Ultrasensitive Detection of Single-Walled Carbon Nanotubes Using

Nov 25, 2015 - ... Engineering, Korea University, Seoul 136-701, Republic of Korea ..... Islam , M. F.; Rojas , E.; Bergey , D. M.; Johnson , A. T.; Y...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/ac

Ultrasensitive Detection of Single-Walled Carbon Nanotubes Using Surface Plasmon Resonance Daeho Jang,∥,† Wonhwi Na,∥,‡ Minwook Kang,† Namjoon Kim,§ and Sehyun Shin*,† †

School of Mechanical Engineering, ‡Department of Micro/Nano Systems, and §School of Biomedical Engineering, Korea University, Seoul 136-701, Republic of Korea S Supporting Information *

ABSTRACT: Because single-walled carbon nanotubes (SWNTs) are known to be a potentially dangerous material, inducing cancers and other diseases, any possible leakage of SWNTs through an aquatic medium such as drinking water will result in a major public threat. To solve this problem, for the present study, a highly sensitive, quantitative detection method of SWNTs in an aqueous solution was developed using surface plasmon resonance (SPR) spectroscopy. For a highly sensitive and specific detection, a strong affinity conjugation with biotin−streptavidin was adopted on an SPR sensing mechanism. During the pretreatment process, the SWNT surface was functionalized and hydrophilized using a thymine-chain based biotinylated single-strand DNA linker (B-ssDNA) and bovine serum albumin (BSA). The pretreated SWNTs were captured on a sensing film, the surface of which was immobilized with streptavidin on biotinylated gold film. The captured SWNTs were measured in real-time using SPR spectroscopy. Specific binding with SWNTs was verified through several validation experiments. The present method using an SPR sensor is capable of detecting SWNTs of as low as 100 fg/mL, which is the lowest level reported thus far for carbon-nanotube detection. In addition, the SPR sensor showed a linear characteristic within the range of 100 pg/mL to 200 ng/mL. These findings imply that the present SPR sensing method can detect an extremely low level of SWNTs in an aquatic environment with high sensitivity and high specificity, and thus any potential leakage of SWNTs into an aquatic environment can be precisely monitored within a couple of hours.

N

an increase in mass due to capturing self-aggregated SWNTs on the cantilever. However, even these cutting-edge methods and techniques yield some drawbacks in terms of sensitivity, specific selectivity, and practicality for the detection of carbon nanotubes in water. In fact, a quantitative detection of SWNTs at the subnanogram-per-liter level is still a challenging task. In addition, specific selectivity is further required even when using a DNA linker. In general, detection using an SPR sensor is considered one of the most precise, qualitative, and quantitative label-free analysis methods for biological and environmental studies. An SPR sensor precisely detects the changes in refractive index within an evanescent field owing to the reactions between the sensor chip surface and the analytes. The strength of an evanescent field is decreased exponentially based on the distance from the SPR sensor surface. The effective range of distance is up to 200 nm, and the sensitivity of an SPR sensor is maximized when the analytes exist within this effective range. Because SWNTs have diameters of less than a few nanometers, detection by an SPR sensor is the most appropriate detection method for nanomaterials including SWNTs. To the best of the

anomaterials have been rapidly applied in various industries as well as consumer applications.1 SWNT is a noteworthy nanomaterial owing to its remarkable mechanical, optical, and electrical properties.2,3 SWNTs are widely used for many applications such as electronic components,4 composites,5 and biosensors.6,7 Under this environment, technologies and applications using SWNTs will be continuously developed. However, recent studies have raised the potential risk of SWNTs associated with their toxicity. Rodriguez-Yanez et al.8 reported the damage to organs and systems exposed to SWNTs. In addition, carbon nanotubes were found to be related with hemotoxicity,9 genotoxicity, and carcinogenicity.8,10 Cell apoptosis in HEK293 cells (human embryo kidney cells) occurs with an SWNT concentration of 25 μg/mL.11 Because a low concentration of SWNTs can cause serious harm to humans after long-term exposure, the development of a highly sensitive method for the detection of low concentrations of SWNTs after long-term accumulation in human organs is urgently needed.12 There have been few studies reporting the detection of carbon nanotubes. Irin et al.13 proposed a method for detecting carbon nanotubes in biological samples through microwaveinduced heating, whereas Mota et al.14 described a method for detecting SWNTs in water using DNA and magnetic fluorescent spheres. In addition, Jang et al.15 developed a microcantilever sensor that monitors shifts in frequency from © 2015 American Chemical Society

Received: October 3, 2015 Accepted: November 25, 2015 Published: November 25, 2015 968

DOI: 10.1021/acs.analchem.5b03722 Anal. Chem. 2016, 88, 968−973

Article

Analytical Chemistry

Figure 1. Schematic illustration for the entire SWNT detection process: (a) pretreatment of aggregated SWNTs and (b) the pretreated SWNTs captured by streptavidin immobilized on the biotin-SAM of a gold sensing chip.

purchased from Sigma-Aldrich (St. Louis, MO, USA). Finally, we designed and fabricated an Au chip. Pretreatment of SWNTs Using B-ssDNA and BSA. A schematic illustration of the proposed detection method is shown in Figure 1. One of our key strategies was to apply a strong, specific binding mechanism between the target materials and the sensor surface. We therefore adopted a biotin− streptavidin pair, which is commonly known to be a strong and specific binding mechanism.16 In the pretreatment process of the SWNTs shown in Figure 1(a), we soaked 5000 μg of SWNTs in 2500 μL of deionized water (DIw) for 24 h. We observed that the degree of dispersion of the SWNTs was improved after the 24-h soaking in DIw. We added 500 μL of a B-ssDNA solution with a concentration of 100 μM in DIw and 125 μL of a BSA solution with a concentration of 2 mg/mL in DIw into the soaked SWNT solution. As a result, we generated a sample mixture composed of SWNTs, B-ssDNA, and BSA with a concentration of 100 μg/mL, 20 μM, and 100 μg/mL in DIw, respectively. The sample mixture was sonicated for 180 min at a power level of 70 W (HD 2070, Bandelin Sonopuls) for the wrapping of B-ssDNA and adsorption of the BSA on the surface of the SWNTs.

authors’ knowledge, there have not been any studies on the detection of SWNTs using an SPR method. Therefore, the present study aims to develop an innovative method for detecting SWNTs in an aquatic medium at extremely low concentration levels. For a highly sensitive and specific selectivity, a biotin−streptavidin conjugation is used for a strong affinity.

■

MATERIALS AND METHODS Materials. The SWNTs and BSA were purchased from Sigma-Aldrich (St. Louis, MO, USA), and the B-ssDNA (5′biotin GGG GGG GGG GTT TTT TTT TTT TTT TTT TTT TTT TTT TTT T) was synthesized at Bioneer (Seoul, Korea) based upon our own design. The B-ssDNA was sequenced for wrapping on the surfaces of the SWNTs. We purchased a 20,000 MWCO dialysis membrane (Slide-A-Lyzer Dialysis Cassette) from Thermo Scientific (Hudson, NH, USA). We used 1x PBS (Gibco PBS, Life technologies, Gaithersburg, MD, USA), chloroform (99.5%, Sigma-Aldrich, St. Louis, MO, USA), and ethanol (99.9%, Burdick & Jackson, Muskegan, MI, USA). Biotin-SAM (B564: Biotin-SAM Formation Reagent) was purchased from Dojindo Laboratories (Kumamoto, Japan), and mercaptoundecanol (MU) was 969

DOI: 10.1021/acs.analchem.5b03722 Anal. Chem. 2016, 88, 968−973

Article

Analytical Chemistry

DIw, and then dried under a nitrogen flow. After the cleaning process, the sensor chip was incubated in 100 μM of a biotin solution dissolved in ethanol overnight and washed with ethanol and DIw sequentially. As shown in Figure 1(b), a selfassembled monolayer (SAM) of biotin was formed on a gold sensing chip. After the biotin-SAM modification, the sensor chip was assembled using an SPR microfluidic system, and 10 μg/mL of streptavidin was injected into the microchannel of the SPR system for 20 min at a flow rate of 40 μL/min. Finally, the sensor chip was washed with PBS for 30 min at a flow rate of 40 μL/min. As shown in Figure 1(b), streptavidin was immobilized on the biotin-SAM. To maximize the binding efficiency between the biotin and streptavidin, the biotin-SAM density on the gold sensor chip was optimized through additional experiments by varying the biotin-SAM to MU molar ratio. Quantification of SWNTs Using an SPR Sensor. The 1 mL of pretreated SWNT samples, within a concentration range of 100 fg/mL to 400 ng/mL, were serially injected into a microchannel inlet of the sample placed on the streptavidinmodified gold sensing film at a flow rate of 40 μL/min, and the real-time signals from the interface of the gold sensing film induced by the interactions between the streptavidin on the film and the pretreated SWNTs were then measured. Any attachment of SWNTs on the sensor film caused a change in the resonance angle, which was monitored over time. The standard deviation for each concentration of the pretreated SWNTs was determined through five repeated measurements (n = 5). Instrument. We developed a Kretschmann-configuration based SPR spectroscopy with an angular interrogation.17 The sensor consists of a light source, a sensing part, and a 1/1.8-in. (1,280 pixels × 1,024 pixels) CMOS camera (IDS Co., Germany). The light source included an NIR-LED of 770 nm (Opnext Inc., Japan), an aspherical lens, a polarizer, and a bandpass filter. A three-channel microfluidic module 5(l) mm × 1(w) mm × 0.15(d) mm in dimensions and a gold sensor chip were sequentially organized on a hemicylindrical prism for the sensing part. The sensor chip was made using a 2 nm layer of sputtered chrome followed by a 48 nm layer of gold on a BK7 glass substrate with a thickness of 0.5 mm. The CMOS camera recorded the responding intensity of reflected light on each pixel as an image. The images were processed using lab-made MATLAB software programed in a previous work.18

The pretreatment of SWNTs, including biotinylation and hydrophilization using B-ssDNA and BSA, was verified through various validation experiments. First, the color of the solution including the SWNTs pretreated with B-ssDNA and BSA under sonication was observed by the naked eye. The solution of the pretreated SWNTs was a homogeneous black in color, whereas the solution of natural SWNTs under sonication showed the appearance of aggregated SWNTs, as shown in Figure S1. These results indicate that the SWNT solution is well dispersed owing to the successful surface hydrophilization of the SWNTs. We next verified the surface properties of the pretreated SWNTs using an SPR sensor. As an additional validation test, the pretreated SWNTs were loaded into a microchannel of the SPR sensor with a hydrophilic surface fabricated using plasma treatment and the hydrophobic surface of a natural gold film. It is worth noting that the natural surface property of an SWNT is hydrophobic, whereas the surface of an SWNT pretreated with B-ssDNA and BSA is comparatively hydrophilic. As a result, the pretreated SWNTs show a high affinity with a hydrophilic surface. Figure S2(a) shows comparison results of the responding SPR signal for the pretreated SWNT solution on both surfaces. The signal on the hydrophilic surface was much higher than that on the hydrophobic surface, at 4.89 and 0.52 pixels, respectively. These results indicate that the surfaces of the pretreated SWNTs were sufficiently hydrophilized. Dialysis Process for Removing Unbound B-ssDNA from the Sample Solution. After the pretreatment of the SWNTs using B-ssDNA, unbound B-ssDNA remained in the pretreated SWNT solution, some of which attached to the surfaces of the SWNTs through streptavidin immobilized on the sensor chip. These undesired attachments were detected by the SPR sensor and caused a false-positive signal, and the unbound B-ssDNA therefore had to be eliminated from the pretreated SWNT solution. Unbounded B-ssDNA in the pretreated SWNT solution was removed using a 20,000 MWCO dialysis membrane. One milliliter of the pretreated SWNT solution was injected into a dialysis cassette. The dialysis cassette was then immersed in 300 mL of a dialysis buffer, 1x PBS. The sample was dialyzed for 2 h, and the dialysis buffer was then changed. This dialysis process was repeated three times, and the final dialysis process was conducted overnight. To verify the dialysis performance, we identified the existence of B-ssDNA in the dialysis buffer. The dialysis buffer from the first step of the dialysis process was analyzed using an SPR sensor. The dialyzed pretreated SWNT solution was stored at room temperature. Figure S2(b) shows a comparison between the responding SPR signal of the solution before and after the dialysis process was applied. After the dialysis processing, the responding SPR signal of the solution became 13.06% lower than that of the nondialyzed solution. We checked the presence of B-ssDNA in the first dialysis buffer using an SPR sensor with streptavidin applied to the sensing surface after the dialysis processing was complete. Figure S2(c) shows the positive signal observed. The results suggest that the unbounded B-ssDNA was successfully removed through the dialysis processing, and we therefore confirmed the successful removal of unbounded B-ssDNA in a solution of pretreated SWNTs using a dialysis process. Fabrication of Sensor Surface Using Immobilizing Streptavidin. The sensor chip was first cleaned with chloroform (99.5%), rinsed with ethanol (99.9%) followed by

■

RESULTS AND DISCUSSION Optimization of the Streptavidin Density on the Sensor Chip. Before describing the experimental results, it is necessary to indicate that the sensitivity of an SPR assay depends on the density of the surface molecules. To achieve a large amount of immobilized receptors, it was necessary to optimize the accessibility between the biotin and the streptavidin. Using varying molar ratios of biotin-SAM to MU, multiple sensor chips were prepared and tested for optimizing the immobilization of the streptavidin and biotinylated analyte. Here, biotinylated BSA was used as the analyte. As shown in Figure 2, the streptavidin immobilization and biotinylated BSA coverage were highest when using only biotin-SAM (30 pixels and 12 pixels) and continuously decreased with a decrease in the molar ratios of biotin-SAM to MU. Thus, because the amount of biotin-BSA binding was the highest when only biotin-SAM was utilized, we decided to use only a biotin-SAM surface for the capturing of the 970

DOI: 10.1021/acs.analchem.5b03722 Anal. Chem. 2016, 88, 968−973

Article

Analytical Chemistry

increased stepwise from 100 fg/mL to 400 ng/mL. The SPR resonance values in the reference channel did not vary with the SWNT concentrations and yielded a nearly constant value. This result implies that the present SPR sensing mechanism successfully removes any false signals induced through a nonspecific hydrophobic interaction between the SWNTs and the gold-based sensing surface. It is worth noting that nonspecific binding has caused a serious drawback in the detection of the carbon nanotubes in aquatic media.20 Meanwhile, the detection channel showed significantly different signals compared with the reference channel. The solid-blue line in Figure 3(c) represents the data obtained from the detection channel as the concentrations of the SWNT solutions were stepwise increased. As depicted in Figure 3(b), the SWNTs wrapped with biotin were strongly bound with the streptavidin coated on the sensor surface of the detection channel. As the concentrations of the SWNT solutions increased, the SPR signals were apparently increased. Surprisingly, the present SPR sensor was able to detect a concentration of SWNTs of as low as 100 fg/mL. Figure 4(a) shows a normalized graph of SWNT detection using an SPR biosensor. This normalized graph was obtained

Figure 2. Comparison of the capturing efficiency for different biotinSAM to MU molar ratios.

biotinylated SWNTs. However, previous research has reported that a sensor chip surface with a 1:9 biotin-SAM:MU ratio shows the best results when assessing the PSA.19 We believe these mismatched results were caused by the different structures of the biotin-SAM. Unfortunately, the manufacturer of the biotin-SAM agent, Dojindo, did not share such information, and thus the exact reason for the discrepancy could not be determined. Detection of SWNT Quantification Using the SPR Sensor. To characterize the SPR sensor, we prepared various concentrations of SWNT solutions, i.e., 100 fg/mL, 1 pg/mL, 10 pg/mL, 100 pg/mL, 1 ng/mL, 10 ng/mL, 100 ng/mL, 200 ng/mL, and 400 ng/mL. The different concentrations of the SWNT solutions were introduced sequentially into both the detection channel and a reference channel. For the reference channel, it is notable that the sensor surface coated with streptavidin was completely blocked with biotinylated BSA, as shown in Figure 3(a), and that none of the SWNTs were designed to be captured on the surface of the reference sensor. The dotted-red line in Figure 3(c) shows the signals obtained from the reference channel as the concentration of SWNTs was

Figure 4. (a) Responding sensing signal of specific binding obtained by an SPR sensor with respect to SWNT concentrations of 0.0001, 0.001, 0.01, 0.1, 1, 10, 100, 200, and 400 ng/mL (p < 0.05) and (b) SWNT detection signal and linearity in a particular section with respect to concentrations of 0.1, 1, 10, 100, and 200 ng/mL.

by subtracting the value of the reference signals (dotted line) from that of the detection channel signal (solid line) in Figure 3(c). As described above, we were able to confirm that the presented SPR sensor clearly detects a SWNT-solution concentration of as low as 100 fg/mL, which to the best of the authors’ knowledge has yet to be achieved in any other studies. In a 100 fg/mL concentration solution, the net difference from the control value was 0.470 + 0.008 pixels, and

Figure 3. (a) Blocked surface, (b) sensing surface, and (c) real-time SPR angle-shift data obtained from a continuous flow-type experiment on a single gold sensing chip. 971

DOI: 10.1021/acs.analchem.5b03722 Anal. Chem. 2016, 88, 968−973

Analytical Chemistry

■

thus the limit of detection (LOD) in our measurements was determined to be 100 fg/mL of SWNTs. In fact, the present LOD was smaller by five orders than the best LOD found in the previous results of carbon-nanotube detection using a microcantilever of an atomic force microscopy.15 This high sensitivity was achieved by high affinity and strength of biotin− streptavidin system having a dissociation constant, kd, in the order of 4 × 10−14 M.16 It is worth noting that the toxic concentration of SWNTs (25 μg/mL) is much higher than the LOD of our system in terms of human HEK293 cells.11 However, a highly sensitive detection of SWNTs is still urgently required for the inspection of nanomaterial-free drinking water and food owing to the wide dispersal of nanomaterials in human environments. According to other research, CNTs can accumulate in human cells.21 As the normalized graph in Figure 4(a) shows, the saturation concentration was observed to be 400 ng/mL. In fact, any high concentration of SWNTs of above 1 μg/mL does not need to be measured using highly sensitive methods because other lowsensitive sensors such as a general spectroscope can be used. The linearity of SPR data within the range of 0.1 to 200 ng/mL was shown to be satisfactory at R2 = 0.9836, as shown in Figure 4(b), whereas that of the cantilever measurements15 was shown to be satisfactory at R2 = 0.8630. In our study, SWNTs are wrapped by B-ssDNA due to binding between aromatic nucleotide bases in B-ssDNA and side-wall of SWNTs through π-stacking.22 This binding through π-stacking also can be applied to other carbon nanomaterials, such as multi-walled carbon nanotube, graphene, and carbon quantum dot. Therefore, these carbon nanomaterials can be detected using our methods. In other words, carbon nanomaterials cannot be selectively identified with DNA wrapping method. However, detection sensitivity which is affected by affinity between analytes and receptors on the sensing surface might be different because of the structural differences between SWNT and other carbon nanomaterials.

■

Article

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b03722. Additional experimental data (Figure S1, comparison between undispersed SWNTs and well-dispersed SWNTs; Figure S2, validation of pretreated SWNT properties) (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Phone: 82 2 3290 3377. Fax: 82 2 928 5825. E-mail: [email protected]. Author Contributions ∥

These authors contributed equally.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This research was supported by the Nano Material Technology Development Program (Green Nano Technology Development Program) through the National Research Foundation of Korea (NRF), which is funded by the Ministry of Education, Science, and Technology (NRF-2011-0020090). One of authors (Daeho Jang) was also supported by a grant from Korea University.

■

REFERENCES

(1) Lee, J.; Mahendra, S.; Alvarez, P. J. ACS Nano 2010, 4, 3580− 3590. (2) Baughman, R. H.; Zakhidov, A. A.; de Heer, W. A. Science 2002, 297, 787−792. (3) Islam, M. F.; Rojas, E.; Bergey, D. M.; Johnson, A. T.; Yodh, A. G. Nano Lett. 2003, 3, 269−273. (4) Frackowiak, E.; Béguin, F. Carbon 2002, 40, 1775−1787. (5) Clayton, L. M.; Sikder, A. K.; Kumar, A.; Cinke, M.; Meyyappan, M.; Gerasimov, T. G.; Harmon, J. P. Adv. Funct. Mater. 2005, 15, 101− 106. (6) Rastogi, R.; Kaushal, R.; Tripathi, S. K.; Sharma, A. L.; Kaur, I.; Bharadwaj, L. M. J. Colloid Interface Sci. 2008, 328, 421−428. (7) Tang, X.; Bansaruntip, S.; Nakayama, N.; Yenilmez, E.; Chang, Y. L.; Wang, Q. Nano Lett. 2006, 6, 1632−1636. (8) Rodriguez-Yanez, Y.; Munoz, B.; Albores, A. Toxicol. Mech. Methods 2013, 23, 178−195. (9) Bussy, C.; Methven, L.; Kostarelos, K. Adv. Drug Delivery Rev. 2013, 65, 2127−2134. (10) Toyokuni, S. Adv. Drug Delivery Rev. 2013, 65, 2098−2110. (11) Cui, D.; Tian, F.; Ozkan, C. S.; Wang, M.; Gao, H. Toxicol. Lett. 2005, 155, 73−85. (12) Cherukuri, P.; Gannon, C. J.; Leeuw, T. K.; Schmidt, H. K.; Smalley, R. E.; Curley, S. A.; Weisman, R. B. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 18882−18886. (13) Irin, F.; Shrestha, B.; Canas, J. E.; Saed, M. A.; Green, M. J. Carbon 2012, 50, 4441−4449. (14) Mota, L. C.; Urena-Benavides, E. E.; Yoon, Y.; Son, A. Environ. Sci. Technol. 2013, 47, 493−501. (15) Jang, K.; Park, J.; Bang, D.; Lee, S.; You, J.; Haam, S.; Na, S. Chem. Commun. (Cambridge, U. K.) 2013, 49, 8635−8637. (16) Holmberg, A.; Blomstergren, A.; Nord, O.; Lukacs, M.; Lundeberg, J.; Uhlen, M. Electrophoresis 2005, 26, 501−510. (17) Schasfoort, R. B.; Tudos, A. J. Handbook of surface plasmon resonance; Royal Society of Chemistry: 2008. (18) Jang, D.; Lim, D.; Chae, G. H.; Yoo, J. Sens. Actuators, B 2014, 199, 488−492.

CONCLUSION

In summary, we proposed a highly sensitive, quantitative SWNT detection method in an aqueous solution using SPR spectroscopy. For a highly sensitive and specific detection, a strong conjugation affinity with biotin−streptavidin was adopted on an SPR sensing mechanism. The pretreatment of SWNTs, including biotinylation and hydrophilization using BssDNA and BSA, and purification using a dialysis process, was verified through various validation experiments. The novel approach achieved a high sensitivity, and the LOD was determined to be 100 fg/mL, which was smaller by five orders as compared to the best LOD found in carbon-nanotube detection approaches described thus far. In addition, the range between 0.1 and 200 ng/mL showed a particularly essential linear trend at R2 = 0.9836. The findings reported herein indicate that an SPR biosensor for SWNT detection can be applicable to a wide variety of biomedical and environmental fields. Conclusively, an SPR sensor was successfully developed for detecting SWNTs in an aquatic medium at extremely low levels, and thus the present method can be further utilized for an SWNT toxicity assessment following the leakage of SWNTs into water. 972

DOI: 10.1021/acs.analchem.5b03722 Anal. Chem. 2016, 88, 968−973

Article

Analytical Chemistry (19) Cao, C.; Kim, J. P.; Kim, B. W.; Chae, H.; Yoon, H. C.; Yang, S. S.; Sim, S. J. Biosens. Bioelectron. 2006, 21, 2106−2113. (20) Lee, E. G.; Park, K. M.; Jeong, J. Y.; Lee, S. H.; Baek, J. E.; Lee, H. W.; Jung, J. K.; Chung, B. H. Anal. Biochem. 2011, 408, 206−211. (21) Simon-Deckers, A.; Gouget, B.; Mayne-L’hermite, M.; HerlinBoime, N.; Reynaud, C.; Carriere, M. Toxicology 2008, 253, 137−146. (22) Zheng, M.; Jagota, A.; Semke, E. D.; Diner, B. A.; McLean, R. S.; Lustig, S. R.; Richardson, R. E.; Tassi, N. G. Nat. Mater. 2003, 2, 338− 342.

973

DOI: 10.1021/acs.analchem.5b03722 Anal. Chem. 2016, 88, 968−973