Article pubs.acs.org/ac
Enhancement in the Sensitivity of Microfluidic Enzyme-Linked Immunosorbent Assays through Analyte Preconcentration Naoki Yanagisawa and Debashis Dutta* Department of Chemistry, University of Wyoming, Laramie, Wyoming 82071, United States S Supporting Information *
ABSTRACT: In this Article, we describe a microfluidic enzyme-linked immunosorbent assay (ELISA) method whose sensitivity can be substantially enhanced through preconcentration of the target analyte around a semipermeable membrane. The reported preconcentration has been accomplished in our current work via electrokinetic means allowing a significant increase in the amount of captured analyte relative to nonspecific binding in the trapping/detection zone. Upon introduction of an enzyme substrate into this region, the rate of generation of the ELISA reaction product (resorufin) was observed to increase by over a factor of 200 for the sample and 2 for the corresponding blank compared to similar assays without analyte trapping. Interestingly, in spite of nonuniformities in the amount of captured analyte along the surface of our analysis channel, the measured fluorescence signal in the preconcentration zone increased linearly with time over an enzyme reaction period of 30 min and at a rate that was proportional to the analyte concentration in the bulk sample. In our current study, the reported technique has been shown to reduce the smallest detectable concentration of the tumor marker CA 19-9 and Blue Tongue Viral antibody by over 2 orders of magnitude compared to immunoassays without analyte preconcentration. When compared to microwell based ELISAs, the reported microfluidic approach not only yielded a similar improvement in the smallest detectable analyte concentration but also reduced the sample consumption in the assay by a factor of 20 (5 μL versus 100 μL).
T
nonspecific binding of the detection antibody and/or enzyme label in the system through use of various blocking agents.6−8 To reduce the sample volume requirement in these assays, miniaturization and automation technologies have been extensively employed to allow the quantitation of smaller amounts of protein biomarkers. 9,10 The use of these miniaturization/automation technologies has further enhanced the multiplexing capability of the ELISA method leading to the development of high throughput platforms for biological analyses.11−13 To enhance the kinetics of immunoreactions, ELISA methods have been also implemented on micrometer sized beads, which has helped in increasing the capture efficiency of the analyte molecules at reduced incubation times due to their larger surface area-to-volume ratios. Moreover, several detection methods ranging from radiometry to timeresolved fluorescence have been integrated to this immunoassay technique permitting large improvements in the detectability of enzyme reaction product.14 Along the same lines, ELISA methods have been reported in which the initially formed enzyme reaction product dimerizes or polymerizes with other substrate molecules (e.g., 2,2-diaminobenzidine (DAB) or 3amino-9-ethylcarbazole (AEC)) to yield an insoluble chemical
he detailed understanding of many biological processes often requires the detection and quantitation of biomolecules present at extremely low concentrations.1 For several of these applications, techniques that are inherently very sensitive (e.g., fluorescence) may not be sensitive enough to extract the desired biological information. In such situations, it is desirable to apply assay methods in which the analyte signal is amplified in some way relative to the background. While the use of PCR based approaches has become quite common in accomplishing this goal when the analyte is a nucleic acid,2 enzyme-linked immunosorbent assays (ELISA) have the virtue of applicability to a wide variety of targets, essentially anything to which an antibody can be generated. Although highly sensitive relative to nonamplification techniques, increasing the sensitivity of ELISA methods is an important goal.3 The most obvious reason for improving ELISA sensitivity is that it would allow the detection of biomolecules at substantially lower levels than currently possible, while at the same time provide data of better quality and reliability for concentration ranges in current use. To increase the sensitivity of ELISA methods, several approaches have been reported in the literature that focus on optimizing its biochemical and analytical aspects. For example, a variety of ELISA surfaces have been developed to enhance the specific capture of target analytes to a solid support, thereby increasing the signal in the assay.4,5 At the same time, a significant effort has been made to reduce the amount of © 2012 American Chemical Society
Received: May 2, 2012 Accepted: July 30, 2012 Published: July 30, 2012 7029
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry
Article
species that can be seen as a spot.15 In this case, the precipitation of the ELISA reaction product in a small area automatically preconcentrates this signaling molecule improving its detectability and thereby that of the analyte. Recently, electrokinetic forces have been employed to accomplish the same goal for solubilized products of the enzyme reaction leading to similar benefits in the performance of the assay.16 In this Article, we describe an alternative approach to reducing the limit of detection for microfluidic ELISAs through preconcentration of the target analyte around a semipermeable membrane. This strategy has been shown to significantly increase the local analyte concentration in the vicinity of the membrane during its immunocapture process resulting in over a 200-fold enhancement in the rate of signal generation in this region. Not surprisingly, the reported preconcentration method also leads to a greater amount of nonspecific binding of the detection antibody/enzyme label in the system. However, because the nonspecific interactions of biomolecules with the ELISA surface tend to be significantly weaker than that between an antigen−antibody pair, the preconcentration approach described here was observed to increase the signal level in the blank assay only by a factor of 2. Overall, the reported assay has been shown to decrease the smallest detectable analyte concentration for the tumor marker CA 19-9 and Blue Tongue Viral antibody by over 2 orders of magnitude and using only a twentieth of the sample volume compared to that required for commercial microwell based ELISAs (5 μL versus 100 μL). It must be noted that the use of membrane structures for enhancing the sensitivity of microfluidic devices has been previously demonstrated in several applications including sample preconcentration,17−19 gas sensing,20 filtration,21,22 and microreactor design,23 among others. Their utility in improving the detection limit of ELISA methods, however, has not been explored in the past and is the focus of our current work.
Figure 1. Schematic of the microfluidic device used in the current work for performing the CA 19-9 and Blue Tongue Viral (BTV) antibody preconcentration ELISAs.
depth of 30 μm. Following this step, the microfluidic network was sealed off by bringing a cover plate in contact with the bottom substrate in deionized water and allowing the two plates to bond under ambient conditions overnight.25 Membrane Fabrication. Prior to the membrane fabrication process, the microchannels in this fluidic network were thoroughly rinsed with 1N sodium hydroxide for about 20 min. A precursor solution26,27 containing 21% (v/v) (37:1) acrylamide/bisacrylamide and 0.2% (w/v) VA-086 photoinitiator was then introduced through reservoir 3 filling up the shallow region as well as segments C and D. It was observed that the larger capillary force within the shallow segment prevented the precursor solution from flowing beyond it into segments A and B during this step. At this point, the excess precursor material was emptied from segments C and D by applying vacuum to reservoir 4. Note that, during this purging process, the precursor solution within the shallow region did not flow out, again due to a larger capillary force in it. The entire device was then exposed to ultraviolet light for about 10 min to complete the photopolymerization process. It must be pointed out that the fabrication approach described above28−30 for creating photopolymerized membranes did not require any careful alignment of a mask or a laser beam to ascertain the location of the membrane structure. Moreover, because the precursor material was removed from regions around the membrane prior to the photo exposure, the axial extent of this structure was not affected by the propagation of the photopolymerization reaction beyond the shallow segment. On the other hand, as the shallow region in our fluidic network was created by depositing a layer of photoresist manually, membrane structures shorter than 1 mm in axial extent could not be reliably fabricated using our approach. For the 2 mm long membranes used in our current work, the reproducibility in their axial lengths was observed to be around 90%. ELISA Surface Preparation. The analysis channel in our microchip (segments A and B) was prepared for an ELISA by treating it with a solution of (3-aminopropyl)triethoxysilane (Sigma-Aldrich) for an hour under ambient conditions. Subsequently, this fluidic duct was rinsed with methanol and then reacted at room temperature with an aqueous solution containing 5% w/v glutaraldehyde (Sigma-Aldrich) for another 60 min to create a surface that could be covalently bonded to the amine groups on a protein molecule. 12,13 Excess glutaraldehyde was removed from the system by rinsing the microchip with deionized water for about 5 min.
■
EXPERIMENTAL SECTION Device Preparation. For fabricating the microfluidic devices employed in this work, bottom substrates and cover plates made from borosilicate glass were purchased from Telic Company (Valencia, CA). While the purchased cover plates had both their faces unprotected, the bottom substrates came with a thin layer of chromium and photoresist laid down on one of their surfaces. The fabrication process for the microchips was initiated by photolithographically patterning24 the desired channel layout (see Figure 1) on the bottom substrate using a custom designed photomask created through Fineline Imaging Inc. (Colorado Springs, CO). The channel width in our device was chosen to be 500 μm for the entire microfluidic network while the analysis channel in front of the membrane (segments A and B) was about 1.5 cm long. After completion of the photopatterning process, the photoresist layer was cured in microposit developer MF-319 (Rohm and Haas) and the chromium layer was removed along the channel network with a chromium etchant (Transene Inc.). In order to accommodate the preconcentrating membrane in our device, a shallow region was created at the location where this structure was to be placed (see Figure 1). This was accomplished by first etching the entire channel network to a depth of 2 μm and subsequently manually covering up about a 2 mm long region at this location with a layer of photoresist. After drying this photoresist layer by heating the microchip at 80 °C for 20 min, the remaining fluidic network was further etched to a final 7030
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry
Article
BTV Antibody Assay. The device for the Blue Tongue Viral (BTV) antibody assay was realized by first incubating excess amounts of a polyclonal BTV capture antibody prepared in 0.1 M carbonate−bicarbonate buffer (pH 9.4) in the glutaraldehyde coated microfluidic network for 2 h. These antibodies were produced by inoculating recombinant BTV31 into a rabbit specimen and drawing 10 mL of its blood two weeks after infection following standard procedures.32 The blood sample was then centrifuged at 1000g for 15 min, and the supernatant fluid was used as the mother solution for the capture species. Preliminary experiments showed (data not presented) that the concentration of this capture species did not affect the assay results provided a 100-fold (100×) or lesser dilution of its mother solution was used. In this situation, a 50× dilution of this source liquid was incubated in segments A and B of our microchip to ensure the excess of the BTV capture antibodies on the ELISA surface. This microchannel was then incubated with a solution containing 0.1 M lysine in 0.1 M carbonate− bicarbonate buffer for 1 h and subsequently treated for another hour with a 1% (w/v) BSA solution in carbonate−bicarbonate buffer to block off the unreacted sites on the channel surface. These fluidic ducts were later exposed to a solution containing excess amounts of the BTV antigen prepared in 0.1 M phosphate buffer (pH 7.4). This virus was expressed in Sf9 insect cells by infection with recombinant baculovirus containing the cloned VP7 gene of BTV using the procedure describe by Mecham and Wilson.33 A 10× dilution of the supernatant fluid from the Sf9 cell culture mentioned above was used to coat our microchannel with the BTV antigen. As before, this dilution factor was arrived at through preliminary experiments (data not shown), which showed that incubation with a 10-fold dilution of the BTV antigen mother solution for an hour ensured the excess of this virus on the ELISA surface in our assays. At this point, different dilutions of the BTV antibody (analyte) prepared in 0.1 M phosphate buffer was introduced in segments A and B of our device. The mother solution for this antibody was the supernatant fluid obtained from a hybridoma culture prepared using the spleen of an adult female BALB/c mice that had been immunized by a series of intraperitoneal and intravenous inoculations with the recombinant BTV antigen.31,34 For the preconcentration assays, high voltages were applied to reservoirs 1/2 while electrically grounding reservoirs 3/4 during the incubation of the BTV antibody sample (1 h). The direction of this electric field was chosen on the basis of the fact that electrokinetic transport of protein molecules in segments A and B was governed by electroosmosis which occurred from a region of high voltage to a region of low voltage in our system. This mechanism for protein transport was confirmed in our work by observing the flow of different fluorescently tagged proteins, e.g., FITClabeled BSA (Sigma-Aldrich), and dylight-488 monoclonal human and mouse antibodies (Kirkegaard & Perry Laboratories, Inc.), toward the polyacrylamide membrane under the influence of an electric field. Following the immobilization of the analyte, excess amounts of biotinylated goat antimouse immunoglobulin (10× dilution of a mother solution obtained from BioGenex Laboratories) prepared in phosphate buffer was introduced into segments A and B of our microchip and incubated for 1 h. Finally, the microchannels were treated with excess amounts of streptavidin-horseradish peroxidase conjugate (10× dilution of a mother solution obtained from BioGenex Laboratories) prepared in phosphate buffer containing 0.05% v/v Tween 20 (Sigma-Aldrich) for 1 h to complete
the ELISA surface. Note that each of the incubation steps described above were performed at room temperature and were separated by rinsing steps in which segments A and B were washed with a 0.1 M buffer solution for 5 min. While the carbonate−bicarbonate buffer was used in these rinsing procedures until the ELISA channel was incubated with BSA, the washes following this incubation step were carried out with the phosphate buffer. Cancer Marker CA 19-9 Assays. For the cancer marker CA 19-9 assays, the glutaraldehyde surface in segments A and B of our microchip was reacted with 0.1% (w/v) streptavidin (Sigma-Aldrich) overnight in a 0.1 M carbonate−bicarbonate buffer at 4 °C. This microchannel was then incubated with a solution containing 0.1 M lysine in 0.1 M carbonate− bicarbonate buffer for 1 h and subsequently treated with a 1% (w/v) BSA solution in carbonate−bicarbonate buffer for 60 min to block off the unreacted sites on the channel surface. Following this derivatization process, the analysis channel was incubated for an hour with excess amounts of biotinylated capture antibody for CA 19-9 that was obtained as part of an ELISA kit from Alpha Diagnostic International Inc. The analysis channel was then reacted with different concentrations of the analyte CA 19-9 and later with an excess amount of antiCA 19-9 horseradish peroxidase conjugate for an hour each to complete the ELISA surface. As before, for preconcentrating the CA 19-9 species around the polyacrylamide membrane, high voltages were applied to reservoirs 1/2 while electrically grounding reservoirs 3/4 over the entire sample incubation period (1 h). Also, each of the incubation steps described above were performed at room temperature and were separated by rinsing steps in which segments A and B were washed with a 0.1 M buffer solution for 5 min. While the carbonate− bicarbonate buffer was used in these procedures until the ELISA channel was incubated with streptavidin, the washes following this incubation step were carried out with the phosphate buffer. Device Operation. The enzyme-linked immunosorbent assays were performed by introducing a solution containing 10 μM Amplex Red and 5 μM hydrogen peroxide in 0.1 M phosphate buffer. In our experiments, this mixture was entered into the analysis channel via capillary forces with no fluid reservoirs attached to the channel terminals. The minimal amount of liquid introduced at the channel ports (∼1 μL) eliminated any measurable flow within the conduits due to differences in hydrostatic heads preventing advective transport of the enzyme reaction product during the enzyme reaction period. Liquid evaporation from the channel terminals was minimized by sealing these ports with adhesive tapes. The enzyme reaction was performed in our assays by maintaining an air temperature of 37 °C around the microchip (measured using a thermometer) through placement of a heating fan close to it. It must be noted that, without this temperature control, the assay-to-assay variability in the ELISA signal was significantly larger than that reported here likely due to fluctuations in the room temperature. In our experiments, the fluorescence measurements at the detection region were made using an epifluorescence microscope (Nikon) with bandpass excitation (528−543 nm) and emission (590−650 nm) optical filters. The fluorescence images thus obtained were recorded using a CCD camera (Roper Scientific) and analyzed with the Adobe Photoshop software. To minimize any unwanted signal generation in our system through spontaneous photo-oxidation of Amplex Red,35 the ELISA region was exposed to the 7031
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry
Article
fluorescence image from one such CA 19-9 assay has been presented after an enzyme reaction period of 25 min for an analyte concentration of 5 U/mL. As expected, the image shows a significantly higher fluorescence around the preconcentration membrane which decays off rapidly as we move away from this structure. We have quantitated this image by plotting the observed brightness on a gray scale along the centerline of our analysis channel, which was determined using an in-house written MATLAB code. The x-coordinate in this graph represents the distance (in mm) measured from the junction of segments A and B along the centerline of these conduits. The fluorescence profile presented in Figure 2 shows that, in spite of the existence of a sharp gradient in the ELISA signal around the polyacrylamide membrane, this quantity is relatively uniform within the region |x| < 0.25 mm. The variation in the signal intensity within this region was observed to be less than 3% of the change in fluorescence going from the center of our analysis channel (x = 0) to its edges in the image in Figure 2 (x = ± 3 mm). Consequently, we decided to quantitate our immunoassays by monitoring the fluorescence intensity at x = 0 over an enzyme reaction period of 25−30 min. In Figure 3a, we have presented data from a series of such CA 19-9 ELISA experiments which show a linear increase in this fluorescence intensity with time. While it is possible that the spatial nonuniformity in the concentration of resorufin molecules may have affected the measured signal in our detection region, this inhomogeneity did not seem to alter its linear variation with time. In this situation, the temporal rate of change in the fluorescence intensity in our detection region was chosen as the measure for the analyte concentration in our ELISA device (kinetic ELISA format).36 As is the case with most ELISA systems, the smallest detectable analyte concentration (limit of detection or LOD) in our device was dictated by the amount of nonspecific binding of the detection antibody and/or the streptavidin−horseradish peroxidase complex to the ELISA surface. This undesirable immobilization represented by the rate of signal generation in the blank assay was observed to be substantially larger (by over a factor of 5) than the noise in the system (measured by the amount of scatter in the data). As a result, the response curves for the analytes (CA 19-9 and BTV antibody) were generated
excitation beam for about 1s during the imaging process through the use of a mechanical shutter. The camera exposure time was chosen to be 100 ms in all our measurements.
■
RESULTS AND DISCUSSION Feasibility Study. The ELISA method reported in this Article relies on the trapping of analyte molecules around a semipermeable membrane during its incubation period and, therefore, automatically introduces a spatial gradient in the concentration of this species along the surface of channel segments A and B in our device. Consequently, the rate of generation of the fluorescent ELISA reaction product (resorufin) was observed to be spatially nonuniform in this conduit leading to its diffusive migration away from the membrane structure during the enzyme reaction period.12 In this situation, it becomes important to establish that such diffusion effects do not restrict our ability to reliably quantitate the reported preconcentration assays. To this end, we initiated our study by mapping the fluorescence profile in the analysis channel (segments A and B) as a function of both space and time in our preconcentration ELISA experiments. In Figure 2, a
Figure 2. Fluorescence profile in segments A and B of our microfluidic device during a CA 19-9 preconcentration assay. The measurement reported here was made for a sample containing 5 U/mL of CA 19-9 after an enzyme reaction period of 25 min.
Figure 3. (a) Temporal variation in the measured fluorescence at the detection point (channel center at the junction of segments A and B) for the preconcentration assays performed with samples containing different amounts of CA 19-9. The CA 19-9 concentration reported here refers to that in the bulk sample. (b) Response curves for CA 19-9 as determined using the reported microchip device (with and without preconcentration) as well as the commercial microwell plates obtained from Alpha Diagnostic International Inc. The chosen range of concentrations was within the linear dynamic range of the assay. 7032
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry
Article
Figure 4. (a) Temporal variation in the measured fluorescence at the detection point (channel center at the junction of segments A and B) for the preconcentration assays performed with samples containing different amounts of BTV antibody. (b) Response curves for the BTV antibody as determined using the reported microchip device (with and without preconcentration) as well as commercial microwell plates. The chosen range of concentrations was within the linear dynamic range of the assay.
preconcentration assays from 1 μL to about 5 μL for an hour long analyte incubation process. To provide the 5 μL sample aliquot in these experiments, 1 cm tall fluid reservoirs with an internal diameter of 1.2 mm were attached to each of the channel terminals 1 through 4. For the assays without preconcentration, the use of these reservoirs was not necessary and the entire sample (∼1 μL) could be introduced into the analysis channel right at the start of the analyte incubation period. The microwell assays shown in Figure 3b were performed using an ELISA kit obtained from Alpha Diagnostic International Inc. following the instructions recommended for it. The only procedural change made in these microwell assays involved the switching of the enzyme substrate from 3,3′,5,5′ tetramethylbenzidine (provided by the vendor) to 10 μM Amplex Red/5 μM hydrogen peroxide prepared in 0.1 M phosphate buffer as was used in the microchip ELISAs. Also, to ensure a fair comparison between our microchip and microwell experiments, the latter were quantitated using the same fluorescence microscope system (instead of a microplate reader) as was used for the former. Interestingly, the sensitivity of the microwell plate ELISAs thus performed was found to be comparable to that of our microchip ones without analyte preconcentration.12,13 It must be noted that the range of analyte concentrations chosen for the experiments in Figure 3 was within the linear dynamic range of our system. In this situation, the response curve in our assays was fitted to a straight line rather than a more complicated curve such as that yielded by the 4-parameter logistic equation.37 To determine the smallest detectable analyte concentration (LOD) in our assays, the signal-to-noise ratio (S/N) in our system was estimated for the experiments shown in Figure 3b and plotted against CA 19-9 concentration in the bulk sample (see Figure S1 in the Supporting Information). In this plot, while the ELISA signal was set equal to the temporal rate of change in the measured fluorescence minus that for the corresponding blank, the noise in our experiments was evaluated as the standard deviation in this signal. The LOD for our assays in this situation was the analyte concentration at which the signal-to-noise ratio equaled 3. The approach outlined above yields LOD values of 0.037 ± 0.002 U/mL and 5.3 ± 0.3 U/mL for the CA 19-9 microchip ELISAs with and without analyte preconcentration. This quantity for the microwell based assays was determined to be 3.8 ± 0.1 U/mL
in this work by plotting the rate of signal generation for a sample minus that for the corresponding blank against the analyte concentration in the bulk. It must be pointed out that the kinetic ELISA format employed here proved to be a reliable way of quantitating our assays as it eliminated the effect of any variation in the background fluorescence, i.e., observed fluorescence at the start of the enzyme reaction period, in going from one microchip device to the other. This variation, although small for the preconcentration experiments, acquired values comparable to the change in the fluorescence signal over the ELISA reaction period for assays without preconcentration. Moreover, because the temporal rate of change in the fluorescence signal was determined on the basis of multiple data points in the kinetic format of our assays rather than a single measurement as in end-point ELISAs, the established correlation between this quantity and the analyte concentration was observed to be quite reliable.12,13 Assay Performance. Having established the quantitative nature of our preconcentration assays, we focused on estimating the improvement in LOD yielded by them over similar ones performed without preconcentration. To this end, the response curves for the analytes (CA 19-9 and BTV antibody) were determined in our microchip device with and without preconcentration in addition to doing the same on commercial microwell plates. In Figure 3b, we have presented data from these experiments for the tumor marker CA 19-9 which shows a 241-fold increase in sensitivity for the response curve (measured in terms of its slope) upon analyte preconcentration in our microchip device. It is important to note that, while the reported approach enhanced the observed fluorescence in our microchip ELISAs by over 2 orders of magnitude, it also doubled the rate of resorufin generation in the blank assay. This is likely due to a greater nonspecific capture of other proteins in the detection region during the sample preconcentration process, which in turn resulted in larger unwanted immobilization of the enzyme label around the membrane structure. For these preconcentration experiments, 1000 V was applied between reservoirs 1/2 and 3/4 over the analyte incubation period. The application of this voltage was observed to generate an electroosmotic flow through the polyacrylamide membrane which was established by monitoring the flow of the neutral dye Rhodamine B through it. Such a flow increased the sample volume requirement in our 7033
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry
Article
ELISAs was 100 μL as compared to 1 μL in the case of the microfluidic assays. In all of the preconcentration ELISAs described above, the electrical voltage applied across reservoirs 1/2 and 3/4 in our microchip was 1000 V. Surprisingly, this operating parameter did not affect the assay performance in any significant way both for the CA 19-9 and the BTV antibody assays. In the Supporting Information (see Figure S2), we have presented some representative results that depict the effect of the preconcentration voltage on the rate of signal generation in the ELISA method reported here. The figure shows that, upon increasing the electric potential from 1000 to 1500 V, the latter quantity only increased by 8% for a 1.5 ng/mL BTV antibody sample. On the other hand, when this preconcentration voltage is reduced from 1000 V to 500 V, the ELISA signal in the system dropped by about 5% for the same sample. A possible reason for the weak dependence of these two quantities on each other is the existence of fluid circulations around the membrane during the preconcentration period. These circulations likely originating from a mismatch in the electroosmotic flow at the membrane interface were observed in our experiments whenever the neutral dye Rhodamine B was electrokinetically flown across the polyacrylamide structure.28,29 In this situation, although the increase in preconcentration voltage may have resulted in the transport of more analyte molecules toward the polymer membrane in our assays, it probably also strengthened the fluid circulations around it, thereby reducing the capture efficiency of the analyte species in the detection region. For applied electric potentials of the order of 1000 V, these two factors appear to be nullifying each other leading to an insignificant improvement in performance of our microchip device with increase in the preconcentration voltage. It is important to point out that, while commercial ELISA kits available today can detect BTV antibody in serum samples from cattle and sheep about a week after infection, there is a need for improving the sensitivity of these assays for a couple of reasons. First, such an improvement would allow reliable detection of this viral infection at an early stage than currently possible. With no effective treatment available for this disease, early diagnosis can be key to preventing a BTV outbreak among artiodactyls in zoos and livestock farms, thereby minimizing economic damages. Second, a more sensitive BTV antibody assay could help us understand the role, if any, of animals like dogs in the epidemiology of this viral condition. This is because the level of this antibody in the bodily fluids of such animals tends to be substantially lower compared to that in artiodactyls and often undetectable using commercial ELISA kits.39 In this situation, the reported strategy to detect smaller BTV antibody concentrations can be of high significance to both diagnostic and basic research applications. It must be noted that, in the BTV antibody assays presented here, many of the biological reagents were grown in our laboratory and are therefore unavailable to other researchers interested in reproducing our results. The primary purpose for including the CA 19-9 experiments (for which the assay reagents can be purchased from a commercial vendor) in the manuscript was to eliminate this limitation of our work as well as demonstrate that our approach is applicable to other biological molecules. For the CA 19-9 assays, the sensitivity of commercial ELISA kits available today is usually sufficient to assess the state of health of an individual, and the reported microfluidic approach can only improve the reliability of this information. However, it is quite likely that the smaller limit of detection accomplished in
which compared well to that reported in the commercial kit from Alpha Diagnostic International Inc. (3 U/mL). Notice that, although the analyte preconcentration strategy in our device led to over 200-fold improvement in the signal, it also increased the noise level in the system by about 50%. Overall, this corresponded to a 143-fold improvement in LOD over microchip ELISAs without preconcentration and a 102-fold improvement in the same quantity over assays performed using commercial ELISA kits. The detection limits reported above can be expressed in mass concentration units based on the fact that 1 U of CA 19-9 corresponds to approximately 0.8 ng of the antigen.38 The LOD values in this situation turn out to be about 30 pg/mL and 4.2 ng/mL for our microfluidic CA 19-9 ELISAs with and without preconcentration, respectively. For the microwell based assays, the corresponding quantity is estimated to be 3.0 ng/mL. In addition to performing ELISAs for assessing the level of CA 19-9 in biological samples, we also applied our preconcentration assays to the detection of Blue Tongue Viral (BTV) antibody. We initiated our work with BTV antibody by first estimating its mass concentration in the original sample so as to be able to compare our assay results to those reported for other analytes in the literature. To this end, we ran our original sample through a sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) system (Life Technologies Corporation) and measured the protein content in the band corresponding to the BTV antibody using a Coomassie Brilliant Blue G based protein quantification kit (Sigma-Aldrich). This procedure yielded a mass concentration value of 4.7 ± 0.5 μg/mL for the viral antibody in the original sample obtained from the hybridoma culture. Now adopting an approach identical to that employed in the CA 19-9 assays, the results obtained from the BTV antibody ELISAs have been summarized in Figure 4. Figure 4b shows that the reported preconcentration strategy enhanced the sensitivity of the BTV antibody ELISAs by over a factor of 200 compared to those performed on microchips without preconcentration and on commercial microwell plates. In terms of mass concentration, this corresponded to LOD values of 96 ± 4 pg/mL and 11.6 ± 0.2 ng/mL for the microchip assays with and without analyte preconcentration, respectively. The preconcentration voltage used in obtaining these results was 1000 V which was identical to that chosen for the CA 19-9 samples. For the microwell plate based experiments, the estimated LOD for the BTV antibody sample was 31 ± 2 ng/mL. It must be noted that while a majority of the steps in preparing the microwell plates (Immulon II, Dynatech Laboratories) for a BTV antibody ELISA was identical to those employed in the microfluidic version of the assay, there were three important distinctions between the two cases. First, the BTV capture antibody was laid down on the surface of the microwell plate without the use of any linkers such as (3-aminopropyl)triethoxysilane and glutaraldehyde as was employed in the glass microchannels. Second, these capture antibodies were incubated for 24 h in the microwell plates as compared to 60 min in the microfluidic ducts so as to be able to generate a measurable signal in the ELISA experiments. Note that this incubation process was carried out at 4 °C in the former case while implemented at room temperature in the latter. The incubation periods and conditions for the remaining biological reagents were chosen to be identical for both the microchip and microwell based ELISAs in our work. Finally, the volume of biological reagents used in all of the incubation steps for the microwell based 7034
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry this work could help scientists in the future obtain important biological information related to CA 19-9, which is not accessible using current technology. Finally, because the reported approach to enhancing ELISA sensitivity is generic to all analytes, its significance in advancing our knowledge in several areas of science is likely to be high.
CONCLUSIONS To conclude, we have demonstrated a novel approach to improving the sensitivity of microfluidic ELISA through preconcentration of the analyte molecules around a semipermeable membrane. The reported strategy has been shown to increase the amount of captured analyte and nonspecific binding next to the membrane interface by about a factor of 200 and 2, respectively. Overall, this corresponded to a decrease in the smallest detectable analyte concentration by over 2 orders of magnitude compared to that realized on commercial microwell plates and using only about a twentieth of the sample (5 μL versus 100 μL). In our current work, the reported improvement in ELISA sensitivity was demonstrated using the tumor marker CA 19-9 and BTV antibody samples. The highly sensitive preconcentration assays, however, yielded a lower dynamic range due to local saturation of the ELISA surface around the membrane region. In the case of the CA 19-9 marker, the maximum analyte concentration that could be quantitated using the reported preconcentration approach under the chosen assay conditions was about 12 U/mL as compared to >600 U/mL for microfluidic ELISAs without preconcentration. Interestingly, this dynamic range is determined by the surface concentration of the capture antibodies around the membrane interface and can be increased by preconcentrating these molecules using electrokinetic forces. A detailed study on this aspect of our microfluidic ELISA is currently underway and will be reported in a future publication. Our experiments also show that the transport of protein species during the reported preconcentration process is dominated by electroosmosis in the analysis channel. While this fact minimizes any biases in the preconcentration process based on the electrophoretic mobility of the target analyte, it leads to generation of fluid circulations around the membrane interface which can deteriorate the capture efficiency of the protein target in the preconcentration region. Finally, it must be noted that the approach reported here to improving ELISA sensitivity is complementary to what has been conventionally pursued in the literature, i.e., designing surfaces that minimize nonspecific binding and enhance the capture of target analytes on the ELISA surface. In this situation, the strategy developed in this work could be implemented on high performance ELISA surfaces to achieve assay sensitivities that are even greater than those accomplished here.
■
REFERENCES
(1) Zhang, H. Q.; Zhao, Q.; Li, X. F.; Le, X. C. Analyst 2007, 132, 724−737. (2) Mackay, I. M. Clin. Microbiol. Infect. 2004, 10, 190−212. (3) Kemeny, D. M.; Challacombe, S. J. ELISA and other solid phase immunoassays: Theoretical and practical aspects; John Wiley & Sons: New York, 1988. (4) Bai, Y.; Koh, C. G.; Boreman, M.; Juang, Y. J.; Tang, I. C.; Lee, L. J.; Yang, S. T. Langmuir 2006, 22, 9458−9467. (5) Seurynck-Servoss, S. L.; White, A. M.; Baird, C. L.; Rodland, K. D.; Zangar, R. C. Anal. Biochem. 2007, 317, 105−115. (6) Steinitz, M. Anal. Biochem. 2000, 282, 232−238. (7) Frederix, F.; Bonroy, K.; Reekmans, G.; Laureyn, W.; Campitelli, A.; Abramov, M. A.; Dehaen, W.; Maes, G. J. Biochem. Biophys. Methods 2004, 58, 67−74. (8) Huber, D.; Rudolf, J.; Ansari, P.; Galler, B.; Fuhrer, M.; Hasenhindi, C.; Baumgartner, S. Anal. Bioanal. Chem. 2009, 394, 539− 548. (9) Stoll, D.; Templin, M. F.; Bachmann, J.; Joos, T. O. Curr. Opin. Drug Discovery Dev. 2005, 8, 239−252. (10) Ng, A. H. C.; Uddayasankar, U.; Wheeler, A. R. Anal. Bioanal. Chem. 2010, 397, 991−1007. (11) Jin, H. J.; Zangar, R. C. Cancer Biomarkers 2010, 6, 281−290. (12) Yanagisawa, N.; Mecham, J. O.; Corcoran, R. C.; Dutta, D. Anal. Bioanal. Chem. 2011, 401, 1173−1181. (13) Yanagisawa, N.; Dutta, D. Biosensors 2011, 1, 58−69. (14) Wild, D. The Immunoassay Handbook, 3rd ed.; Elsevier Ltd.: Oxford, U.K., 2005. (15) Verlander, P. C. Detection of Horseradish Peroxidase by Colorimetry. In Nonisotopic Probing, Blotting, and Sequencing, 2nd ed.; Kricka, L. J., Ed.; Academic Press: San Diego; 1995, pp 217−236. (16) Cheow, L. F.; Ko, S. H.; Kim, S. J.; Kang, K. H.; Han, J. Anal. Chem. 2010, 82, 3383−3388. (17) Khandurina, J.; Jacobson, S. C.; Waters, L. C.; Foote, R. S.; Ramsey, J. M. Anal. Chem. 1999, 71, 1815−1819. (18) Song, S.; Singh, A. K.; Kirby, B. J. Anal. Chem. 2004, 76, 4589− 4592. (19) Herr, A. E.; Hatch, A. V.; Throckmorton, D. J.; Tran, H. M.; Brenan, J. S.; Giannobile, W. V.; Singh, A. K. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 5268−5273. (20) Mitrovski, S. M.; Nuzzo, R. G. Lab Chip 2005, 5, 634−645. (21) Kim, J. E.; Cho, J. H.; Paek, S. H. Anal. Chem. 2005, 77, 7901− 7907. (22) Floriano, P. N.; Christodoulides, N.; Romanovicz, D.; Bernard, B.; Simmons, G. W.; Cavell, M.; McDevitt, J. T. Biosens. Bioelectron. 2005, 20, 2079−2088. (23) Cui, T.; Fang, J.; Zheng, A.; Jones, F.; Reppond, A. Sens. Actuators, B 2000, 71, 228−231. (24) Reyes, D. R.; Iossifidis, D.; Auroux, P. A.; Manz, A. Anal. Chem. 2002, 74, 2623−2636. (25) Szekel, L.; Guttman, A. Electrophoresis 2005, 26, 4590−4594. (26) Hatch, A. V.; Herr, A. E.; Throckmorton, D. J.; Brennan, J. S.; Singh, A. K. Anal. Chem. 2006, 78, 4976−4984. (27) Jong, J.; de; Lammertink, R. G. H.; Wessling, M. Lab Chip 2006, 6, 1125−1139. (28) Toh, G. M.; Corcoran, R. C.; Dutta, D. J. Chromatogr., A 2010, 1217, 5004−5011.
ASSOCIATED CONTENT
S Supporting Information *
Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.
■
ACKNOWLEDGMENTS
This research work was supported by start-up funds from the University of Wyoming and a grant from the National Science Foundation (DBI 0964211). N.Y. also acknowledges a graduate assistantship through the Wyoming INBRE program (grant # P20RR016474). The authors would like to thank Dr. James Mecham from the Arthropod Borne Animal Disease Research Laboratory, Laramie, for providing the reagents for the BTV antibody assay.
■
■
■
Article
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest. 7035
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
Analytical Chemistry
Article
(29) Toh, G. M.; Yanagisawa, N.; Corcoran, R. C.; Dutta, D. Microfluid. Nanofluid. 2010, 9, 1135−1141. (30) Wadsworth, C. J.; Yanagisawa, N.; Dutta, D. J. Power Sources 2010, 195, 3636−3639. (31) Mecham, J. O. Am. J. Vet. Res. 1993, 54, 370−372. (32) Hanly, W. C.; Artwohl, J. E.; Bennet, B. T. ILAR J. 1995, 37, 93−118. (33) Mecham, J. O.; Wilson, W. C. J. Clin. Microbiol. 2004, 42, 518− 523. (34) Mecham, J. O.; Dean, V. C.; Wigington, J. G.; Nunamaker, R. A. J. Med. Entomol. 1990, 27, 602−606. (35) Gorris, H. H.; Walt, D. R. J. Am. Chem. Soc. 2009, 131, 6277− 6282. (36) Tsang, V. C. W.; Wilson, B. C.; Peralta, J. M. Method Enzymol. 1983, 92, 391−403. (37) Baud, M. Data analysis, mathematical modeling. In Methods of Immunological Analysis Vol. 1: Fundamentals; Masseyeff, R. F., et al., Eds.; VCH Publishers, Inc.: New York, NY, 1993; pp 656−671. (38) Del Villano, B. C.; Brennan, S.; Brock, P.; Bucher, C.; Liu, V.; McClure, M.; Rake, B.; Space, S.; Westrick, B.; Schoemaker, H.; Zurawski, V. R. Clin. Chem. 1983, 29, 549−552. (39) Howerth, E. W.; Dorminy, M.; Dreesen, D. W.; Spires, E. A.; Stallknecht, D. E. J. Vet. Diagn. Invest. 1995, 7, 393−394.
7036
dx.doi.org/10.1021/ac3011632 | Anal. Chem. 2012, 84, 7029−7036
