Highly Sensitive and Rapid Detection of ... - ACS Publications

Dec 28, 2010 - Ivan Maguire , Jenny Fitzgerald , Daniel McPartlin , Brendan Heery , Caroline ... International Journal of Thermal Sciences 2014 79, 60...
0 downloads 0 Views 3MB Size
ARTICLE pubs.acs.org/est

Highly Sensitive and Rapid Detection of Microcystin-LR in Source and Finished Water Samples Using Cantilever Sensors Yanjun Ding and Raj Mutharasan* Department of Chemical and Biological Engineering, Drexel University, Philadelphia, Pennsylvania 19104, United States ABSTRACT: Microcystin-leucine-arginine (MCLR) is one of the toxic microcystin congeners produced by the common cyanobacteria, blue-green algae. A piezoelectric-excited millimeter-sized cantilever (PEMC) sensor was developed for the sensitive detection of MCLR in a flow format using both monoclonal and polyclonal antibodies that bind specifically to MCLR. PEMC is a resonant cantilever sensor whose resonant frequency decreases as target analyte binds to its surface. Monoclonal antibody against MCLR was immobilized on the sensor surface via amine coupling. As the toxin in the sample water bound to the antibody, resonant frequency decreased proportional to toxin concentration. Three water matrices, namely buffer, tap water, and river water, were spiked with MCLR standards and were successfully detected in the dynamic range of 1 pg/mL to 100 ng/mL (effective concentration -250 fg/mL to 25 ng/mL). The sensor response was characterized by a log-linear relationship between resonant frequency change and MCLR concentration. Positive verification of MCLR detection was confirmed by a sandwich binding on the sensor with a second antibody binding to MCLR on the sensor (attached in first detection step) which caused a further resonant frequency decrease. We show for the first time that MCLR in various water samples can be detected at 1 pg/mL.

1. INTRODUCTION Microcystins (MCs) are a family of stable heptapeptides (95% HPLC purity) were provided by Abraxis (Warminster, PA). Cysteamine, glutaraldehyde (GLU), and bovine serum albumin (BSA) were purchased from Sigma-Aldrich (Alletown, PA). Phosphate buffered saline (PBS, 10 mM, pH 7.4) was prepared in deionized water (18 MΩ, Milli-Q system, Millipore). All other chemicals used in the experiment were of analytical grade and were purchased from Sigma. 2.2. PEMC Sensor Fabrication and Functionalization. Details of PEMC fabrication can be found in previous publications.22 Briefly, the sensor consists of a PZT layer (Piezo Systems, Woburn, MA) and a glass layer (SPI, West Chester, PA), bonded by a nonconductive adhesive. The free-end dimensions of PZT and glass were 1.2  1.0  0.127 mm3 and 3.0  1.0  0.16 mm3 (L  W  T), respectively. See Figure 1B. The top and bottom surfaces PZT was connected to excitation voltage source. The cantilever was anchored in a 6 mm glass tube by a nonconductive epoxy. Subsequently the sensor was spin-coated with polyurethane for providing electrical insulation. Both sides of glass (0.9 mm) were sputtered with 100 nm gold (99.9%) in a Denton Desk Sputtering System (Denton Vacuum, New Jersey). The gold sensing area was about 1.8 ( 0.1 mm2. For surface functionalization, the freshly gold-coated sensor was first immersed in 10 mM cysteamine for 1 h at 37 C, followed by incubation in 2.5% GLU for 1 h at 37 C. The sulfur in cysteamine binds to gold Æ111æ sites whose bond energy is considered equivalent to covalent bond and is stable in nonoxidizing environment.26 The two aldehyde groups in GLU react with the amine groups on cysteamine and antibody forming a covalent link between the two. The sensor was rinsed with DI water and incubated in monoclonal anti-MCs solution (17 μg/mL) for 1 h at 37 C. After rinsing with PBS and DI, the sensor was exposed to 10 mg/mL BSA for blocking the nonspecific binding sites. After a second rinsing step with DI and PBS, the sensor was installed in the sample flow cell for the detection experiments. All of the tests were done in the flow apparatus, a schematic of which is in Figure 1. 2.3. Experimental Apparatus and Methods. A schematic of the experiment setup and experimental apparatus was described previously.22 It consists of an impedance analyzer (HP 4192A or

Figure 1. A. Spectra of PEMC sensor in the 50-120 kHz region in air and in DI water. Resonant frequency in air was 102 kHz (Q = 34) and decreased to 66.5 kHz in DI water (Q = 33). The sensor was excited at 100 mV. Phase angle shown is the angle between the excitation voltage and the resulting current in the PZT layer. Sharpe change in phase angle occurs at resonance. B. A picture of PEMC sensor. The sensor width is 1.0 mm, overall length is 3.0 mm, and the PZT is 1.2 mm. C. Schematic of Flow apparatus. All tubings are 3 mm, and the flow rate of running buffer or the introduced sample water was maintained by the peristaltic pump. The fluid from the sample reservoirs is connected to the inlet of the pump via a 4-port manifold connector. The incubator was maintained at 30 C, and a custom written software managed the impedance analyzer connected to the sensor. By opening one of the valves (V1 V4) the desired reagent or buffer is admitted to the sensor flow cell.

Agilent, HP4294A), a peristaltic pump, a homemade flow cell, and several fluid reservoirs. The flow cell has a hold-up volume of 120 μL and was maintained in an incubator at 30 ( 0.1 C to ensure isothermal conditions. Prior to an experiment, the entire flow system was rinsed with 100% ethanol followed by copious amount of DI water. The resonant frequency change during an experiment was monitored continuously by a custom-written LabView program. The flow loop was operated in either a single pass mode or in a recirculation mode. 1491

dx.doi.org/10.1021/es1020795 |Environ. Sci. Technol. 2011, 45, 1490–1496

Environmental Science & Technology In a typical detection experiment, each of the functionalized sensors was vertically installed in the flow cell, and the flow rate was set at 1.0 mL/min. For detection in PBS (or tap water or Schuylkill river water), the flow loop was filled with PBS (or tap water or Schuylkill river water) and the resonant frequency was allowed to reach steady state. Subsequently, samples spiked with MCLR were switched into the flow loop in sequence while the frequency changes were monitored continuously. To confirm that the sensor response was due to MCLR binding, control experiments including both positive and negative control samples were carried out at the same temperature and flow rate. The positive control response is the response of a clean PEMC sensor (without anti-MCs immobilized) exposed to 1 pg/mL MCLR. Negative control response is the response of anti-MCs immobilized PEMC sensor exposed to a 1 mL sample of MCLR-free buffer or water. 2.4. Sensor Mass Calibration. The sensor was mounted in a 30 C dry incubator, and the sensor's resonant frequency was measured continuously. A 0.2 μL droplet of aqueous glycerol solution (780 pg/mL) was dispensed on the sensor surface. The frequency decreased immediately due to the mass loading. Subsequently, it rose as the water evaporated. Glycerol being nonvolatile remained on the sensor surface after water evaporated, leaving the sensor with an added mass of 156 fg glycerol. Several similar mass additions were made, and the measured resonant frequency changes were plotted against mass additions to obtain mass-change sensitivity value. The resonant mode at ∼102 kHz exhibited sensitivity of 3 ( 0.5  10-15 g/Hz in air (n = 5 data points). Sensitivity in liquid is not significantly compromised and was previously shown to be lower compared to the sensitivity in air by a factor of 2 to 3.27 This is because the sensor Reynolds number at 102 kHz is high (∼105), and the response depends on mass change. 2.5. Water Sample Experiments. Water samples tested included tap water obtained in our laboratory (supplied by Philadelphia Water Supply Company) and river water (Schuylkill River, collected near 30th and Market Sts., Philadelphia, PA). The Schuylkill is the source water for about a third of Philadelphia city. Water samples were spiked with MCLR standard to final concentrations in the range of 1 pg/mL and 100 ng/mL, which brackets WHO target of 1 ng/mL. The flow loop of the experimental apparatus was filled with the appropriate water matrix and the sensor was prestabilized, and then various MCLR spiked samples were introduced. The flow loop was set in recirculation mode. The resonant frequency of the sensor was recorded at a rate of two measurements per minute and later analyzed for frequency shifts.

3. RESULTS AND DISCUSSION 3.1. Characterization of PEMC Sensor. Several PEMC sensors (n > 30) were fabricated, and their resonant frequency spectra in air and in DI water were determined. Although there were small variations in spectral properties due to dimensional differences in the individually and manually made sensors, the selected sensors for detection had similar spectral properties and exhibited two resonant modes in 1-150 kHz. The first observed mode is the fundamental mode, and its peak was relatively small in DI water. The second mode in air and in DI water were found to be significant and stable and are illustrated in Figure 1A. The resonant frequency decreased from 102 kHz in air to 66.5 kHz in DI water due to density of the surrounding medium. The change

ARTICLE

in resonant frequency gives a measure of mass-change sensitivity because the oscillating sensor carries with it the liquid that is in close adjacency to sensor surface. In Figure 1B a picture of a PEMC sensor whose PZT length is 1.2 mm and glass is 3.0 mm is shown. Both PZT and glass were anchored in epoxy. The quality factor (Q) of the resonant peak is one of the key performance parameters of the sensor. Q is a measure of peak sharpness and is the ratio of resonance frequency to the width frequency at half phase angle peak height. The second mode used in this study exhibited Q of 34 in air and decreased slightly to 33 in DI water. Such a small decrease indicates that viscous damping at this frequency is small. In liquid, resonant frequency change of PEMC sensor is directly attributable to mass change because the Reynolds number of PEMC at ∼100 kHz is high (∼105), indicating negligible viscous effects. The baseline phase angle is -89 suggesting that the sensor behaved electrically as a capacitor at nonresonant frequencies. Because the second resonant mode was relatively noise-free and exhibited excellent masschange sensitivity, it was selected for detection experiments. Once the resonant mode was found to be useful for sensing, mass-change sensitivity was experimentally determined by adding 0.2 μL droplet of dilute aqueous glycerol sample to the cantilever tip and then observing the resulting steady state change in resonant frequency upon evaporation of water. Typically, upon addition of 0.2 μL sample, the resonant frequency decreased by ∼10 kHz and recovered as the water evaporated from the cantilever tip. Once the resonant frequency reached a constant value and was lower than the initial value, we attributed the decrease to the added nonvolatile glycerol mass. The glycerol residue formed a film on the glass part of the sensor tip with an area of ∼1 mm2. For a glycerol addition of 156 fg, resonant frequency shift in air was 52 ( 5 Hz (n = 4). The resonant frequency changes plotted against various mass additions yielded a straight line whose slope is termed mass change sensitivity in air. For the sensors used in the current study, mass change sensitivity were in the range of 0.4 to 4  10-15 g/Hz in air. The range of measured sensitivity was large for the entire set of sensors, but the selected sensors for detection were of a much narrower range and had a sensitivity of 3 ( 0.5  10-15 g/Hz (n = 5 sensors and m = 7 tests). Each experiment reported in this paper is an average of several experiments, typically 3 or 5. 3.2. Response to MCLR in Three Water Media. To test and compare the performance of the PEMC sensor detection responses in matrices that contain contaminants, detection experiments in PBS, tap water, and river water were carried out using functionalized sensors. Typical responses obtained in the three different water media are shown in Figure 2. In these experiments, an antibody was immobilized via amine coupling. As shown in Figure 2A, after a sensor was prepared with anti-MCs and blocked with BSA, it was first stabilized in PBS, and then 1 mL of various MCLR in PBS samples was introduced into the flow loop sequentially. The flow loop was set in recirculation mode to allow sufficient time for binding as indicated by reaching steady state value for resonant frequency. The total volume of the flow circuit was 4 mL, thus the effective concentration that sensor was exposed to was diluted by a factor of 4. Resonant frequency responses to sequential additions of 1 pg/mL, 10 pg/mL, and 1 ng/mL in PBS were 75, 43, and 50 Hz, respectively. The total frequency response for the cumulative addition of 1.011 ng MCLR was 168 Hz. One notes in Figure 2A, the sensor showed a rapid and large resonant frequency decrease upon the addition of the first 1 pg/mL MCLR sample followed by a slightly smaller 1492

dx.doi.org/10.1021/es1020795 |Environ. Sci. Technol. 2011, 45, 1490–1496

Environmental Science & Technology

ARTICLE

Figure 3. Resonant frequency responses to various concentrations of MCLR in PBS, tap water, and river water. The responses in three media show log-linear correlation with the MCLR concentration. The data given above were from repeated experiments (n = 2-5) and used of multiple sensors (n = 18).

Figure 2. Sequential detection of MCLR in PBS (A), tap water (B), and river water (C). For detection in PBS, 1 mL sample at 1 pg/mL, 10 pg/mL, and 1 ng/mL of MCLR were introduced into the flow loop sequentially. For detection in tap water, a 1 mL sample prepared in tap water at 1 pg/mL, 10 pg/mL, and 1 ng/mL of MCLR were injected sequentially. For detection in river water, a 1 mL sample prepared in river water at 1 pg/mL and 10 pg/mL of MCLR were sequentially introduced into the flow system. The control shown is the response of a bare PEMC sensor exposed to 1 ng/mL MCLR, and the sensor response was less than 8 Hz. A 10-min PBS or water rinsing step was introduced between each sample injection and is not shown in the graph. The total volume in the flow circuit was 4 mL. Thus each sample concentration was diluted by a factor of 4 in the flow circuit.

response to subsequent samples. From earlier work we know that sensor response is log-linear suggesting that response to higher concentration is not linearly proportional to concentration.24 The number of binding sites on the sensor is estimated at ∼1011, assuming full surface (1.8 mm2 gold area) coverage with antibody of average size 4-5 nm. If each antibody binds to two MCLR

molecules, the maximum mass of MCLR that can bind to the sensor under the most optimistic condition is ∼0.3 ng. Similar sequential addition experiments were individually conducted in tap water and in river water by serially introducing various concentrations of MCLR and are shown in Figure 2B and Figure 2C, respectively. After the functionalized sensor was stabilized in tap water or river water, various MCLR samples spiked in tap water or in river water were injected into the flow loop (1.0 mL/min) in recirculation mode. Resonant frequency responses (Figure 2B) to sequential additions of 1 pg/mL, 10 pg/mL, and 1 ng/mL MCLR in tap water were 20, 12, and 19 Hz, respectively. Note that signal-to-noise ratios in these responses were >12. Similarly, in river water for sequential additions of 1 pg/mL and 10 pg/mL MCLR resulted in decreases of 48 and 28 Hz, respectively (Figure 2C). Following MCLR attachment, the flow loop was rinsed with MCLR-free tap water or river water, and no further change in resonant frequency occurred. The control experiments shown in Figure 2 are responses of a bare sensor (without antibody attached to it) exposed to 1 ng/ mL MCLR and yielded a response less than 8 Hz and are within measurement noise. The second control experiment shown was conducted with a functionalized sensor exposed to MCLR-free river water and tap water which gave a response of 2 ( 5 Hz which is at noise level (data not shown). Based on the results in Figure 2, we conclude that sensor response is proportional in a nonlinear fashion to the amount of MCLR in the sample and that nonspecific adsorption to the sensor was very small in the three water matrices tested. 3.3. Estimates of Detection Limit and Dynamic Range. Sensor responses to MCLR spiked in the three water matrices (PBS, tap and river water) are collectively presented in a semilogarithmic graph in Figure 3. The experimental data obtained in this study correlated well with the following equation ð- Δ f Þ ¼ AlogðCÞ þ B

ð1Þ

where A and B are sensor constants and depend on cantilever geometry, antibody surface concentration, and antibody-antigen immuno-binding constant. The (-Δf) is the total steady state frequency shift, and C is MCLR concentration added to the flow loop. In Figure 3 we present the results obtained using various 1493

dx.doi.org/10.1021/es1020795 |Environ. Sci. Technol. 2011, 45, 1490–1496

Environmental Science & Technology sensors (n = 18) that exhibited similar spectral property and mass-change sensitivity. From Figure 3, one can see that the functionalized PEMC sensors show good detection sensitivity in the three water media. The detection range was 1 pg/mL to 100 ng/mL, and the fitted logarithmic regression relationships are (-Δf) = 18.94 log (C) þ 605.8 in PBS (R2 = 0.986), (-Δf) = 8.07 log (C) þ 240.4 in tap water (R2 = 0.989), and (-Δf) = 13.03 log (C) þ 404.7 in river water (R2 = 0.901), respectively. As can be seen in Figure 3, the sensor frequency response magnitude was lower in tap water than in PBS for the same MCLR concentration. The reason for this was not investigated in the current study, but we believe that the difference may be due to lower binding property of the antibody in a nonbuffered medium. Additionally residual chlorine or other disinfectants might affect the binding of the MCLR with the antibody. For detection of 1 pg/mL MCLR in river water, the sensor produced a measurable response. In the four experiments at 1 pg/mL the sensor responses were 48 Hz, 36 Hz, 52, and 43 Hz, which gave an average of 45 ( 9 Hz. The signal-to-noise ratio was 18 to 26 as noise level was (2 Hz. Thus we find that a low concentration of 1 pg/mL MCLR (effective concentration of 250 fg/mL) in river water can be successfully measured. The results suggest that the sensitivity of the PEMC sensor is not significantly comprised in source and finished water samples. 3.4. Confirmation of Detection Using Sandwich Assay. One method of confirming that the sensor response is indeed caused by the antigen (MCLR) binding to the antibody on the sensor is to determine whether a second antibody binding to the bound MCLR will produce a further sensor response. This approach is similar to the sandwich method used in immuno-assays such as ELISA except that the second antibody is not labeled and the sensor itself produces the detection signal due to increased mass on the sensor. A monoclonal antibody against MCLR became available from a commercial vendor during the course of the current study. Since MCLR is a small molecule, it seemed reasonable to immobilize the mAb on sensor and use pAb in the second binding step. The rationale is that MCLR bound to mAb is more likely to leave open antigenic sites for pAb than the other way around. We know from experiments that both antibodies bind to MCLR, but their binding sites are unknown. Therefore, monoclonal anti-MC immobilized on the sensor was used as a capture antibody, and the polyclonal rabbit anti-MC served as the indicator antibody. A typical secondary antibody binding response in PBS is shown in Figure 4A. We first introduce mAb (17 μg/mL, 1 mL) into the flow loop with the sensor prepared with cysteamine and glutaraldehyde. The reaction of mAb with the exposed aldehyde group occurs rapidly which was measured by sensor's resonant frequency decrease (55 Hz). After the monoclonal antibody-functionalized sensor was blocked with BSA and stabilized in PBS, 1 mL of 1 pg/mL MCLR sample was introduced into the flow loop in a recirculation mode. A 50 Hz resonant frequency decrease was observed in 60 min. After PBS rinse of the flow loop, 1 mL of polyclonal anti-MC (19.0 μg/mL) was introduced into the flow loop in a recirculation mode. A further decrease of 40 Hz occurred in 70 min. The decrease was due to the secondary antibody binding and indicated that initial response observed during detection was indeed due to MCLR binding to the sensor. In separate experiments, monoclonal antiMC immobilized sensor was exposed to 1 mL of polyclonal antiMC (19.0 μg/mL), and the response was within a noise level of 6 Hz and served to indicate that the response to secondary antibody was due to binding to the antigen, MCLR and not to the immobilized mAb.

ARTICLE

Figure 4. Confirmation of MCLR detection using polyclonal anti-MCs as a secondary antibody in PBS (A), tap water (B), and river water (C). Sensor was first functionalized with monoclonal anti-MCs and then blocked with BSA. After the detection response to MCLR at 1 pg/mL in PBS, 1 pg/mL in tap water and 100 pg/mL in river water, respectively, it was then exposed to 1 mL of 19 μg/mL polyclonal anti-MCs. The sensor showed a further resonant frequency decrease of 40 Hz in PBS, 20 Hz in tap water and 30 Hz in river water, respectively. In the control experiment, 19 μg/mL polyclonal anti-MCs was injected to monoclonal anti-MCs immobilized sensor, and no significant response occurred and the result was not added in the figure. A 10-min PBS or water rinsing step was used between each injection and is not labeled in the graph.

Similar confirmation experiments were carried out in tap water (Figure 4B) and in river water (Figure 4C). After the monoclonal 1494

dx.doi.org/10.1021/es1020795 |Environ. Sci. Technol. 2011, 45, 1490–1496

Environmental Science & Technology

ARTICLE

water (Figure 5). One of the reasons for slower binding rate in water could be due to poor buffering of tap water. Complex medium impedes both the accessibility of antibody binding site and the transport of MCLR to the sensor surface. The other reason for a slower binding rate in tap water could be the chorine in tap water that can potentially reduce the immuno-reactivity of the MCLR. Moreover, comparing the reaction Kobs value of MCLR with other big cells such as Escherichia coli O 157:H7 that has been previously investigated to be larger than 0.1 min-1,29 the Kobs value of MCLR that obtained here is much lower. This confirms the fact that MCLR is a kind of small biomolecular and the immuno-reactive site on its surface is not as much as that of big cells.

’ AUTHOR INFORMATION Corresponding Author Figure 5. Kinetic analysis for the detection of 1 pg/mL MCLR in PBS, tap water and river water. The slope of each line gives the observed binding rate constant, Kobs. See eq 2.

antibody-functionalization step which yielded a frequency decrease of 90 Hz, the sensor was blocked with BSA. Subsequently, the flow cell was rinsed with tap water or river water which caused the resonant frequency increase due to the lower density of water (0.998 g/cm3) compared to 10 mM PBS (1.008 g/cm3). The response to BSA blocking depends on available reactive sites on the sensor. In Figure 4A the response was 10 Hz, while in the other two cases it was negligible. We believe the variability occurs due to how densely antibody occupies on the sensor surface. The range of response to BSA blocking varied from 0 to 10 Hz. Sample (one mL) containing 1 pg/mL MCLR prepared in tap water or 100 pg/mL MCLR prepared in river water was introduced into the flow loop, and the flow was set in recirculation mode to allow sufficient time for binding reaction to occur. The attachment of MCLR in tap water and in river water resulted in a frequency decrease of 20 and 75 Hz, respectively. A second antibody introduction (pAb, one mL of 19 μg/mL) caused a further decrease of 20 and 30 Hz, for the two cases. Similar experiments were conducted in the concentration range of 1 pg/mL to 1 ng/mL in the two waters. These results indicate that tap water and river water offer a similar favorable secondary binding environment as PBS. In practical cases, one would prefer to conduct the secondary binding under favorable condition such as in PBS for maximizing sensor response. 3.5. Kinetic Analysis of Sensor Response. Comparing the detection responses in Figure 2, one notes that the resonant frequency response for the same amount of MCLR (1 pg) in PBS was ∼3 times greater than the response obtained in tap water and ∼1.5 times greater than the response obtained in river water. The sensor binding response in Figure 2 can be analyzed using Langmuir kinetic model28, given by ! Δf ¥ - Δf ln ð2Þ ¼ - Kobs τ Δf ¥ where (-Δf) is the change in resonant frequency at time τ, Δf¥ is the total steady state resonant frequency shift, and Kobs is the observed binding rate constant. Binding rate constant was obtained by fitting a straight line to experimental data during the first 8 min of the sensor responses and was found as 0.0279 min-1 in PBS, 0.0215 min-1 in river water, and 0.0085 min-1 in tap

*Phone: (215)895-2236; fax: (215)895-5837; e-mail: mutharasan@ drexel.edu.

’ ACKNOWLEDGMENT The authors thank Environmental Protection Agency STAR Grant R833829 for the financial support. The authors thank Fernando Rubio (Abraxis) for providing purified antibody and for helpful discussions on MCLR. ’ REFERENCES (1) Carmichael, W. W. The Cyanotoxins. In Advances in Botanical Research; Academic Press: 1997; Vol. Vol. 27, pp 211-256. (2) Lawton, L. A.; Chambers, H.; Edwards, C.; Nwaopara, A. A.; Healy, M. Rapid detection of microcystins in cells and water. Toxicon 2010, 55 (5), 973–978. (3) Hernandez, J. M.; Lopez-Rodas, V.; Costas, E. Microcystins from tap water could be a risk factor for liver and colorectal cancer: A risk intensified by global change. Med. Hypotheses 2009, 72 (5), 539–540. (4) Li, X.-Y.; Wang, J.; Liang, J.-B.; Liu, Y.-D. Toxicity of microcystins in the isolated hepatocytes of common carp (Cyprinus carpio L.). Ecotoxicol. Environ. Saf. 2007, 67 (3), 447–451. (5) Pinho, G. L. L.; Moura da Rosa, C.; Yunes, J. S.; Luquet, C. M.; Bianchini, A.; Monserrat, J. M. Toxic effects of microcystins in the hepatopancreas of the estuarine crab Chasmagnathus granulatus (Decapoda, Grapsidae). Comp. Biochem. Physiol., Part C: Toxicol. Pharmacol. 2003, 135 (4), 459–468. (6) Zegura, B.; Sedmak, B.; Filipic, M. Microcystin-LR induces oxidative DNA damage in human hepatoma cell line HepG2. Toxicon 2003, 41 (1), 41–48. (7) Dawson, R. M. The toxicology of microcystins. Toxicon 1998, 36 (7), 953–962. (8) Khreich, N.; Lamourette, P.; Renard, P.-Y.; Clave, G.; Fenaille, F.; Creminon, C.; Volland, H. A highly sensitive competitive enzyme immunoassay of broad specificity quantifying microcystins and nodularins in water samples. Toxicon 2009, 53 (5), 551–559. (9) Magalh~aes, V. F.; Marinho, M. M.; Domingos, P.; Oliveira, A. C.; Costa, S. M.; Azevedo, L. O.; Azevedo, S. M. F. O. Microcystins (cyanobacteria hepatotoxins) bioaccumulation in fish and crustaceans from Sepetiba Bay (Brasil, RJ). Toxicon 2003, 42 (3), 289–295. (10) Aguete, E. C.; Gago-Martínez, A.; Le~ao, J. M.; RodríguezVazquez, J. A.; Menard, C.; Lawrence, J. F. HPLC and HPCE analysis of microcystins RR, LR and YR present in cyanobacteria and water by using immunoaffinity extraction. Talanta 2003, 59 (4), 697–705. (11) Xie, L.; Park, H.-D. Determination of microcystins in fish tissues using HPLC with a rapid and efficient solid phase extraction. Aquaculture 2007, 271 (1-4), 530–536. 1495

dx.doi.org/10.1021/es1020795 |Environ. Sci. Technol. 2011, 45, 1490–1496

Environmental Science & Technology

ARTICLE

(12) Triantis, T.; Tsimeli, K.; Kaloudis, T.; Thanassoulias, N.; Lytras, E.; Hiskia, A. Development of an integrated laboratory system for the monitoring of cyanotoxins in surface and drinking waters. Toxicon 2010, 55 (5), 979–989. (13) Yuan, M.; Carmichael, W. W.; Hilborn, E. D. Microcystin analysis in human sera and liver from human fatalities in Caruaru, Brazil 1996. Toxicon 2006, 48 (6), 627–640. (14) Ferranti, P.; Fabbrocino, S.; Cerulo, M. G.; Bruno, M.; Serpe, L.; Gallo, P. Characterisation of biotoxins produced by a cyanobacteria bloom in Lake Averno using two LC-MS-based techniques. Food Addit. Contam., Part A 2008, 25 (12), 1530–1537. (15) Oehrle, S. A.; Southwell, B.; Westrick, J. Detection of various freshwater cyanobacterial toxins using ultra-performance liquid chromatography tandem mass spectrometry. Toxicon 2010, 55 (5), 965–972. (16) Bouaïcha, N.; Maatouk, I.; Vincent, G.; Levi, Y. A colorimetric and fluorometric microplate assay for the detection of microcystin-LR in drinking water without preconcentration. Food Chem. Toxicol. 2002, 40 (11), 1677–1683. (17) Heresztyn, T.; Nicholson, B. C. Determination of cyanobacterial hepatotoxins directly in water using a protein phosphatase inhibition assay. Water Res. 2001, 35 (13), 3049–3056. (18) Sheng, J.-W.; He, M.; Shi, H.-C.; Qian, Y. A comprehensive immunoassay for the detection of microcystins in waters based on polyclonal antibodies. Anal. Chim. Acta 2006, 572 (2), 309–315. (19) Zeck, A.; Eikenberg, A.; Weller, M. G.; Niessner, R. Highly sensitive immunoassay based on a monoclonal antibody specific for [4-arginine]microcystins. Anal. Chim. Acta 2001, 441 (1), 1–13. (20) Lindner, P.; Molz, R.; Yacoub-George, E.; Durkop, A.; Wolf, H. Development of a highly sensitive inhibition immunoassay for microcystin-LR. Anal. Chim. Acta 2004, 521 (1), 37–44. (21) Lindner, P.; Molz, R.; Yacoub-George, E.; Wolf, H. Rapid chemiluminescence biosensing of microcystin-LR. Anal. Chim. Acta 2009, 636 (2), 218–223. (22) Maraldo, D.; Rijal, K.; Campbell, G.; Mutharasan, R. Method for label-free detection of femtogram quantities of biologics in flowing liquid samples. Anal. Chem. 2007, 79 (7), 2762–2770. (23) Craighead, H. G. Nanoelectromechanical Systems. Science 2000, 290 (5496), 1532–1535. (24) Maraldo, D.; Mutharasan, R. Detection and Confirmation of Staphylococcal Enterotoxin B in Apple Juice and Milk Using Piezoelectric-Excited Millimeter-Sized Cantilever Sensors at 2.5 fg/mL. Anal. Chem. 2007, 79 (20), 7636–7643. (25) Maraldo, D.; Garcia, F. U.; Mutharasan, R. Method for Quantification of a Prostate Cancer Biomarker in Urine without Sample Preparation. Anal. Chem. 2007, 79 (20), 7683–7690. (26) Qian, X.; Metallo, S. J.; Choi, I. S.; Wu, H.; Liang, M. N.; Whitesides, G. M. Arrays of Self-Assembled Monolayers for Studying Inhibition of Bacterial Adhesion. Anal. Chem. 2002, 74 (8), 1805–1810. (27) Campbell, G. A.; Mutharasan, R. Method of measuring Bacillus anthracis spores in the presence of copious amount of Bacillus thuringiensis and Bacillus cereus. Anal. Chem. 2007, 79 (3), 1145–1152. (28) Campbell, G. A.; Mutharasan, R. Near real-time detection of Cryptosporidium parvum oocyst by IgM-functionalized piezoelectricexcited millimeter-sized cantilever biosensor. Biosens. Bioelectron. 2008, 23 (7), 1039–1045. (29) Campbell, G. A.; Uknalis, J.; Tu, S.-I.; Mutharasan, R. Detect of Escherichia coli O157:H7 in ground beef samples using piezoelectric excited millimeter-sized cantilever (PEMC) sensors. Biosens. Bioelectron. 2007, 22 (7), 1296–1302.

1496

dx.doi.org/10.1021/es1020795 |Environ. Sci. Technol. 2011, 45, 1490–1496