FFD immunoassays have been developed with a focus on using sandwich immunoassays for rapid, sensitive, and selective antigen detection. In the most basic form of these assays, an antibody is immobilized on the flow cell surface, the antigen is then captured to the surface, and the bound complex is labeled with a secondary antibody. Antibody conjugated microbeads are then added to the flow cell that recognize the secondary antibody [28]. FFD is performed and the amount of beads bound to the surface quantified. However, for small molecules such as TTX, sandwich immunoassays are not amenable to detection, as it is rarely possible to bind two antibodies to these small analytes. To alleviate the size and antibody constraints, an immunoassay platform can be designed that takes advantage of single antibody binding in an inhibition immunoassay format.

To design the substrate for FFD detection of TTX, modifications were made to a previously optimized biosensor platform (Figure 1a) [25]. A 1 part NH₂-oligoethylene glycol (NH₂-OEG) thiol is mixed with 9 parts OH-OEG thiol to create a substrate that is both bioresistant [30, 31] and amenable to conjugation of TTX [23, 32]. The immobilization reaction is performed by using the crosslinker formaldehyde to covalently attach TTX to the self-assembled monolayer (SAM). By immobilizing TTX on the surface in this manner, the CH₂OH functionality and antibody-reactive portion of the molecule faces up and is available to bind anti-TTX. Based on this orientation then, an inhibition immunoassay can be performed where the mouse anti-TTX is captured and subsequently labeled by anti-mouse microbeads as shown in Figure 1b. Due to the nature of the inhibition immunoassay, the larger the concentration of TTX in the analyte solution, the less free antibodies there will be to bind to the substrate, and consequently the fewer microbeads observed upon readout.

FFD assay instrumentation is shown in Figure 2. All assay steps occur in a microfluidic cell formed when the functionalized surface is mounted in a compression fit cartridge. This custom built platform mounts on an upright microscope and has five individually addressable sensing areas per substrate. Using a peristaltic pump in either push or pull mode, buffer and reagents can be added to and removed from the sensing areas. Following FFD, captured beads are imaged with low power microscopy (5×), and images are captured on a CCD camera. The bench top instrument shown here is a manual workhorse, well designed for assay development; however, in the future it should be feasible to develop an automated, shoe-box sized, portable instrument. Most of the components of this system have already been developed for the compact Bead Array Sensor System (cBASS^®), a field portable bioassay platform employing FFD assays and using magnetoresistive sensors to detect captured magnetic bead labels [33].

After performing the inhibition immunoassay with TTX, the substrate is imaged through a 5× microscope objective, and the image is captured with a CCD camera. To interrogate the quantity of microbeads that are immobilized on the surface, the image is processed using Image J software (Figure 3). The images on the far left represent raw data files where the top image has high coverage (antibody present), and the bottom image has low coverage (no antibody). The images are transformed with the software to yield a pure black and white representation of the surface. By using the software, the percent area covered by dark space is then determined, and this value is representative of microbead binding. From these data, it is seen that at full coverage, the substrate had approximately 73% of the surface covered by microbeads. At the other extreme, when there was no antibody (Ab) present in the assay, the substrate exhibited about 12% coverage. This allows a range of 61% for the immunoassay calibration curve. To determine the linear dynamic range and assay sensitivity, the assay is then performed by mixing various concentrations of TTX with the antibody in the immunoassay format shown in Figure 1b, and the data processing is performed as described herein.

Upon completion of these immunoassays and normalization of the data, the results can be shown in a calibration curve (Figure 4). As expected, when TTX concentration is low there are many antibodies available to bind to the surface, resulting in a larger number of beads and thus a high signal. The inverse is true in which a high TTX concentration consumes the antibodies in solution, and therefore no signal is seen. The curve, ranging from 1 to 100,000 ng/mL TTX, shows a linear trend when the normalized signal is plotted with respect to the log of TTX concentration. This large dynamic range may offer advantages over other techniques as fewer sample dilutions would be necessary to determine an unknown concentration. Furthermore, the sensitivity of the assay, as defined by the IC₂₀ (inhibitory concentration of 20% at 0.8 normalized signal), is 14 ng/mL which compares well to HPLC and LC/MS methods.

One drawback seen in the current data arises from the high standard deviations of each concentration. These relative errors, on the order of 10–15%, could lead to a less quantitative assay. We attribute these errors to nonspecific adsorption of the microparticles to the gold surface, as previous FFD experiments using non-gold surfaces yielded much lower standard deviations [28, 33]. The gold surface, however, is not essential for FFD performance and was only utilized because of the well-established surface chemistry for TTX on gold platforms [25]. Current research is focused on redesigning the substrate conjugation chemistry to eliminate the gold platform and decrease the error while maintaining the specificity, sensitivity and large dynamic range of this assay.

Further experiments indicated an advantage to the TTX immunoassay not previously incorporated into FFD assays. In surface plasmon resonance biosensors, assay platforms can be reconditioned by removing the antibody from the substrate at the end of each experiment [34] which can minimize error, reduce cost per sample, and decrease analysis time. This potential regeneration of substrates was investigated by immersing the used surfaces in a 50 mM NaOH solution and placing on a shaker for 1 hr. After rinsing the slides with water and drying with N_2(g), it is expected that the microbeads and anti-TTX will be removed from the surface leaving a substrate with just TTX immobilized to the SAM. Upon regeneration and running further immunoassays on an individual slide, only minor changes in nonspecific binding (5%) and signals (7.5%) that appeared uniform throughout additional regeneration cycles were observed. This process then also appears to add a source of error in that regeneration was not complete. As such, regeneration was found to be feasible but the optimal conditions, such as using acidic regeneration solutions to avoid potential base sensitivity of TTX, are being investigated as an area of improvement for this immunoassay.

Anti-TTX monoclonal antibody (1 mg lyophilized protein) was purchased from Hawaii Biotech, Inc. (Aiea, HI), and TTX was procured from Sankyo Co, Ltd (Tokyo, Japan). Monobasic potassium phosphate (100mM), dibasic potassium phosphate (100 mM), 37% formaldehyde solution, triethylamine, and glacial acetic acid were obtained from Sigma-Aldrich (St. Louis, MO). HS-(CH₂)₁₁-EG₄-OH and HS-(CH₂)₁₁-EG₆-NH₂ hydrochloride were synthesized by ProChimia Surfaces (Sopot, Poland). Ethanol (absolute, anhydrous, ACS/USP grade) was purchased from Pharmaco AAPER (Shelbyville, KY). 50% w/w sodium hydroxide solution was procured from J. T. Baker (Phillipsburg, NJ). Sheep anti-mouse, 2.8 μm Dynabeads were purchased from Invitrogen (Carlsbad, CA). Gold substrates (146 Å titanium adhesion layer, 3000 Å gold on glass microscope slides) were prepared using the AJA Sputtering Unit at the University of Maryland Nanocenter (College Park, MD). All water used in these experiments was 18.2 MΩ·cm (Millipore MilliQ Academic, Billerica, MA).

A mixed SAM of oligoethylene glycol (OEG) alkanethiol (hydroxyl-terminated and amine-terminated forms) on gold is prepared by modifying a procedure to form TTX surfaces for surface plasmon resonance biosensors [25]. This monolayer allows for covalent linking of tetrodotoxin to the surface (NH₂-OEG) via formaldehyde coupling [23, 32] as well as minimization of nonspecific protein adsorption (OH-OEG) [30, 31]. Gold slides are ozone cleaned (20 min, NovaScan PSD-UV, Ames, IA), rinsed with water and then ethanol, and dried with N_2(g). The dry slides are immersed into the mixtures (described next) for 18 hours. Briefly, 5 mM ethanolic solutions of each thiol are prepared. To generate the SAM with a 1:9 mixture of NH₂-OEG thiol to OH-OEG thiol, 46.5 mL ethanol, 1.5 mL triethylamine, 0.2 mL NH₂-OEG thiol, and 1.8 mL OH-OEG thiol are mixed together. The triethylamine addition to the SAM formation allows for a smoother surface with less unbound thiols [35].

After monolayer formation, the slide is removed from solution, rinsed with ethanol, 10% acetic acid in ethanol, followed by ethanol and dried with N_2(g). This substrate is then reacted with a mixture of 932.6 μL 100 mM pH 7.0 phosphate buffer, 41.8 μL of 10 mg/mL TTX, and 69.6 μL 37% formaldehyde. The formaldehyde functions as a one-carbon crosslinker which is able to covalently bind the TTX to the thiolate surface. Furthermore, this conjugation does not tie up the antibody binding motif located on the CH₂OH side of TTX [25]. The reaction proceeds for 72 hrs at 37 °C, and after formation, the substrates may be stored at 4 °C until use. Upon removal, the chip is rinsed with DI water and dried with N_2(g).

FFD assays were performed with a custom built instrument mounted on an upright microscope. The TTX-functionalized substrate is housed in a compression-sealed fluid cartridge with five individually addressable flow cells [36]. At the front of each flow cell is a 200 μL well used for introduction and exchange of reagents. The opposing side of the cartridge is connected to a buffer reservoir and peristaltic pump. FFD substrate conditioning begins by pumping buffer through the cartridge to the base of the reagent well.

As TTX is a small molecule that is challenging to detect directly, an inhibition immunoassay was employed. One part TTX sample in buffer (1× PBS, 5% Carnation dried skim milk) is mixed with 9 parts 1.1 μg/mL anti-TTX in buffer for 10 min. This inhibition assay mixture is then pulled into the FFD cartridge until the flow cell is filled, and the mixture is exposed to the TTX-functionalized surface for 5 min. During this time, antibodies not occupied with TTX antigen from the sample can bind to the TTX molecules on the substrate. Next, the mixture is pumped out of the cartridge and removed from the reagent well which simultaneously refills the flow cell with buffer. Then, the 2.8 μm diameter beads conjugated with sheep anti-mouse antibodies are pulled into the flow cell and allowed to settle for 3 min. This allows the beads to bind to any surface-bound TTX monoclonal antibodies. Finally, buffer is pumped through the flow cell at a controlled speed to remove nonspecifically attached beads, as previously described in detail [28, 33]. For each substrate, which had five sample flow cells, one cell is used as a positive control (no TTX, Ab), one cell as a blank (no TTX, no Ab), and three cells as immunoassay samples (TTX and Ab).

Beads remaining after the FFD assay are imaged with low power (5×) optical microscopy, and images are captured with a CCD camera. Using Image J software (http://rsbweb.nih.gov/ij/index.html), the images are flattened, thresholded, and digitized to distinguish beads and background. A fixed area (100 × 100 pixels) is selected, and the percent area of beads determined. For each channel, an average of five images was taken, and the coverage of beads determined. The images are then compared to both a blank signal (i.e., no TTX and no Ab) and a no inhibition, positive control signal (i.e., no TTX but Ab in solution) on the same assay substrate. For uniformity between chips, the average sample signal had the nonspecific adsorption, as indicated in the blank run average signal, subtracted. Next, this result was divided by the positive control average signal that is due to just antibody binding. This process yields a normalized signal response. These normalized values for a single concentration with multiple sample runs (as indicated by N) are averaged together and plotted versus original concentration of TTX in the sample prior to antibody dilution.