from a few tens of nanomoles per liter of analytes. But most of these preconcentration techniques lack the capability to selectively separate and concentrate specific target molecules from a complex biomolecule analyte; traditional electrophoretic separation methods in a microchip can identify/separate specific target molecules from the concentrated analytes, but they require additional separation columns, and the concentrated biomolecules are unavailable for other postanalysis.17–19 To address the drawbacks of the nanoscale biosensors and the nanofluidic preconcentration techniques above, we employ microtubules (one of the major constituents of the cytoskeleton of cells) as a specific biomolecule carrier/detector and a nanofluidic structure as a biomolecule concentrator. This combination enables both sensitive detection and high concentration of specific (7) Laurell, T.; Marko-Varga, G. Proteomics 2002, 2, 345–351. (8) Lion, N.; Rohner, T. C.; Dayon, L.; Arnaud, I. L.; Damoc, E.; Youhnovski, N.; Wu, Z. Y.; Roussel, C.; Josserand, J.; Jensen, H.; Rossier, J. S.; Przybylski, M.; Girault, H. H. Electrophoresis 2003, 24, 3533–3562. (9) Alivisatos, P. Nat. Biotechnol. 2004, 22, 47–52. (10) Bruchez, M.; Moronne, M.; Gin, P.; Weiss, S.; Alivisatos, A. P. Science 1998, 281, 2013–2016. (11) Patolsky, F.; Zheng, G. F.; Lieber, C. M. Anal. Chem. 2006, 78, 4260– 4269. (12) Bao, Y. P.; Wei, T. F.; Lefebvre, P. A.; An, H.; He, L. X.; Kunkel, G. T.; Muller, U. R. Anal. Chem. 2006, 78, 2055–2059. (13) Nam, J. M.; Stoeva, S. I.; Mirkin, C. A. J. Am. Chem. Soc. 2004, 126, 5932– 5933. (14) Chen, R. J.; Bangsaruntip, S.; Drouvalakis, K. A.; Kam, N. W. S.; Shim, M.; Li, Y. M.; Kim, W.; Utz, P. J.; Dai, H. J. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 4984–4989. (15) Wong, S. S.; Joselevich, E.; Woolley, A. T.; Cheung, C. L.; Lieber, C. M. Nature 1998, 394, 52–55. (16) Khandurina, J.; Jacobson, S. C.; Waters, L. C.; Foote, R. S.; Ramsey, J. M. Anal. Chem. 1999, 71, 1815–1819. (17) Foote, R. S.; Khandurina, J.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 2005, 77, 57–63. (18) Sheehan, P. E.; Whitman, L. J. Nano Lett. 2005, 5, 803–807. (19) Kim, S. M.; Burns, M. A.; Hasselbrink, E. F. Anal. Chem. 2006, 78, 4779– 4785.

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5383

Figure 1. Basic mechanisms of our device. (A) A mixture of analytes containing specific target molecules, nontarget molecules, and functionalized microtubules is loaded into the microchannel at the left side of the nanochannel, whereas the right channel is filled with a buffer solution. At this step functionalized microtubules bind specific target molecules. (B) Since the electrophoresis (EP) of nontarget molecules is governed by the electroosmosis (EO) of a buffer solution in a glass substrate, in the presence of an electric potential they are separated from the target moleculebound microtubules towards the cathode. On the other hand, since the EP of microtubules dominates EO, the microtubules are transported towards the anode, even though target molecules are bound to their surface. Subsequently, the target molecule-bound microtubules are concentrated by the nanochannel structure, enabling the concomitant concentration of target molecules. (C) The micro/nanochannel network of the device. (D) The detailed top view of the device indicates a nanochannel where microtubules are concentrated. Here, Hµ is ∼7 µm but hn is ∼30 nm.

target biomolecules from a complex mixture without an additional separation process. Microtubules hold significant potential for biosensors and preconcentration techniques; first of all microtubules that are 24 nm in diameter but of varying length up to several micrometers long can provide ∼1600 binding sites per micrometer for specific target biomolecules, showing a large value of surface to volume ratio (Figure 1A). In addition, microtubules are biocompatible and their molecular structures are well-established so that they can be easily functionalized to capture specific target biomolecules. That is, the basic building blocks of microtubules are R- and β-heterodimers (∼8 nm tubulin subunits), which associate head5384

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

to-tail to form a protofilament. The protofilaments then associate laterally to form cylindrical tubes that are microtubules.20 Moreover, the electrokinetic motion of microtubules is fast and controllable21,22 and their transporting direction toward the anode is opposite to typical nontarget biomolecules because microtubules are highly negatively charged in a physiological buffer solution, subject to electrophoresis (EP).21–23 As illustrated in Figure 1B, the transporting direction of other nontarget biomolecules is usually governed by electroosmosis (EO), which is generated from the anode to the cathode because the glass surface has a negative (20) Howard, J. Mechanics of Motor Proteins and the Cytoskeleton; Sindauer: Sunderland, MA, 2001.

ζ-potential (∼25 mV).24 These different electrokinetic behaviors are hypothesized to rapidly separate the target molecule-bound microtubules from nontarget molecules. To concentrate the target molecule-bound microtubules, a flat nanochannel structure is created between the microchannels. Since the depth of the nanochannel (∼30 nm) is close to the diameter of microtubules, the nanochannel structure is hypothesized to physically filter the microtubules. At the same time, the nanochannel allows ionic current and electroosmotic flow (EOF) to pass through it to enable the microtubules to transport toward the anode continuously.2,19,25 The accumulation of the microtubules by the nanochannel can result in the concomitant concentration of the target biomolecules. To verify the hypotheses we develop a new protocol to functionalize microtubules to bind only specific target biomolecules and design and fabricate a device that consists of two microchannels (analyte and buffer, see Figure 1C) and a flat nanochannel that electrokinetically connects those adjacent microchannels (Figure 1D). We optimize the depth of the nanochannel to filter/concentrate the microtubules and then applied the device to the extraction and concentration of the target biomolecules such as streptavidin and bovine serum albumin (BSA). We characterize the performance of the device by quantifying concentration factors and formulating concentration rates of the target molecules. Lastly, we describe key design parameters related with the application of microtubules and nanofluidics to biomolecule preconcentration. EXPERIMENTAL SECTION Reagents. A solution of 80 mM BRB80 was mainly used for a buffer solution, adjusted to pH 6.8 with potassium. For all experiments, a buffer solution was mixed with 10 µM Taxol to prevent the depolymerization of microtubules. As target biomolecules, tetramethylrhodamine (TMR)-labeled streptavidin (STV) (Molecular Probes, Oregon) was diluted with the buffer solution to prepare three different initial concentrations of analytes (1 pM, 100 fM, and 10 fM). In the same way, fluorescein isothiocyanate (FITC)-labeled BSA (Sigma-Aldrich, St. Louis, MO) was diluted, and the initial concentrations were 1.49 nM, 14.9 pM, and 149 fM. Microtubule Preparation. Details of tubule purification and microtubule polymerization are found in our previous work.22,26 Briefly, tubulin was purified from cow brain by three cycles of microtubule polymerization and depolymerization followed by phosphocellulose ion-exchange chromatography, and fluorescently labeled tubulin (TMR-tubulin) was prepared by reacting polymerized microtubules with a 20-fold excess of TMR at room temperature for 30 min. Labeled tubulin was purified from this mixture by repeated depolymerization and polymerization. Mi(21) Kim, T.; Kao, M.-T.; Hasselbrink, E. F.; Meyho ¨fer, E. Nano Lett. 2007, 7, 211–217. (22) Kim, T.; Kao, M.-T.; Hasselbrink, E. F.; Meyho¨fer, E. Biophys. J. 2008, 94, 3880–3892. (23) Stra¨cke, R.; Bohm, K. J.; Wollweber, L.; Tuszynski, J. A.; Unger, E. Biochem. Biophys. Res. Commun. 2002, 293, 602–609. (24) Kirby, B. J.; Hasselbrink, E. F. Electrophoresis 2004, 25, 187–202. (25) Lee, J. H.; Chung, S.; Kim, S. J.; Han, J. Y. Anal. Chem. 2007, 79, 6868– 6873. (26) Kim, T.; Kao, M.-T.; Meyho ¨fer, E.; Hasselbrink, E. F. Nanotechnology 2007, 18, 025101.

crotubules were polymerized by incubating 2 mg/mL tubulin (equal ratios of TMR-labeled and unlabeled tubulin), 1 mM GTP, and 4 mM MgCl2 in BRB80 buffer at 37 °C for 20 min and then stabilized by the addition of 10 µM Taxol. The concentration of TMR-labeled microtubules was estimated from their average length, which was measured to be ∼10 µm in general gliding assays, and their lattice structure.22,26 As introduced above, microtubules consist of R- and β-heterodimers (∼50 kDa), which are about 8 nm in size and associate head-to-tail to form a protofilament. In mammals, 13 protofilaments associate laterally to form microtubules. Since microtubules were diluted with the buffer solution up to 100 times during all experiments, the molar concentration of microtubules was approximately (2 mg/mL)/ (50 kDa)(100-1)(8 nm/10 µm)(13-1) ) 25 pM. Microtubule Functionalization. Microtubules were functionalized in two ways. First, microtubules were biotinylated to capture and concentrate only streptavidin molecules from a complex mixture of analytes. Biotinylated microtubules were polymerized from 2 mg/mL biotinylated, unlabeled tubulin by following the same protocol used for TMR-labeled microtubules. The stoichiometry of tubulin to biotin was measured to be ∼100:1 using a spectrophotometer (BioSpec-1601, Shimadzu, Japan). Therefore, every single biotinylated microtubule (∼10 µm long) may have ∼160 biotins () (10 µm/8 nm)(13/100)). Second, microtubules were conjugated with an antibody system (BSA antibody) to concentrate a specific target protein (BSA) as follows. Unlabeled and polymerized microtubules were mixed with 5 µM EGS (ethylene glycol disuccinate di(N-succinimidyl) ester, SigmaAldrich), which is a homobifunctional cross-linking reagent and couples with molecules containing primary amine groups by amide bond. After 5 min of incubation, approximately 0.5 µM BSA antibody (ab3781, Abcam Inc., Cambridge, MA) was added, followed by 5 min of incubation. Free binding sites of EGS were quenched by the addition of 10 mM glycine, followed by 5 min of incubation, so that nontarget biomolecules containing amine groups were not allowed to bind the functionalized microtubules, producing the high selectivity of only the target biomolecules. Three different initial concentrations of FITC-labeled BSA as antigen were mixed with the functionalized microtubules, and then experiments were conducted. Micro/Nanofabrication of the Device. Standard photolithography techniques as well as reactive ion etching (RIE) were used to fabricate microchannels and nanochannels, respectively. Microchannels were photopatterned on a 400 µm thick glass wafer using a first mask and then etched with HF (49%) up to ∼7 µm. After all patterns and deposited metal coatings were removed, another photoresist mask layer was patterned by using a second mask for RIE etching to make nanochannels between the adjacent microchannels. The RIE etching rate of a borosilicate glass wafer (D263, S.I. Howard Glass Co., Inc. Worcester, MA) was about 17.5 nm/min at the conditions of 150 W power, 20 mTorr pressure, and 20 sccm CHF3. About 30 nm deep nanochannels (1.5-2.0 min etching) were mainly used except where otherwise noted. An electrochemical drilling technique was used to access the microchannels and load sample solutions. This processed glass wafer was bonded with a 100 µm thick, bare glass wafer (Erie Scientific Co., Portsmouth, NH) using a glass-glass bonding technique; first they were piranha-cleaned (1:1 ) H2SO4/H2O2) Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5385

Figure 2. Mixture of FITC-labeled BSA and TMR-labeled microtubules is injected into the analyte channel, whereas a buffer solution is injected into the buffer channel. Panels A-D were recorded at blue excitation light to see the movement of FITC-labeled BSA. As soon as 400 V (dc) is applied, FITC-labeled BSA is expelled from the microtubule concentrator by EOF. On the other hand, panels E-G were recorded at green excitation light to see the movement of TMR-labeled microtubules. Most of the microtubules are concentrated by the nanochannel, and the fluorescence intensity continuously increases.

for 15 min and rinsed for another 5 min with DI water. Second, they were treated with 85 °C KOH (49%) solution for 12 min, rinsed, spin-dried for 10 min, and then bonded together by applying pinpoint pressure with tweezers. Third, a high-temperature annealing process (550 °C for 5 h) was conducted to bond these two glass wafers permanently. Lastly, tube-shaped glass pipet tips (∼7.5 mm in diameter) were bonded concentrically over the drilled holes with UV-glue and epoxy. It is noted that the device structure is quite simple and very robust because all processes were performed in glass substrates. Fluorescence Microscopy and Experimental Setup. An inverted epifluorescence microscope (Axiovert 200, Carl Zeiss Microimaging, New York) equipped with a 40× oil immersion objective, a 100 W mercury arc lamp, and two dichronic filters for FITC and TMR was used for fluorescence microscopy. Imaging was performed using a digital CCD camera (Orca ER II, Hamamatsu, Japan), and quantification was carried out using Image J. To eliminate the effect of nonspecific, fluorescent species on the quantification of the protein concentration, all experiments were conducted using newly fabricated devices and background signals were subtracted from sample signals. To further clear any contaminations of the channels, BRB80 buffer was loaded into both the analyte channel and the buffer by applying a pressure differential of ∼50 kPa between the reservoirs for 20 min with a hand vacuum pump with gauge (S94224, Fisher Scientific). Subsequently, 100 µL of the buffer solution containing 0.14 mg/ mL of casein was introduced into the analyte channel for 20 min in the same way and then incubated for 5 min in the absence of a pressure differential before analytes were loaded. On the other hand, the buffer channel was filled with the buffer solution containing 10 µM Taxol without casein. Casein molecules were adsorbed onto the entire surface of the analyte channel. This casein coating turned out that nonspecific bindings of proteins and microtubules were dramatically reduced, but the ζ-potential of glass surfaces was not influenced. A high-voltage power supply (PS350, Stanford Research Systems, Sunnyvale, CA) was used to apply electric potentials to 5386

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

the microchannels via clean, bright platinum electrodes by placing them into the reservoirs. The strength of the applied electric potential was controlled using a custom-built electrical circuit. RESULTS Microtubule Concentration with the Nanochannel. First of all we verified the feasibility of the device by demonstrating that microtubules are concentrated by the nanochannel/concentrator and simultaneously nontarget molecules (e.g., FITC-labeled BSA) are removed from the concentrator. As shown in Figure 2, we loaded a mixture of FITC-labeled BSA and TMR-labeled microtubules into the analyte channel and a physiological buffer solution containing no fluorescein into the buffer channel. In the absence of an electric potential (Figure 2A, t ) 0 s), the analyte channel contained a large amount of FITC-labeled BSA and thereby was much brighter than the buffer channel. However, as soon as 400 V (dc) was applied between the microchannels, FITClabeled BSA was transported toward the cathode by EOF, and approximately 40 s later, most of the BSA molecules were removed from the microtubule concentrator (Figure 2A-D). On the other hand, TMR-labeled microtubules continued to move toward the anode but were unable to penetrate the nanochannel, resulting in their accumulation (Figure 2E-G). This physical filtering mechanism can be explained as follows: the orientations of the microtubules migrating in the analyte channel (7 µm in depth) are observed to be random. However, as they approach the nanochannel (30 nm in depth), the leading ends of them are mostly perpendicular to the entrance of the nanochannel (aligned in parallel with the electric fields, see Figure 2E-G). Given that the nanochannel is too narrow, the leading ends of microtubules are trapped/blocked by the nanochannel and cannot pass through it. This filtering mechanism enabled the nanochannel to act as a microtubule concentrator and EOF to separate nontarget molecules from the concentrated microtubules; EOF flows through the nanochannel and then carries nontarget molecules (BSA, streptavidin, casein, etc.) away from the concentrator toward the cathode. Quantification of these qualitative results further validated

Figure 3. Image sequences show the qualitative results of the concentration of TMR-labeled streptavidin analyte (1 pM and 10 fM) with biotinylated microtubules in the presence of the electric potential of 400 V. The fluorescence intensities gradually increase, and the concentration area continues to expand with time. This is entirely due to the electrokinetic transport of functionalized microtubules, implying that streptavidin can be selectively separated from the analytes and highly concentrated by the nanochannel.

our working mechanism (see Figure S1 in Supporting Information). Hence, we were encouraged to apply the device to the extraction and concentration of TMR-labeled streptavidin and FITC-labeled BSA. Extraction and Concentration of Streptavidin. As the first target molecule, we attempted to concentrate TMR-labeled streptavidin with biotinylated microtubules because this is a widely accepted model system; streptavidin strongly binds biotin (see the illustration in Figure 3). When TMR-labeled streptavidin and biotinylated microtubules were loaded into the analyte channel, apparently it was impossible to tell whether the analyte channel contained target molecules (streptavidin) or not because they were extremely diluted with the buffer solution down to 1 pM. However, in the presence of 400 V, the fluorescence intensity at the microtubule concentrator gradually increased and the concentrated area continued to widen with time up to 3 h. This is obviously because the biotinylated microtubules are successfully bound to TMR-labeled streptavidin and then concentrated at the concentrator, resulting in the higher concentration of TMR-labeled streptavidin. When we further diluted TMR-labeled streptavidin down to 10 fM, the concentrated areas at the same time frames appeared to be similar to those of 1 pM, but the fluorescence intensities showed large difference (e.g., Figure 3, part C vs part F). This can be accounted for by the different initial concentrations of TMR-labeled streptavidin. Of course, no fluorescence signal was detected at the concentrator when the analyte contained no biotinylated microtubules; the electrokinetic mobility of streptavidin is governed by EO in physiological buffer. Extraction and Concentration of BSA. In addition to concentrating streptavidin, we tested the application of the device to the extraction and concentration of BSA; as demonstrated in Figure 2, the electrokinetic mobility of BSA is governed by EO. For this experiment, microtubules were conjugated with a BSA antibody system via EGS (see the Experimental Section). These differently functionalized microtubules were premixed with FITClabeled BSA and then loaded into the analyte channel. Again, it was impossible to detect the fluorescence signal from FITC-labeled

BSA ranging from 1.49 nM to 149 fM. However, in the presence of the same voltage (400 V), the signal gradually increased, implying the concentration of FITC-labeled BSA (see Figure S2 in Supporting Information). This is because the electrophoretic mobility of the microtubules still predominates over EO although FITC-labeled BSA molecules are bound to their surfaces. Therefore, we confirm that other target proteins can be also extracted and concentrated by differently functionalizing microtubules to be conjugated with corresponding antibody systems. We note that the protocol developed in this work is very flexible to replace the BSA antibody system with other antibody systems. Again, without the functionalized microtubules, no fluorescence signal was detected at the concentrator in the control experiment. Characterization of Concentration Factors. To characterize the ability of the device to concentrate diluted analytes we quantified the fluorescence intensities of the concentrator, converted them to local molar concentrations, and then compared them with initial concentration. We used the microscope calibration method that was previously established based on the stepwise photobleaching of single dye molecules (1 AU ∼ 3.1 molecules/ µm2).27 For example, since the average fluorescence intensity of Figure 3G over the concentration area (Ac ) 30 µm × 50 µm, it varies with time) corresponds to ∼1000 AU, the total number of the concentrated sample is about (1000)(3.1 molecules/µm2)(Ac) ) 4.65 × 106 molecules. By assuming the height of the concentrated sample to be hc ) 2 µm, the volume of the sample is approximated to be Vc ) Achc; this seems reasonable because the focal depth of the images is ∼2 µm. Therefore, the resulting local molar concentration becomes 4.65 × 106molecules/Vc ) 2.58 µM. As shown in Figure 4A, approximately 106-108-fold concentrations were achieved from the 1 pM to 10 fM analytes, which are better or comparable compared to those of other previous techniques.2,19 Similarly, the concentrations of FITC-labeled BSA were also quantified and resulted in 105-106-fold concentrations, as shown in Figure 4B. The range of the concentration factors of BSA is lower than streptavidin by a factor of 10-100. In addition, there (27) Fan, C. Y.; Kurabayashi, K.; Meyho ¨fer, E. Nano Lett. 2006, 6, 2763–2767.

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5387

Figure 4. (A) Quantification of the concentration of TMR-labeled streptavidin with biotinylated microtubules, as illustrated in the inset. Three different initial concentrations of TMR-labeled streptavidin were tested over 3 h, and their resulting, averaged fluorescence intensities were measured and converted to local molar concentrations. Approximately 106-108-fold concentrations are achieved. (B) Quantification of the concentration of FITC-labeled BSA with BSA antibody-functionalized microtubules, as illustrated in the inset. In the similar way, three different initial concentrations of FITC-labeled BSA were tested over 3 h, and their resulting, averaged fluorescence intensities were measured and converted to local molar concentrations. Approximately 105-106-fold concentrations are achieved.

seems to be a linear relationship between the initial concentration of BSA and the final concentration of it, whereas we can see that the higher initial concentration of streptavidin, the higher final concentration is obtained (nonlinear). These differences may result from the fact that the association constant (affinity) between BSA (antigen) and BSA antibody (∼108 M-1) is 7 orders of magnitude weaker than that between streptavidin and biotin (∼1015 M-1) although the range of initial concentrations of BSA is close to that of streptavidin (at most 3 orders of magnitude).28,29 Also, we note that the stoichiometry ratio of BSA antibody to tubulin that may be less than that of biotin to tubulin due to the EGS connection can contribute the differences in part. Nevertheless, it is obvious that the concentration factors (>105-fold) and the sensitivity (approximately a few tens of femtomoles per liter) of the device for BSA are high enough to be applied to preconcentrating various biomolecules. DISCUSSION Concentration Rates. We theoretically formulate the concentration rate (rateMT, molecules/s) of microtubules (or target molecules) to better describe the performance of the device as follows. rateMT ) 2CMT(µEP_MT + µEOF)EAxNavo

(1)

Here, CMT is the concentration of functionalized microtubules (or active binding sites for target molecules), µEP_MT is electrophoretic mobility of microtubules (µEP_MT ) -(3.0 ± 0.2) × 10-4 cm2/V · s), µEOF is the electroosmotic mobility of the buffer solution on glass surfaces (µEOF ) (1.49 ± 0.06) × 10-4 cm2/V · s), E is the strength of the electric fields induced between the analyte channel ends (E ) 50 V/cm), Ax is the cross-sectional area of the analyte channels (Ax ) 7 µm × 300 µm), and Navo is Avogadro’s number. Equation 1 contains both experimental conditions (CMT, µEP_MT + µEOF, and E) and design/fabrication parameters (Ax). As the first (28) Gonzalez, M.; Bagatolli, L. A.; Echabe, I.; Arrondo, J. L. R.; Argarana, C. E.; Cantor, C. R.; Fidelio, G. D. J. Biol. Chem. 1997, 272, 11288–11294. (29) Lee, B. S.; Krishnanchettiar, S.; Lateef, S. S.; Gupta, S. Electrophoresis 2005, 26, 511–513.

5388

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

experimental condition, the initial concentration of the microtubules (CMT) was determined to be 25 pM to provide sufficient binding sites for target molecules. From all values used in this work, rateMT can be approximated to be 4700 microtubules/s and, in turn, the concentration rate of streptavidin can be approximated to be (4700 microtubules/s)(160) ) 7.5 × 105 molecules/s; a single microtubules has 160 biotin binding sites (see the Experimental Section).22 Therefore, the highest concentration of streptavidin over an hour becomes ∼1.5 mM ((7.5 × 105 molecules/s)(1 h)/ Vc). This value represents a theoretically possible maximum concentration rate, which can be defined to be the concentration capacity of the device, and can be a useful guideline to determine CMT for a certain concentration of analytes. The second experimental condition is the net electrokinetic mobility (µEP_MT + µEOF). Even though we change buffer conditions such as ionic strength, pH, and temperature, it is not easy to manipulate the net electrokinetic mobility because the change of the electrophoretic mobility of microtubules (µEP_MT) is reverse to the electroosmotic mobility (µEOF).30 Instead, a biochemical approach seems suitable to improving µEP_MT, but it will be challenging future work. As the third experimental condition, the strength of the electric field (E) appears to be easily adjustable, but it is coupled with the design parameters because rateMT is proportional to the crosssectional area (Ax) of the analyte but E is inversely proportional. In addition, since the electrical resistance (R) of a channel is determined by Ax, the length (L), and the conductance of a buffer solution (σ) (i.e., R ) L/(σAx), in fact, σ seems to vary along the channels because of the ion depletion and/or exclusion–enrichment effect (EEE),2,31 but the conductance was assumed to be a constant), E may be limited to Joule heating of the nanochannel. Hence, as a key design parameter, we determined the dimension of the nanochannel first to be 30 nm × 40 µm × 4 µm (D × W × L). And then the analyte channel (7 µm × 300 µm × 50 mm, Ra) was in turn designed to have ∼10 times higher electrical resistance than the nanochannel (Rn) as well as the buffer channel (7 µm × 1000 µm × 10 mm, Rb) (i.e., Ra/Rn ) ∼10 and Ra/Rb ) ∼10) in (30) Probstein, R. F. Physicochemical. Hydrodynamics, 2nd ed.; Wiley-Interscience: New York, 1994; pp 190-202. (31) Plecis, A.; Schoch, R. B.; Renaud, P. Nano Lett. 2005, 5, 1147–1155.

Figure 5. Sixty nanometer deep nanochannel concentrator shows the concentration process of a 10 pM of TMR-labeled streptavidin analyte with biotinylated microtubules. Streptavidin-bound microtubules are concentrated at the left (cathodic) side of the nanochannel, as demonstrated in 30 nm deep concentrators. However, some microtubules pass through the 60 nm deep nanochannels and then accumulate at the anodic side. Flow vortices and instability observed only at the anodic side disperse the accumulation microtubules, failing in gradual concentration of target molecules.

order to induce most of the applied voltage at the analyte channel (Ra/(Ra + Rn + Rb) ) ∼1, E ) ∼50 V/cm). These optimal resistance ratios enhanced rateMT and the electrical efficiency of the device and allowed us to avoid the effects of Joule heating and electrolysis near the nanochannel. Effect of Nanochannel Depth on Microtubule Concentration. We hypothesized that microtubules would be physically filtered by the nanochannel. However, we found that permselectivity induced by the nanochannel in the presence of electric fields might help concentrate target molecule-bound microtubules at the cathodic side because the nanochannel can be in a state of cation selectivity, hindering microtubules (anionic molecules) from penetrating through it.2,19 However, the accumulation of microtubules by the nanochannel is still largely due to physical filtering because no microtubules have been observed to pass through 30 nm deep nanochannels right after an electric field was applied (before double layer overlap occurred or permselectivity was induced). A control experiment verified our hypothesis again. After we loaded analyte and buffer solution, respectively, we applied pressure between analyte and buffer channels using a hand vacuum pump to remove any bubbles trapped around nanochannels, but no microtubule passage through the nanochannel was observed in 30 nm deep nanochannels even though pressuredriven flow may carry some microtubules from the cathodic side to the anodic side. Thus, we confirmed that our hypothesis is true. The depth of the nanochannel significantly affects the concentration rate of the device according to the electrical resistance analysis above as well as the filtering mechanism. In general, a deep nanochannel appears to be more efficient because it generates a larger value of Ra/Rn than a shallow nanochannel because it induces a higher electric potential between the analyte channel. However, the maximum depth of the nanochannel may be limited to a certain extent. To find an optimal nanochannel depth, we have changed this depth from 30 to 60 nm and then tested the function of the nanochannel. In the case of 60 nm nanochannels, we have observed the similar microtubule concentration at the cathodic side as seen in 30 nm nanochannels, but some microtubules pass through the nanochannel and then aggregate at the anodic side or transport toward the anode, as shown in Figure 5. This observation can be explained by the fact that the nanochannel is deep enough (about 2 times greater than the microtubule diameter) to be penetrated by the microtubules and the electrical double layer (∼1 nm at BRB80 buffer)30 may

not affect the passage of them. Thus, from the viewpoint of the concentration factor, 60 nm deep nanochannels are less efficient than 30 nm because they cannot completely filter/concentrate microtubules, decreasing the amount of the microtubules including target molecules at the cathodic side. Thus, 30 nm deep nanochannels appear to be an optimized microtubule concentrator. A question may arise why the microtubules were able to be accumulated at the anodic side of the nanochannel as seen in Figure 5. Although our approach used the cathodic side of the nanochannel, similar nanofluidic structures have used the anodic side to preconcentrate proteins and peptides by exploiting the phenomena of ion depletion and nonlinear EOF2,32 that also seem to be explained by an EEE.31 In part, these phenomena may explain the accumulation of the microtubules at the anodic side of the 60 nm deep nanochannel, and the nonspecific bindings of the microtubules can be another reason as well; the microtubules can be cross-linked by the STV that has four biotin binding sites. The previous work that employed the phenomena of ion depletion and nonlinear (second kind) electrokinetic flow reported that flow vortices as well as flow circulations were observed at the anodic side of the nanochannel.32–35 We also have observed similar flow vortices as well as flow instability (similar to flow circulation) in both 30 and 60 nm nanochannels. However, no vortices and instability were observed at the cathodic side of such nanochannels even though we used a high electric potential (>400 V compared to ∼25 V) so that our approach seems more stable than others. Moreover, since the concentrated microtubules were dispersed by flow vortices and frequently some of them were detached from the nanochannel and then transported along the buffer channel, the target molecule-bound microtubules failed to gradually accumulate at the anodic side. Detection and Concentration Limit of the Device. It would be important to discuss the detection and concentration limit of our device. First there can be certain biomolecules that can be transported toward the anode in the presence of an electric field such as microtubules; their electrophoretic mobilities (µEPprotein < 0) are more dominant than µEOF > 0. In this case, microtubules are not necessary to extract these molecules. However, their (32) Jin, X. Z.; Joseph, S.; Gatimu, E. N.; Bohn, P. W.; Aluru, N. R. Langmuir 2007, 23, 13209–13222. (33) Kim, S. J.; Wang, Y. C.; Lee, J. H.; Jang, H.; Han, J. Phys. Rev. Lett. 2007, 99, 035901. (34) Pundik, T.; Rubinstein, I.; Zaltzman, B. Phys. Rev. E 2005, 72, 061502. (35) Rubinstein, I.; Zaltzman, B.; Lerman, I. Phys. Rev. E 2005, 72, 011505.

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

5389

concentration may depend on their size compared to the nanochannel. All biomolecules used in our experiments passed through the 30 nm deep nanochannel and then continued to migrate toward the anode when a reverse electric field was applied except nonspecific bindings around the nanochannel; we run the experiments no more than 1 min to exclude the effect of ion depletion/ permselectivity. On the other hand, for certain other biomolecules of which electrophoretic mobilities (µEPprotein > 0) are much dominant than µEP_MT + µEOF < 0, they cannot be concentrated with microtubules because the net mobility of target moleculebound microtubule construct is a positive value (it migrates toward the anode). In this case the anodic side of the nanochannel can be employed, but the mechanism would be significantly complex because it is coupled with ion depletion/permselectivity. We have estimated the electrophoretic mobilities of FITC-labeled BSA, FITC-labeled casein, and TMR-labeled streptavidin, but all these molecules turned out that their net mobilities are governed by µEOF; they migrate toward the cathode in the presence of an electric filed, but microtubules can carry them toward the anode. Capability of Label-Free Biomolecule Detection and Postanalysis. We labeled target biomolecules with fluorescein prior to mixing them with other nontarget molecules to quantify the concentration factors. However, this labeling step would not be allowed in detecting/concentrating specific target molecules already dissolved in a complex mixture. Indeed, our method does not require labeling samples because unlabeled biomolecules can be bound to functionalized microtubules and then concentrated by the nanochannel in the same manner as demonstrated in this work. Thus, it is emphasized that our device can be applied to selectively detecting and concentrating unlabeled biomolecules from a complex mixture. If necessary, the concentrated target biomolecules bound to microtubules can be extracted out of the device by depolymerizing microtubules and then performing conventional ion-exchange chromatography for other analysis.

high-selective extraction and a nanofluidic structure for highthroughput concentration. We developed a powerful protocol that functionalizes microtubules to be conjugated with specific antibody systems so that we enabled the application of the device to various target proteins. We achieved that the detection sensitivity of the device was as high as a few tens of femtomoles per liter, and the concentration factors were as high as 105-108-fold. Whereas concentration takes place prior to separation in order in other devices, our device performed separation (extraction) and concentration simultaneously so that we significantly reduced analysis time and efforts and simplified the device structure; no separation columns were required. In addition, our device holds a unique potential to be integrated with other postanalysis systems in series because not only the separated, concentrated target molecules are useful for following processes but also microtubules are biocompatible and their molecular structures are well-established. Our approach seems electrokinetically more stable than other similar nanofluidic applications using the anodic side, because we concentrate target molecules at the cathodic side of the nanochannel where no flow vortices and instability have been observed. Finally, we believe that microtubules will be engineered for biocompatible biomolecule detection and extraction methods and the nanofluidic concentration structure will be widely applied to filtering small biomolecules in biochemical analysis systems.

CONCLUSION We successfully developed a novel micro/nanofluidic biomolecule preconcentration device by incorporating microtubules for

Received for review February 25, 2008. Accepted May 1, 2008.

5390

Analytical Chemistry, Vol. 80, No. 14, July 15, 2008

ACKNOWLEDGMENT This work was funded by NSF Award BES-0428090. SUPPORTING INFORMATION AVAILABLE Figure S1, supporting the experimental results of Figure 2 by quantifying fluorescence intensities of TMR-labeled microtubules and FITC-labeled BSA, and Figure S2, showing the experimental results of FITC-labeled BSA concentrations using BSA antibodyfunctionalized microtubules. This material is available free of charge via the Internet at http://pubs.acs.org.

AC8003874