Anal. Chem. 2010, 82, 4015–4019
Detection of Nitroaromatic Explosives Using a Fluorescent-Labeled Imprinted Polymer R. Cody Stringer,† Shubhra Gangopadhyay,‡ and Sheila A. Grant*,† Department of Biological Engineering and Department of Electrical and Computer Engineering, University of Missouri, Columbia, Missouri 65211 Optical sensors have proven to be a useful method in identifying explosive devices by recognizing vapors of explosive compounds that become airborne and emanate from the device. To detect high explosive compounds such as TNT, a molecularly imprinted polymer (MIP) sensing mechanism was developed. This mechanism consists of MIP microparticles prepared using methacrylic acid as the functional monomer. The MIP microparticles are then combined with fluorescent quantum dots via a simple cross-linking procedure. The result is a highly robust optical sensing scheme that is capable of functioning in an array of environmental conditions. To study the sensing mechanisms’s ability to detect nitroaromatic analytes, the fluorescent-labeled MIP particles were tested for their performance in detecting aqueous 2,4-dinitrotoluene (DNT), a nitroaromatic molecule very similar to TNT, as well as TNT itself. These preliminary data indicate that the system is capable of detecting nitroaromatic compounds in solution with high sensitivity, achieving lower limits of detection of 30.1 and 40.7 µM for DNT and TNT, respectively. The detection mechanism also acted rapidly, with response times as low as 1 min for TNT. Due to the results of this study, it can be concluded that the fluorescent-labeled MIP system is a feasible method for detecting high explosives, with the potential for future use in detecting vapors from explosive devices. The concern to accurately detect explosive devices in an array of environments has led to a myriad of methods for sensing trace amounts of explosive compounds. These methods include gas chromatography (GC), high-pressure liquid chromatography (HPLC), mass spectrometry, infrared absorption spectroscopy, and Raman scattering, just to name a few.1 Of particular interest is the ability to detect the nitroaromatic explosive 2,4,6-trinitrotoluene (TNT). Methods for detection of TNT must not only be capable of trace detection so that airborne explosive vapor is detected but must also be capable of field deployment. Therefore, the sensing system must be robust enough to function in many different environments, must be scalable to relatively small sizes, and must also be inexpensive. Of the sensing schemes available, fluorescence sensors are the most likely candidates to meet these
requirements.2 Fluorescence sensors are especially useful for detection of TNT due to the ability of nitroaromatic molecules to act as electron acceptors, quenching the emission of nearby excited fluorescence species.3 Antibody-based detection methods that utilize fluorescence have shown to be highly sensitive with high specificity, but these systems are limited to the aqueous phase only and are not suitable for vapor detection.4 In addition, antibodies are sensitive to extremes in temperature, which limits the scope of applications to which they can be employed.5 Other fluorescence sensors include those that rely on conjugated polymers or polymer films.6 These schemes are very sensitive to TNT but have no means to ensure analyte-specific fluorescence quenching. Attempts to render conjugated polymers analyte-specific via molecular imprinting have shown promise, but like other systems that rely on organic fluorescent species, the polymers photobleach with repeated or sustained excitation, making continued sensor interrogation difficult.7 Unlike conjugated polymers, molecularly imprinted polymers (MIPs) have an innate specificity for the analyte of interest. MIPs are biomimetic polymers that are imprinted with a template molecule for the purpose of retaining a memory of that template molecule. The template is introduced during solution polymerization and later chemically extracted, leaving three-dimensional template binding sites that are specific to both the shape and chemical functionality of the template.8 Although most applications of MIPs have been limited to separation of mixtures, most notably with capillary electrochromatography (CEC),9-12 MIP-based sensing systems have also been developed.13,14 Many diagnostic applications of MIPs have relied on mass sensing, commonly (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
* To whom correspondence should be addressed. E-mail: grantsa@ mail.missouri.edu. † Department of Biological Engineering. ‡ Department of Electrical and Computer Engineering. (1) Moore, D. S. Rev. Sci. Instrum. 2004, 75, 2499–2512. 10.1021/ac902838c 2010 American Chemical Society Published on Web 04/19/2010
(12) (13) (14)
Germain, M. E.; Knapp, M. J. Chem. Soc. Rev. 2009, 38, 2543–2555. Goodpaster, J. V.; McGuffin, V. L. Anal. Chem. 2001, 73, 2004–2011. Smith, R. G.; D’Souza, N.; Nicklin, S. Analyst 2008, 133, 571–584. Shriver-Lake, L. C.; Breslin, K. A.; Charles, P. T.; Conrad, D. W.; Golden, J. P.; Ligler, F. S. Anal. Chem. 1995, 67, 2431–2435. Yang, J. S.; Swager, T. M. J. Am. Chem. Soc. 1998, 120, 11864–11873. Li, J.; Kendig, C. E.; Nesterov, E. E. J. Am. Chem. Soc. 2007, 129, 15911– 15918. Mosbach, K.; Ramstrom, O. Nat. Biotechnol. 1996, 14, 163–170. Nilsson, J.; Spe´gel, P.; Nilsson, S. J. Chromatogr. B 2004, 804, 3–12. Priego-Capote, F.; Ye, L.; Shakil, S.; Shamsi, S. A.; Nilsson, S. Anal. Chem. 2008, 80, 2881–2887. Schweitz, L.; Andersson, L. I.; Nilsson, S. J. Chromatogr. A 1998, 817, 5– 13. Schweitz, L.; Andersson, L. I.; Nilsson, S. Anal. Chim. Acta 2001, 435, 43–47. Dickert, F. L.; Hayden, O. TrAC Trends Anal. Chem. 1999, 18, 192–199. Yano, K.; Karube, I. TrAC Trends Anal. Chem. 1999, 18, 199–204.
Analytical Chemistry, Vol. 82, No. 10, May 15, 2010
4015
utilizing MIPs applied to a quartz crystal microbalance (QCM).15,16 These systems, however, are quite delicate and are not suitable for field deployable devices. Some of the more pertinent MIPbased sensors that utilize fluorescence are those that make use of quantum dots as the fluorescent species. Quantum dots have distinct advantages over organic fluorescent dyes in that they retain stable fluorescence over time and in an array of environments, thereby making them resistant to photobleaching effects and temperature effects. In addition, the surface of quantum dots may be functionalized with a chemical moiety of interest, allowing them to be integrated easily into analyte sensing systems. The quenching of quantum dot fluorescence by nitroaromatic explosive compounds is also well-documented.17 MIP-based sensors have utilized quantum dots by functionalizing the surface of the quantum dot with 4-vinylpyridine, allowing the quantum dot to polymerize directly into the polymer matrix of the MIP.18,19 The shelf life of quantum dots functionalized with 4-vinylpyridine, though, was very short and the solution turned an opaque black within hours of preparation. Another method for pairing quantum dots with MIP sensing used a prepolymerization step, in which the quantum dot was rapidly polymerized in the presence of the functional monomer for a short time, creating a partially formed oligomer cap that allows the quantum dot to remain dispersed during the remainder of the preparation of the MIP.20 In general, both of these methods demonstrated considerable downsides. In particular, microscopic evaluation showed poor quantum dot dispersion, creating nonuniform fluorescence of the MIP. In an attempt to produce a fluorescent MIP that utilizes quantum dots as the fluorescent species, a novel method for preparing the fluorescent polymer is described. Rather than preparing the quantum dots beforehand and attempting to incorporate them into the polymer matrix, the quantum dots were added as a postprocessing step. The methacrylic acid-based polymer used for fabrication of the MIP in this study presents two side groups: a methyl moiety and a carboxylic acid moiety. Using the carboxylic acid functional group of the polymer, the MIP was labeled with commercially available amine-functionalized quantum dots using a traditional bioconjugation procedure. The resulting fluorescent MIP was utilized as a detection tool for nitroaromatic compounds. EXPERIMENTAL SECTION All chemical reagents and solvents used were purchased from Sigma-Aldrich (St. Louis, MO). Amine-functionalized CdSe quantum dots with a fluorescence emission wavelength of 605 nm were purchased from eBioscience (San Diego, CA). For the fluorescent labeling procedure, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC) and N-hydroxysuccinimide (NHS) were purchased from Pierce Protein Research Products (Rockford, IL). Fluorescence measurements were taken using a Horiba Jobin (15) Bunte, G.; Hu ¨ rttlen, J.; Pontius, H.; Hartlieb, K.; Krause, H. Anal. Chim. Acta 2007, 591, 49–56. (16) Ebarvia, B. S.; Sevilla Iii, F. Sens. Actuators, B 2005, 107, 782–790. (17) Shi, G. H.; Shang, Z. B.; Wang, Y.; Jin, W. J.; Zhang, T. C. Spectrochim. Acta Part A: Mol. Biomol. Spectrosc. 2008, 70, 247–252. (18) Lin, C. I.; Joseph, A. K.; Chang, C. K.; Lee, Y. D. J. Chromatogr. A 2004, 1027, 259–262. (19) Lin, C. I.; Joseph, A. K.; Chang, C. K.; Lee, Y. D. Biosens. Bioelectron. 2004, 20, 127–131. (20) Pang, L.; Shen, Y.; Tetz, K.; Fainman, Y. Opt. Express 2005, 13, 44–49.
4016
Analytical Chemistry, Vol. 82, No. 10, May 15, 2010
Yvon Fluoromax-2 spectrofluorometer (Edison, NJ). Electron microscopy imaging was carried out using a Joel-1400 transmission electron microscope (Tokyo, Japan). Surface area measurements were conducted using a Quantachrome Autosorb gas adsorption system (Boynton Beach, FL). MIP Preparation. The imprinted polymer was prepared by combining 1.0 mmol of methacrylic acid monomer, 4.0 mmol of ethylene glycol dimethacrylate (EGDMA) cross-linker, and 0.025 mmol of template in 50% (v/v) chloroform. To this solution was added 0.05 mmol of 1-hydroxycyclohexyl phenyl ketone photoinitiator. The prepolymerization solution was briefly bubbled with nitrogen, and polymerized under 365 nm UV illumination for 5 min. After polymerization, the bulk polymer was pulverized with a mortar and pestle and sieved to exclude any particles with a diameter above 20 µm using a nylon net filter. Next, the particles were dispersed in methanol and sonicated at 60 °C for approximately 2 h to extract the template, as well as any free monomer, cross-linker, or initiator. The MIP particles were then centrifuged at 2000 rcf for 10 min, and the methanol was removed. The particles were allowed to dry overnight at room temperature. A nonimprinted polymer (NIP) control was synthesized similarly, but without the addition of the template. Fluorescent Labeling. The dried MIP particles were dispersed in phosphate-buffered saline (PBS) and sonicated for approximately 1 h to ensure complete particle dispersion. Once the particles were appropriately dispersed in PBS, 2.0 mM EDC and 7.5 mM NHS were added. In this reaction, the carboxylic acid side groups of the MIP particles were activated to allow binding to available amines. After a 30-min incubation, 0.03 µM amine-functionalized CdSe quantum dots were added. Upon introduction of the quantum dots, a zero-length cross-linking reaction took place between the activated carboxyl groups of the MIP microparticles and the amine-functionalized surface of the quantum dot. The result was a fluorescently labeled MIP microparticle solution. The fluorescent MIP particles were then centrifuged at 10 000 rcf for 10 min and the PBS was removed, along with any unbound quantum dot, EDC, or NHS. The particles were then dispersed in distilled water for use. The NIP particles were fluorescently labeled in the same manner. Template Rebinding. In order to examine the effectiveness of the sensor, the fluorescent MIP particles were re-exposed to the template. The templates examined in this study were DNT and TNT. The fluorescent MIP particles were exposed to an aqueous solution containing the maximum concentration of 0.5 mM template, which is approximately the saturation limit of TNT in water. After introduction, the template bound to the imprinted binding sites of the MIP. This decreased the distance between the electron-accepting template and the quantum dots conjugated to the MIP particles, causing fluorescence quenching of the quantum dots. The quantum dot fluorescence spectrum was monitored both before and after introduction of the template. For dose-response experiments, the aqueous template concentration ranged from 0 to 0.5 mM and the quantum dot fluorescence spectra were captured after an exposure time of 10 min. The data was analyzed by comparing the peak fluorescence intensity of each solution and plotting these intensity values with respect to the concentration of template or duration of exposure.
RESULTS AND DISCUSSION MIP Particle Characterization. An important aspect of the characteristics of the MIP particles, for both fluorescent labeling as well as template extraction and rebinding, is surface area. Surface area analysis was conducted for both the MIP particles and the NIP negative control particles. The surface area value for the MIP particles was determined to be 10.59 m2/g, while the value for the NIP particles was determined to be 178.92 m2/g. It was initially thought that the imprinted binding sites of the MIP would enhance the porosity of the microparticles. However, surface area analysis of the MIP as compared to the NIP, which was synthesized without the presence of the template, illustrates that the opposite is true. The decreased surface area of the microparticles with imprinted binding sites is likely due to the strong UV absorption of the template. Because UV light is being absorbed by the template during polymerization, the cross-linking reaction will proceed slowly, decreasing the porosity. After fluorescent labeling of the MIP particles was complete, TEM images of the microparticles were taken to ensure sufficient quantum dot dispersion and degree of labeling. The aqueous MIP particles were drop-cast onto a standard TEM grid and allowed to dry. The grids were then inserted into the microscope and imaged. Quantum dot labeling was analyzed by comparing the particles before and after the bioconjugation procedure. Both samples are shown in Figure 1. The quantum dots are clearly observable as small dots distributed within the polymer matrix of the MIP particle in Figure 1a, while the particles lack these features before quantum dot labeling, as shown in Figure 1b. The TEM images also do not seem to show a higher density of quantum dots around the exterior, indicating consistent dispersion of the nanocrystals throughout the polymer matrix of the MIP particles. MIP Solvent Selection. An important factor determining the efficiency of molecular imprinting is the solvent selection. The polymerization solvent facilitates the molecular interactions between the dissolved template and monomer and is a primary determinant of the strength of these interactions. A limited number of solvents are available due to the characteristics that are desired for the polymerization reaction. Most importantly, the solvent must readily dissolve all reagents, including monomer, cross-linker, template, and initiator. Second, the solvent should be a poor solvent for the cross-linked polymer so that a porous matrix is created, allowing the template to be readily extracted. Solvents that meet these criteria include acetonitrile, chloroform, toluene, and 2,4-dimethylformamide (DMF). To determine the most appropriate solvent for imprinting of nitroaromatic template molecules, three of these solvents were examined: chloroform, acetonitrile, and toluene. DMF was not considered due to a color change reaction that occurs when mixed with nitroaromtic compounds in high concentrations. The labeled MIP microparticles were exposed to a saturated concentration of DNT in water and the peak fluorescence intensity was observed both before and after template introduction. Additionally, these results were compared to a nonimprinted polymer control that was prepared using the same solvents. The percent change in fluorescence intensity values, or degree of quenching values, are shown in Figure 2.
Figure 1. TEM images of the imprinted polymer microparticles labeled with quantum dots (a) and before quantum dot labeling (b).
Figure 2. Degree of quenching of fluorescent-labeled molecularly imprinted polymer (red bars) compared to nonimprinted negative control (blue bars) prepared using various solvents.
The largest decrease in fluorescence occurred when the imprinted polymer was prepared using chloroform as the solvent. Interestingly, this corresponded to the smallest decrease in Analytical Chemistry, Vol. 82, No. 10, May 15, 2010
4017
Figure 3. Time response of fluorescent-labeled molecularly imprinted polymer (red squares) versus nonimprinted negative control (blue triangles) after re-exposure to (a) DNT template and (b) TNT template.
fluorescence for the nonimprinted polymer control. This indicates that the imprinted polymer synthesized using chloroform as the polymerization solvent has the highest imprinting efficiency and is therefore the most suitable solvent for rebinding of nitroaromatic templates. Chloroform was utilized as the solvent for preparation of the imprinted and nonimprinted polymers in subsequent studies. Time Response of Fluorescent MIP Sensor. By exposing the labeled MIP sensor solution to a high concentration of template and examining the quench in fluorescence over time, the time response of the sensor was determined. After addition of 0.5 mM DNT to the MIP sensor, an immediate decrease in fluorescence was observed, with an exponential trend over time, as shown in Figure 3a. This is in contrast to the nonimprinted polymer control, which produced a much smaller initial decrease in fluorescence intensity, which remained constant over time, indicating a lack of binding of the template to the polymer microparticles. After only 1 min of exposure, the p value of the means of the imprinted and nonimprinted polymers is 0.0121, showing a statistical significance between the sensor and the negative control with a confidence interval of 95%. At 5 min after initial exposure to the DNT template, the p value increases to 0.0916 due to the slightly larger error in these samples. At all subsequent time points, however, the differences between means of the DNT imprinted and nonimprinted polymers are significant. Because of the irregularly large error at a time of 5 min after addition of DNT, the response time of the sensor can be considered to be 10 min. The time response of the fluorescent MIP sensor when TNT is used as the template is shown in Figure 3b. After addition of TNT to the solution containing the MIP sensor, a time-response curve similar to that of the DNT-imprinted polymer is observed. This data is then compared to a nonimprinted polymer control, which showed negligible fluorescence quenching after addition of TNT. The means of the imprinted polymer sensor and the 4018
Analytical Chemistry, Vol. 82, No. 10, May 15, 2010
Figure 4. Dose response of fluorescent-labeled molecularly imprinted polymer (red squares) versus nonimprinted negative control (blue triangles) when exposed to various concentrations of (a) DNT template and (b) TNT template.
nonimprinted polymer control were compared at the smallest time value, 1 min. At this point, the p value of the two means is 0.0129 with a confidence interval of 95%. The differences between the MIP sensor quenching and negative control quenching for all subsequent time points are also statistically significant, so the response time of the TNT-imprinted sensor can be considered 1 min. However, no time values below 1 min were studied, so it is possible that the response time is lower than 1 min. DoseResponseofFluorescentMIPSensor.Thedose-response curve of the MIP sensor provides valuable information on the sensitivity and lower limit of detection of the sensor. To ascertain this information, the MIP sensor was exposed to concentration of template ranging from 0 to 0.5 mM and measurements were taken at 10 min. The fluorescence of the DNT imprinted MIP, shown in Figure 4a, decreases notably when exposed to increasing concentrations of template. This quenching effect continues up to a saturation point, above which very little further fluorescence quenching is observed. The decrease in fluorescence intensity of the nonimprinted negative control decreases nearly negligibly with increasing concentrations of DNT, illustrating that binding of the MIP to the template is the cause of the quenching response and not random interactions in the aqueous solution. The maximum decrease in fluorescence intensity of the nonimprinted negative control was used as the benchmark for determining the lower limit of detection of the MIP sensor. To do this, the lowest fluorescence intensity value of the nonimprinted control was applied to the nonlinear regression curve of the imprinted polymer sensor. The DNT concentration was calculated in this manner to be 30.1 µM, which may be considered the lower limit of detection of the fluorescent MIP sensor. This study was repeated with the TNT-imprinted polymer, with the results shown in Figure 4b. Again, a relatively large quenching response was observed for the MIP sensor as various concentra-
tions of TNT were added, while negligible response was observed for the nonimprinted polymer negative control. Using the largest quenching response of the negative control as the benchmark for determining the lower limit of detection of the TNT imprinted polymer, the lowest detectable concentration of TNT was calculated to be 40.7 µM. While this value is comparable to the DNTimprinted polymer sensor, it was expected that the quenching response to TNT would be greater due to the increased electronaccepting properties of the TNT. This irregularity can be accounted for by slight variations in imprinting efficiency between batches of polymer particles. Although the TNT should theoretically produce a stronger quenching response, less of it is being bound by the MIP particles due to a lower imprinting efficiency as compared to the DNT-imprinted polymer. Regardless, the data indicates that the sensor is capable of not only detecting the presence of either nitroaromatic compound but is able to do so in a concentration-dependent manner. The preliminary results for the fluorescent-labeled MIP sensor are indicative of a suitable system for detection of nitroaromatic compounds. The performance of the system illustrates its potential use as a field-deployable sensor for explosives detection. Although the lower limit of detection of the sensor is currently too high for trace detection of these compounds, optimization of the microparticle size would increase the limit of detection. Additionally, it is important to note that the sensitivity of the sensor is relatively high, so the lower limit of detection could be decreased dramatically by lowering the volume of the sensor solution. Such a move would not only likely result in an appropriate lower limit of detection for trace diagnostics but would lead to downscaling of the aqueous sensor solution volume to allow ease of miniaturization and integration into an appropriate sensing platform. Another noteworthy aspect of the results throughout the studies of the fluorescent MIP sensor is that the MIP quenching response was compared with a nonimprinted polymer that acted as a negative control. This negative control was exposed to the nitroaromatic analytes in an identical manner as the sensor solution. While observable and distinct decreases in fluorescence occurred with the MIP sensor solution, negligible quenching was observed with the nonimprinted control. What little quenching effect that was (21) Thomas, S. W.; Joly, G. D.; Swager, T. M. Chem. Rev. 2007, 107, 1339– 1386.
observed with the negative control can be attributed to random interactions of the nitroaromatic analyte with the fluorescent nonimprinted polymer microparticles in the aqueous solution. This, in turn, illustrates that binding interactions between the template and fluorescent MIP were responsible for the observed quenching response. In can be concluded, then, that the proposed fluorescent MIP-based sensor scheme is a potentially viable method for future use in detection of high explosives. CONCLUSIONS A fluorescent-labeled imprinted polymer sensing scheme was developed to detect aqueous concentrations of TNT and DNT. The sensor had notable sensitivity, achieving a lower limit of detection of 30.1 µM for DNT and 40.7 µM for TNT. These concentrations roughly correspond to approximately 0.5-1 ppm. However, in the case of conjugated polymer-based sensors for nitroaromatic explosives, which are a now commercially available method for explosives detection, lower limits of detection in the range of 1-5 ppb and even lower have been achieved.21 These detection limits are significantly lower than those obtained from the quantum dot-coupled MIP studied herein, and it is clear that further optimization is required to enhance the sensitivity of the proposed sensing scheme. On the other hand, the fluorescent MIP method showed a rapid response time, obtaining a statistically significant quenching response as quickly as 1 min. Future studies will focus on optimization and further development of the insolution fluorescence sensor based on MIP microparticles. This work will be particularly focused on increasing performance and potential miniaturization and integration of the sensor. ACKNOWLEDGMENT The authors thank Dr. David B. Henthorn at the Missouri University of Science and Technology for his help and support in this project, the University of Missouri Electron Microscopy Core, and Dr. Rajagopalan Thiruvengadathan of the University of Missouri Remote Testing Facility.
Received for review December 14, 2009. Accepted April 2, 2010. AC902838C
Analytical Chemistry, Vol. 82, No. 10, May 15, 2010
4019
