Highly sensitive detection of glucose: A quantitative approach employing nanorods assembled plasmonic substrate

Talanta 165 (2017) 516–521

Contents lists available at ScienceDirect

Talanta journal homepage: www.elsevier.com/locate/talanta

Highly sensitive detection of glucose: A quantitative approach employing nanorods assembled plasmonic substrate

MARK

Qiulan Chena,b,c,1, Yu Fua,b,1, Weihong Zhanga, Suibo Yea, Hao Zhanga, Fangyan Xiea, Li Gonga, ⁎ Zhanxiao Weib, Haoyu Jinc, , Jian Chena,d,⁎⁎ a

Instrumental Analysis & Research Center, Sun Yat-sen University, Guangzhou 510275, China School of Chemistry and Chemical Engineering, Sun Yat-sen University, Guangzhou 510275, China c Department of Medical Devices, Guangdong Food and Drug Vocational College, Guangzhou 510520, China d Guangdong Province Key Laboratory of Display Material and Technology, Sun Yat-sen University, Guangzhou 510275, China b

A R T I C L E I N F O

A BS T RAC T

Keywords: SERS Plasmonic substrate Glucose Sensing Nanorods Quantitative

Sensitive glucose detection enables indirect blood glucose sensing through easily accessible biofluids such as saliva and sweat. In this work, silver coated gold nanorods (Au@Ag NRs) were synthesized and used to prepare plasmonic substrate for surface-enhanced Raman spectroscopy (SERS) to leverage highly sensitive detection of glucose for quantitative analysis. By synthetically manipulating of gold NRs and the outer silver shell, the size and aspect ratio of Au@Ag NRs were optimized, and the plasmon resonance wavelength was tuned to approximately the excitation wavelength. 4-Mercaptophenyl-boronic acid (4-MPBA) and 4-Cyanophenylboronic acid (4-CPBA) were used as primary and secondary receptors respectively to specifically capture glucose molecules. The distinct Raman peak at 2226 cm−1 of the cyano group in 4-CPBA was used as a signal reporter for glucose sensing. It is located in a biological silent region (1800–2800 cm−1), thus offering specific sensing of glucose, without the interference of other endogenous molecules. Our results showed that the SERS substrate was long-term stable. Glucose in urine solution with additive glucose was quantitatively and specifically determined, with the detection limit down to 10−8 M. Further experiments using urine from mild diabetes shows positive results, demonstrating the feasibility of clinical use.

1. Introduction Nowadays, glucose sensors are widely used for self-monitoring of blood glucose (SMBG) for diabetes [1–5]. According to the report of World Health Organization (WHO) in 2015, the global prevalence of diabetes in 2014 was around 9% for adults, and around 1.5 million deaths were directly caused by diabetes. SMBG is therefore of importance for diabetic patients in the absence of a cure. Apart from diabetes control, glucose sensing is also used to identify both hypoglycemia and hyperglycemia for diagnostic purposes [6–8]. Hyperglycemia may result from acute conditions such as stroke and trauma, while hypoglycemia may be caused by chronic conditions such as hyperinsulinism and ketotic hypoglycemia. In addition, tight glycemic control also plays important roles in increasing clinical outcomes and reducing mortality and morbidity in intensive care unit [9–12]. Commonly used vein blood glucose meters involve enzymatic reaction and the resultant products such as hydrogen peroxide or

nicotinamide adenine dinucleotide phosphate (NADPH) can be determined by resultant optical absorption or electrical current. However, such kind of sensors challenge in precision and accuracy due to a number of affect factors such as ambient environment, patient metabolic factors and medication. Latest reports showed that sweat glucose content could be used to estimate the level of blood glucose [13–16]. However, the content of glucose in such media is only 1–10% of that in blood. Such weak signal is often embedded in background signal of other biomolecules and is difficult to pick up using commercial glucose sensors. Therefore, sensitive and selective glucose detection is of use and may leverage new applications. Surface-enhanced Raman spectroscopy (SERS)-based glucose sensor is a prospective alternative. SERS has been considered as one of the promising sensing techniques due to its distinct advantages in terms of high sensitivity, non-invasiveness, non-labeling, and fingerprint-type way of sensing [17–19]. SERS commonly utilizes noble metal nanostructure, such as Au, Ag and Cu, for excitation of surface plasmon

⁎

Correspondence to: Department of Medical Devices, Guangdong Food and Drug Vocational College, Rm 208, Bldg. B, Longdong Road North, Guangzhou 510520, China Correspondence to: Instrumental Analysis & Research Center, Sun Yat-sen University, Rm 213, No. 135, Xingang Xi Road, Guangzhou 510275, China E-mail addresses: [email protected] (H. Jin), [email protected] (J. Chen). 1 Qiulan Chen and Yu Fu contributed equally to this work and thus share the first authorship. ⁎⁎

http://dx.doi.org/10.1016/j.talanta.2016.12.076 Received 4 October 2016; Received in revised form 23 December 2016; Accepted 26 December 2016 Available online 26 December 2016 0039-9140/ © 2016 Elsevier B.V. All rights reserved.

Talanta 165 (2017) 516–521

Q. Chen et al.

resonance, the associated electromagnetic field is thus enhanced and Raman signal is amplified in several orders of magnitude (typically 106–1014) [19–24], enabling the capability of single molecule sensing. In this method, the preparation of nanostructured SERS active substrate with significant and reproducible electromagnetic field enhancement is the crucial part to optimize the sensing performance. There have existed methods to prepare nanostructured SERS substrate, for example, deep UV lithography etching on silicon wafer [25], glass nanopillar arrays with nanogap-rich silver nano-islands [26], and metal film over nanosphere (MFON) [27,28]. MFON is a promising method to prepare nanostructured SERS substrate. Polystyrene nanospheres were spin-coated onto glass substrate to form spheres monolayer and following sputtering of silver and gold films were performed to form SERS active substrate. Nevertheless, the performance of the substrate highly depended on surface morphology of the nanostructure, which requiring nanometer precision. To achieve optimized electric filed enhancement effect for both excitation and the Raman scattering processes, the plasmon wavelengths of the nanostructures should be tuned in resonance with the excitation laser. This may be the main reason why the sensitivity and limit of detection (LOD) is limited for the above-mentioned work. With this as the background, metal NRs assembled nanostructure have received great interests for its tunable plasmonic properties [29]. In this method, high curvature at the ends of NRs results in strong electromagnetic field enhancement. Furthermore, the gap regions between neighboring NRs act as “hot spots” to provide extremely high SERS signal for molecules in between. However, the use of nanorods for glucose sensing remains unexplored. In this study, SERS active substrate was prepared by depositing silver coated gold nanorods (Au@Ag NRs) onto a cover glass surface. With the size and aspect ratio of NRs optimized, the plasmon resonance wavelength was tuned to extremely close to the excitation wavelength. Experimental results demonstrated that the SERS substrate showed qualitative and quantitative response to not only glucose urine solution, but also urine samples with extreme low glucose content from mild diabetes. By suggesting a new sensitive way, our work facilitates glucose quantitative determination steps further towards single molecule detection.

solution, which containing HAuCl4 (0.47 mM), AgNO3 (0.047 mM), HCl (18.74 mM), AA (0.75 mM) and CTAB solution (93.70 mM). The mixture was kept at room temperature for 12 h for growth. At the presence of Ag+, the aspect ratio of NRs can be tuned by the ratio of gold seed to gold salt. To exhibit spectroscopic characteristics of silver nanoparticles, an additional silver layer is coated on Au NRs by wet chemical reduction method [25]. 4 ml of the as-prepared Au NRs solution was centrifuged to remove the supernatant. Subsequently, sequential addition of CTAC (4 ml, 80 mM), AgNO3 (200 µL, 0.01 M), AA (100 µL, 0.1 M) was performed, which resulted in reduction of Ag+ ions to coat on the Au NRs. The mixture solutions was kept at 65 °C for 6 h for complete reduction of AgNO3. The plasmonic properties of resultant Au@Ag NRs were determined by silver. 2.3. Preparation of SERS substrate As reported previously, Au@Ag NRs can be uniformly and densely deposited on clean cover glass surface [30]. 4 ml of the as-prepared Au@Ag NRs dispersion solution was washed by centrifugation (5000 rpm, 5 min) for removal of the supernatant, and re-dispersed at the same volume in ultrasound. The above procedures were repeated twice to remove the stabilizer as much as possible. Clean cover glass was then immersed in the resultant solution for 12 h for deposition of NRs densely on the substrate. For glucose sensing, the substrate was first incubated in 4-MPBA dissolved in ethanol (10−5 M) for 2 h to allow the thiol group (-SH) covalently bonded with Ag to form Ag-S bond. The functionalized substrate was then incubated in analyte for 3 h for binding of the analyte molecules onto the boronic acid receptor. Afterwards, the substrate was further incubated in 4-CPBA solution (10−5 M) for 3 h. 4-CPBA specifically bound with glucose only to form sandwiched 4MPBA–glucose–4-CPBA complexes, which was not the case for other saccharide such as fructose and galactose. The substrate was then taken out and washed thoroughly with DI water. Finally, the functionalized substrate was dried with N2 flow and placed on a glass slide for Raman measurement. Raman scattering measurements was performed in a confocal Raman microscope (Renishaw inVia) equipped with a 633 nm excitation laser to collect the signal. The data acquisition time was 10 s for each accumulation with 5% of incident laser power.

2. Material and methods 2.1. Reagents

3. Results To perform glucose quantification assay, glucose anhydrous (AR, Guangzhou chemical reagent factory, China) was dissolved in urine from healthy adults or DI water to prepare glucose urine solution or aqueous solution with various concentrations. 4-Mercaptophenylboronic acid (Sigma-Aldrich, USA) and 4-Cyanophenyl boronic acid (Sinopharm Chemical Reagent Co. Ltd., China) were dissolved in ethanol respectively to prepare solutions of 10−5 M. Cetyltrimethylammonium bromide (CTAB, 99%) obtained from Kermel (Tianjin, China), hydrogen tetrachloroaurate (HAuCl4·3H2O, 99.99%), silver nitrate (AgNO3, 99.99%), sodium borohydride (NaBH4, 99.99%), ascorbic acid (AA, 98%), octadecyltri-methoxysilane (OTS, 99.9%), and hexadecyl trimethyl ammonium chloride (CTAC, 99%) were purchased from Sigma-Aldrich (USA).

3.1. Characterization of SERS substrate The preparation of SERS active substrate plays the most crucial role for quantitative analysis. Silver and gold nanocrystals are the most explored in plasmonic applications among the noble metals. Silver nanocrystal shows better performance than that of gold nanocrystal in enhancing Raman signal as silver nanoparticle exhibits much weaker plasmon damping effect in visible regime and larger light scattering due to small imaginary part of the dielectric function [29,31,32]. However, the preparation of elongated anisotropic nanocrystals with uniform size and shape, as well as synthetic control of its aspect ratio, remains a challenge. Gold NRs show good tunability of plasmon resonance bands by synthetically manipulating their size and shape within visible and near-infrared spectral regions [29,33]. In recent years, it has been reported that introducing an additional silver shell to gold nanorods may cause blue-shift of plasmon resonance and narrow the plasmon linewidth, thus enabling tunable resonance wavelength over a broad spectral range and improved sensing performance [34,35]. In this work, we employed silver coated gold NRs for glucose sensing, so that both the plasmonic properties of silver NRs and the attractive advantages of gold NRs in terms of chemical stability and tunable longitudinal plasmon wavelengths can be introduced. The

2.2. Preparation of silver coated gold nanorods (Au@Ag NRs) Gold NRs were synthesized by seed-mediated growth method, which have been reported previously [17]. Briefly, 250 µL HAuCl4 aqueous solution (10 mM) was first added into 9.75 ml CTAB aqueous solution (0.1 M). Subsequently, ice-cold NaBH4 aqueous solution (600 µL, 0.1 M) was rapidly prepared and added into the mixture. Adequate inversion mixing was conducted in each step. CTAB-stabilized Au nanocrystal seed formed at room temperature in 3 h. To obtain Au NRs, 10 µL of the seed solution was then added to 10.67 ml growth 517

Talanta 165 (2017) 516–521

Q. Chen et al.

aspect ratio correlating with the longitudinal plasmon mode was first tuned by synthetically manipulating the gold NRs. Subsequently, an additional silver layer was coated. Blue shift of the extinction peak of the NRs was observed, which agrees with previous publications [17,33]. The reason may be that the silver layer was thicker at the sides than the terminal areas, leaded to a decrease in the aspect ratios of the NRs. Therefore, further minor adjustment of the aspect ratio could be realized by controlling the thickness of the outer shell. The longitudinal plasmon wavelength of as-prepared Au@Ag NRs was finally tuned to 628.5 nm, which was quite close to the excitation wavelength. The average length, width and aspect ratio were 65 nm, 37 nm and 1.77, respectively. The average shell thickness was around 10 nm. Finally, Au@Ag NRs were deposited on a clean cover glass surface and formed a nanostructured SERS substrate, which maintained the plasmonic properties of the individual NRs [30]. The nanostructure of the SERS substrate was characterized to guarantee its plasmonic properties prior to glucose detection. Raman enhancement factor (EF) was evaluated using Rhodamine 6 G (R6G) dissolved in ethanol, 10 µL R6G (10−4 M) solution was dripped on the SERS substrate and blank cover glass respectively. The solvent was evaporated quickly in a minute. Subsequently, Raman measurement was conducted under identical experiment conditions. Assuming that the spread area for the two substrates is the same, EF can be roughly estimated as the ratio of Raman peak intensity. The calculated result was around 1011, which is seven orders of magnitudes higher than that of previously reported [27]. Functionalization of the SERS substrate was performed by immobilizing 4-MPBA molecules on the surface of Au@Ag NRs through Ag-S covalent bonds. For glucose sensing applications towards clinical use, long-term stability and large-area uniformity are desirable for repeated measurements. To validate this, the phenyl-ring breathing mode at 1068 cm−1 in Raman spectrum was monitored to evaluate the activity and stability of the functionalized SERS substrate. SERS mapping of the 4-MPBA functionalized substrate was conducted. As shown in Fig. 1, the Au@Ag NRs distribution is relatively uniform on the substrate. Slight aggregation was found, which was believed to be caused by the inadequate dispersion of the NRs and cleaning of the cover glass. This uniformity was adequate enough to produce enhanced signal at various positions of the substrate with minimal variation and guaranteed its reproducibility. To assess the stability of the 4-MPBA functionalized substrate, the substrate was stored at room temperature and Raman measurements were continuously conducted under identical experiment conditions. As shown in Fig. 2, the signal intensity at 1068 cm−1 in Raman spectrums remained at same magnitude for up to 22 days at 25 °C in vacuum oven, with no significant variations

Fig. 2. SERS spectra of 4-MPBA functionalized substrate with no analyte. The substrate was stable for over 22 days.

observed. This is believed to be mainly due to high binding affinity of 4-PMBA on the glass substrate and subsequent formed strong Ag-S covalent bond. Such long time stability allows potential applications for clinical use.

3.2. Glucose sensing Fructose and galactose are two of the most abundant monosaccharaides in biological analyte [36–38]. To facilitate glucose sensing towards clinical use, selective capture of glucose among saccharides molecules is the crucial aspect to be addressed to guarantee the detection accuracy and specificity [39–41]. In this work, selective sensing of glucose was exploited by adopting a sandwiched probing approach by the use of two carbohydrate receptors. The relevant mechanism was illustrated in Fig. 3. First, 4-MPBA immobilized on SERS substrate acted as primary carbohydrate receptor. Carbohydrate such as glucose, fructose and galactose can be captured by the boronic acid group (B(OH)2) onto the substrates when incubating with the

Fig. 3. Illustration diagram of sensing mechanism. 4-MPBA and 4-CPBA were used as primary and secondary glucose receptor respectively for specific capturing of glucose. The distinct Raman peak of cyano group at 2226 cm−1 was employed as signal reporter for glucose quantification.

Fig. 1. SERS images mapped with the Raman bands at 1068 cm−1. The Au@Ag nanorods was uniformly distributed on the substrate, with only slight aggregation caused by inadequate dispersion of the NRs and cleaning of the cover glass.

518

Talanta 165 (2017) 516–521

Q. Chen et al.

analyte. The difference between glucose and other saccharides is that each glucose molecule has two sets of diol groups in a synperiplanar arrangement at its equatorial positions [33]. Therefore, when the substrate was incubated with the secondary receptor, i.e., Cyano group (C≡N)-functionalized 4-CPBA molecules, specific labeling of glucose occurred, as the boronic acid group straddled the left set of synperiplanar arranged diol group in glucose, leaving other saccharides without labeling. When incubated with the analyte, reversible and rapid formation of stable boronic acid-diol complexes occurred in a quantity-dependent manner, as glucose reacting with two aromatic boronic acids to form boron–oxygen (B-O) covalent bonds, producing an equilibrium mixture of sandwiched 4-MPBA–glucose–4-CPBA complex [42]. We assumed that this sandwiched complex might allow quantification of glucose through fingerprint signal monitoring of cyano group of each complex. The distinct Raman peak of cyano group at 2226 cm−1 was employed as signal reporter because it located in a biological silent region (1800–2800 cm−1). This means glucose can be recognized from biological analyte without interference by other biological molecules. To verify the detection specificity of our protocol, aqueous glucose, galactose, fructose, solutions and their mixture with same concentration were prepared for SERS measurement. As shown in Fig. 4, a significant peak of Cyano group at 2226 cm−1 was observed for both glucose solution and the mixture sample. However, no obvious Raman signal exhibited in the spectral regime for aqueous fructose and galactose solutions. This observation suggested that no 4-CPBA was bound with fructose or galactose molecules, and the assay protocol was highly specific to glucose, with ignorable disturbance of fructose and galactose. To further demonstrate the interference of other biological molecules in clinical samples, fresh urine sample from healthy adult was used to prepare urine glucose solution of 10−4 M and DI water was used to prepare aqueous glucose solution at same concentration as positive control sample. As shown in Fig. 5, the Raman signal of Cyano group was almost the same for urine and aqueous glucose samples, and no other obvious Raman peak was observed. Nevertheless, no clear peak was found throughout this spectral region for the negative control (blank substrate after 4-MPBA treatment). This demonstrated that there is no unlinked 4-CPBA existed on the blank substrate and no significant interference from other biological molecules in urine. Raman signal specifically corresponded to the sandwiched 4-MPBA– glucose–4-CPBA complexes. Furthermore, this also suggested that the biological molecules in urine samples is “silent” enough, and do not have any interference for the detection of glucose. This specific sensing capability suggested great potential applications of this approach for non-invasive glucose monitoring for biological media.

Fig. 5. SERS spectrums of urine glucose, aqueous glucose and negative control sample. Urine glucose was prepared using glucose power dissolved in urine from healthy adult (10−4 M), aqueous glucose solution was prepared with same concentration as positive control. Negative control sample was blank SERS substrate after 4-MPAB treatments.

With the detection specificity discussed above, we moved one step further to evaluate the possibility of urine glucose quantification assay. Urine glucose solutions with various concentrations (10−2 M, 10−3 M, 10−4 M, 10−5 M, 10−6 M, 10−7 M, 10−8 M) were prepared by adding glucose powder to urine from healthy adults for SERS measurement. The SERS peak intensity and peak area of Raman signal at 2226 cm−1 were used to study the concentration dependency. As can be in Fig. 6, the signal peak intensity showed a linear response to the concentrations of glucose. Similar results could be observed for peak area. As mentioned above, the Raman signal specifically corresponded to sandwiched 4-MPBA–glucose–4-CPBA complex, thus can be used for quantitative analysis of glucose content. The fitting results agreed with the assumption. The sensing limit of detection (LOD) was around 10−8 M. This proved that glucose quantification of urine is possible, without any disturbance from other molecules. In previously publications or commercial products, typical LOD is at mM level [13,25,27,47,48], as shown in Table 1. By tuning the aspect ratio of gold NRs and manipulating the thickness of outer silver shell, the Au@ Ag NRs were optimized with extinction spectrum peak extremely close to that of the incident light, thus maximizing the electric field enhancement effect. Furthermore, combining the inner plasmonic properties of silver nanoparticles with better Raman signal enhancing

Fig. 6. Peak intensity and peak area of SERS signal versus various concentrations of urine glucose from 10−8 M to 10−2 M, which were prepared by adding glucose powder to healthy urine. Peak intensity and peak area showed a linear response to glucose concentration. Experiment using aqueous glucose sample shows similar results.

Fig. 4. SERS spectrums of aqueous glucose, galactose, fructose solutions and their mixture with same concentration (10−2 M).

519

Talanta 165 (2017) 516–521

Q. Chen et al.

measured by conventional glucose biosensors listed in Table 1. Therefore, the Raman signal can be used for quantitatively determination of urine glucose. No spectral interference of inherent biomolecules was observed. High specificity of this method also suggests that this protocol is highly applicable in saliva and sweat samples. Our results thus prove that this method offers high competence and great potentiality for on-site glucose detection and may create new opportunities for non-invasive monitoring of glucose with high sensitivity and accuracy.

Table 1 Comparison of sensing performance for different detection methods. Limit of detection (LOD)

Detection method

References

10 nM 0.1 mM

SERS with NRs assembled substrate SERS with Au/Ag coated polystyrene substrate SERS with MFON substrate Near-infrared (NIR) Electrochemical method Commercial glucose biosensors based on Electrochemical method

This work [27]

5 mM 0.11 mM 0.01 mM 0.56 mM

[25] [47] [13] [48]

4. Discussion In this study, we have introduced silver coated gold nanorods assembled to form nanostructured SERS-active substrate for spectroscopic glucose quantitative analysis. In this method, NRs exhibiting plasmonic properties of silver was realized in a tunable manner. Glucose quantity showed a linear relationship with the Raman signal. Interference of biomolecules in biological samples was avoided by adopting a sandwiched probing method. To the best of our knowledge, we are the first to achieve LOD down to 10 nM. In addition, our method provides a number of advantages over traditional practice in terms of high specificity, non-invasive, long-term stability, anti-interference and sensitive detection. Apart from applying the substrate for glucose sensing and quantification in artificially prepared glucose urine solution, we also demonstrated that extreme low glucose content in mild diabetic human urine could be detected and identified. In conclusion, we have demonstrated that our method provides a sensitive, specific approach for quantitative determination of glucose content in biological meida and have great potential and competence in clinical use. Acknowledgement

Fig. 7. SERS spectrum of substrate incubated in urine samples from mild diabetic patient and healthy adult. The tiny amount of glucose can be detected from diabetic urine.

The authors wish to acknowledge funding support from National Natural Science Foundation of China under Grant No. 51373205 and No. 21576301, Project supported by GDHVPS 2016, and the Research Grant of Guangdong Food and Drug Vocational College No. 2014YZ001.

effect than gold, the detection sensitivity of this protocol is thus maximized, and the LOD is increased 4 orders of magnitude compared with previous publications [27,28,43], and is 6 orders of magnitude higher than the threshold regime of blood glucose content for diabetes (7.0–11.1 mM) in the diagnostic criteria of American Diabetes Association in 2010. The significantly increased sensitivity allows potential applications in those high accuracy-demanded areas such as non-invasive glucose monitoring in low-content media and study of the pharmacological effect of hypoglycemic drugs on glucose levels. Urine test of glucose is the easiest way for preliminary screening of diabetes. There is almost no urine glucose for healthy human. Small quantity of urine glucose exists when blood glucose exceeds renal glucose threshold for diabetes and have a correlation with blood glucose concentration. However, current urine test paper strip suffers from the accuracy and environmental interference, and is only used for semi-quantification screening. New quantification applications may be leveraged if sensitive sensing techniques are available because urine is much easier to obtain than blood. On the other hand, saliva and sweat are considered as potential candidate media for clinical use to noninvasively and real-time monitor blood glucose concentration in an indirect manner [44–46]. However, low content of glucose in these biofluids requires new generation of sensing techniques. To demonstrate the feasibility of practical use of our protocol in clinical samples with low content of glucose, freshly collected urine samples from healthy adults and mild diabetes without any pretreatment were used for SERS analysis. As can be seen in Fig. 7, the peak signal at 2226 cm−1 was clearly observed for samples from mild diabetes, while no obvious signal was observed for samples from healthy adults. According to the relationship between Raman intensity and glucose concentration figured out in Fig. 6, the glucose content was around 0.55 mM. Such low content is difficult to be accurately

References [1] [2] [3] [4] [5] [6] [7] [8] [9]

[10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21]

520

W.T. Wu, N. Mitra, E.C.Y. Yan, S.Q. Zhou, ACS Nano 4 (2010) 4831–4839. J.D. Newman, A.P.F. Turner, Biosens. Bioelectron. 20 (2005) 2435–2453. E.H. Yoo, S.Y. Lee, Sensors 10 (2010) 4558–4576. K. Tonyushkina, J.H. Nichols, J. Diabetes Sci. Technol. 3 (2009) 971–980. S.F. Clarke, J.R. Foster, Brit. J. Biomed. Sci. 69 (2012) 83–93. Y.W. Zhang, S. Liu, L. Wang, X.Y. Qin, J.Q. Tian, W.B. Lu, G.H. Chang, X.P. Sun, Rsc. Adv. 2 (2012) 538–545. D. Das, D.M. Kim, D.S. Park, Y.B. Shim, Electroanal 23 (2011) 2036–2041. O. Lyandres, R.P. Van Duyne, J.T. Walsh, M.R. Glucksberg, S. Mehrotra, Analyst 135 (2010) 2111–2118. B.G. Van, den, A. Wilmer, G. Hermans, W. Meersseman, P.J. Wouters, I. Milants, E. Van Wijngaerden, H. Bobbaers, R.N. Bouillon, Engl. J. Med. 354 (2006) 449–461. D. Vlasselaers, I. Milants, L. Desmet, P.J. Wouters, I. Vanhorebeek, et al., Lancet 373 (2009) 547–556. G. Hermans, M. Schrooten, P. Van Damme, N. Berends, B. Bouckaert, W. De Vooght, W. Robberecht, G.V. Berghe, Crit. Care. 13 (2009) 248. D.C. Gore, D. Chinkes, J. Heggers, D.N. Herndon, S.E. Wolf, M. Desai, J. Trauma. 51 (2001) 540–544. H. Lee, T.K. Choi, Y.B. Lee, H.R. Cho, R. Ghaffari, L. Wang, H.J. Choi, T.D. Chung, N.S. Lu, T. Hyeon, S.H. Choi, D.H. Kim, Nat. Nano. 11 (2016) 566–572. J. Moyer, D. Wilson, I. Finkelshtein, B. Wong, R. Potts, Diabetes Technol. The. 14 (2012) 398–402. K. Sakaguchi, et al., J. Diabetes Sci. Tehcnol. 7 (2013) 678–688. O. Olarte, J. Chilo, J. Pelegri-Sebastia, K. Barbé, W.V. Moer, Conf. Proc. IEEE Eng. Med. Biol. Soc. (2013) 1462–1465. J.X. Wen, H.B. Zhang, H.J. Chen, W.H. Zhang, J. Chen, J. Opt. 17 (2015) 114015. L.X. Quang, C. Lim, G.H. Seong, J. Choo, K.J. Do, S.K. Yoo, Lab Chip 8 (2008) 2214–2219. N.P.W. Pieczonka, R.F. Aroca, Chem. Soc. Rev. 37 (2008) 946–954. J.R. Lombardi, R.L. Birke, Acc. Chem. Res. 42 (2009) 734–742. C.L. Haynes, C.R. Yonzon, X.Y. Zhang, R.P. Van Duyne, J. Raman Spectrosc. 36 (2005) 471–484.

Talanta 165 (2017) 516–521

Q. Chen et al.

[35] K.S. Lee, M.A. El-Sayed, J. Phys. Chem. B 110 (2006) 19220–19225. [36] T. Kawasaki, H. Akanuma, T. Yamanouchi, Diabetes Care 25 (2002) 353–357. [37] T. Kawasaki, N. Ogata, H. Akanuma, T. Sakai, H. Watanabe, K. Ichiyanagi, T.M. Yamanouchi, Clin. Exp. 53 (2004) 583–588. [38] C.A. Williams, T. Phillips, I.M. Macdonald, Clin. Exp. 32 (1983) 250–256. [39] T.D. James, K.R.A.S. Sandanayake, S. Shinkai, Angew. Chem., Int. Ed. Engl. 33 (1994) 2207–2209. [40] T.D. James, K.R.A.S. Sandanayake, S. Shinkai, Angew. Chem. Int. Ed. Engl. 35 (1996) 1910–1922. [41] V.V. Karnati, X. Gao, S. Gao, W. Yang, W. Ni, S. Sankar, B.B. Wang, Bioorg. Med. Chem. Lett. 12 (2002) 3373–3377. [42] M.P. Nicholls, P.K.C. Paul, Org. Biomol. Chem. 2 (2004) 1434–1441. [43] V.K. Gupta, N. Atar, M.L. Yola, M. Eryılmaz, H. Torul, U. Tamer, I.H. Boyacı, Z. Üstündag, J. Colloid Interface Sci. 406 (2013) 231–237. [44] V.S. Siu, J. Feng, P.W. Flanigan, G.T.R. Palmore, D. Pacifici, Nanophotonics 3 (2014) 125–140. [45] W. Zhang, Y. Du, M.L. Wang, Sens. Bio-Sens. Res. 4 (2015) 23–29. [46] H. Lee, T.K. Choi, Y.B. Lee, H.R. Cho, R. Ghaffari, L. Wang, H.J. Choi, T.D. Chung, N. Lu, T. Hyeon, S.H. Choi, D.H. Kim, Nat. Nanotechnol. 11 (2016) 566–572. [47] B. Nacht, C. Larndorfer, S. Sax, S.M. Borisov, M. Hajnsek, F. Sinner, E.J.W. ListKratochvil, I. Klimant, Biosens. Bioelectron. 64 (2015) 102–110. [48] E.H. Yoo, S.Y. Lee, Sensors 10 (2010) 4558–4576.

[22] O. Lyandres, R.P. Van Duyne, J.T. Walsh, M.R. Glucksberg, S. Mehrotra, Analyst 135 (2010) 2111–2118. [23] K. Ma, J.M. Yuen, N.C. Shah, J.T. Walsh, M.R. Glucksberg, R.P. Van Duyne, Anal. Chem. 83 (2011) 9146–9152. [24] D.A. Stuart, J.M. Yuen, N.S.O. Lyandres, C.R. Yonzon, M.R. Glucksberg, J.T. Walsh, R.P. Van Duyne, Anal. Chem. 78 (2006) 7211–7215. [25] U.S. Dinish, F.C. Yaw, A. Agarwal, M. Olivo, Biosens. Bioelectron. 26 (2011) 1987–1992. [26] Y.J. Oh, K.H. Jeong, Adv. Mater. 24 (2012) 2234–2237. [27] K.V. Kong, C.J.H. Ho, T.X. Gong, W.K.O. Lau, M. Olivo, Biosens. Bioelectron. 56 (2014) 186–191. [28] K.V. Kong, Z. Lam, W.K.O. Lau, W.K. Leong, M. Olivo, J. Am. Chem. Soc. 135 (2013) 18028–18031. [29] H.J. Chen, L. Shao, Q. Lia, J.F. Wang, Chem. Soc. Rev. 42 (2013) 2679–2724. [30] L. Shao, Q.F. Ruan, R.B. Jiang, J.F. Wang, Small 10 (2014) 802–811. [31] R.J. Stokes, A. Macaskill, P.J. Lundahl, W.E. Smith, K. Faulds, D. Graham, Small 3 (2007) 1593–1601. [32] L.B. Luo, L.M. Chen, M.L. Zhang, Z.B. He, W.F. Zhang, G.D. Yuan, W.J. Zhang, S.T. Lee, J. Phys. Chem. C. 113 (2009) 9191–9196. [33] R.B. Jiang, H.J. Chen, L. Shao, Q. Li, J.F. Wang, Adv. Mater. 24 (2012) OP200–OP207. [34] J. Becker, I. Zins, A. Jakab, Y. Khalavka, O. Schubert, C. Sönnichsen, Nano Lett. 8 (2008) 1719–1723.

521