Design a New Strategy Based on Nanoparticle-Enhanced Chemiluminescence Sensor Array for Biothiols Discrimination

Array-based sensor is an interesting approach that suggests an alternative to expensive analytical methods. In this work, we introduce a novel, simple, and sensitive nanoparticle-based chemiluminescence (CL) sensor array for discrimination of biothiols (e.g., cysteine, glutathione and glutathione disulfide). The proposed CL sensor array is based on the CL efficiencies of four types of enhanced nanoparticle-based CL systems. The intensity of CL was altered to varying degrees upon interaction with biothiols, producing unique CL response patterns. These distinct CL response patterns were collected as “fingerprints” and were then identified through chemometric methods, including linear discriminant analysis (LDA) and hierarchical cluster analysis (HCA). The developed array was able to successfully differentiate between cysteine, glutathione and glutathione disulfide in a wide concentration range. Moreover, it was applied to distinguish among the above analytes in human plasma.

clue to this problem, in the present contribution, we combined nanoparticle enhanced CL system with the sensor array technique.
Array-based sensors have become an interesting set of tools for analytical studies and their application continues to extend rapidly 14 . In recent years, many sensor array designs have been applied by various research groups for sensing and identifying different chemicals, such as biomacromolecules (Rotello et al. 15 and Hamilton et at. 16 ), vapors (Suslick et al. 17 , Lewis et al. 18 , Walt et al. 19 , and Rose et al. 20 ), and ions (e.g., Anslyn et al. 21 , Anzenbacher et al. 22 , and Wolfbeis et al. 23 ). Chemosensors are slowly shifting toward sensor arrays leaving behind expensive and non-versatile selective sensors. In chemical sensor array, also called chemical noses or chemical tongues, the specificity of the identification comes from the response pattern derived from the sensor elements. In sensor arrays, each non-specific sensor element has a special amount of selectivity to each analyte. By using a pattern recognition protocol, signals of all sensor elements are simultaneously analyzed and a fingerprint-like response pattern is generated for each analyte. These unique response patterns are used for the classification and identification of a set of analytes. A large number of chemical array sensors use changes in the optical properties of the sensing elements to provide distinct response patterns. For example, these optical properties may consist of changes in the absorption spectra (color) and the emission intensity. Recently a few number of chemiluminescent (CL) sensor arrays have been reported that are quite interesting. For instance, Cui et al. reported a new nanoparticle-based chemiluminescent sensor array for protein discrimination. In their work CL intensity, time for CL emissions to appear, and the time to reach the CL peak value were used as the sensing elements. Their system provided distinct response patterns which were discriminated using classical principal component analysis (PCA) 24 . A similar system was used for pesticide discrimination by He et al. 25 . Moreover Zhang et al. designed a sensor array based on nanomaterial-assisted thermochemiluminescence (TCL) for protein sensing and cell discrimination. They reported that the TCL of their analytes was generated on the surface of catalytic nanomaterials. Therefore, six catalytic nanomaterials, γ -Al 2 O 3 , Pt/Ba/Al, MgO, MgCO 3 , SrCO 3 , and ZrO 2 , were used as sensing elements. Unique TCL response patterns were obtained from the array and were analyzed by linear discriminant analysis (LDA) 26 .
In this study, three common biothiols for which we have recently introduced a colorimetric sensor array 27 , were chosen as model analytes in order to discriminate among them. The biothiols of interest included glutathione (GSH), glutathione disulfide (GSSG), and cysteine (Cys). These compounds have fundamental roles in biological processes, such as reversible redox reactions and enormous cellular functions such as metabolism and detoxification. Any change in the biothiols redox status, particularly for small molecular biothiols such as cysteine, glutathione disulfide, and glutathione can influence the physiological properties 28 .
Over the past decades, various analytical methods have been developed for the detection of biothiols, for instance, high performance liquid chromatography (HPLC) 29 , capillary electrophoresis 30,31 , mass spectrometry 32 , fluorometry 33,34 and colorimetry 35 . Though all these methods show good capability for the detection of different thiols, they suffer from a variety of disadvantages, such as the use of sophisticated, expensive instruments, time-consuming and preliminary treatment. Several enzymatic methods have also been applied to determine biothiols 36,37 . Although these methods have good selectivity, the chemical stability of the enzymes is poor, which greatly restricts their simple use. Due to various advantages such as simple instrumentation, ease of control, rapidity in signal detection, and simple operation, we chose chemiluminescence (CL)-based assays for our study. In addition in comparison with reported array based systems 27,38 we achieved biothiols discrimination in a lower and wider concentration range. Hence, we developed an approach, which holds the promise of achieving both simplicity and sensitivity.
Different nanomaterials with different CL reagents were chosen, for the first time, as sensing elements. We used luminol-AgNO 3 enhanced chemiluminescence (CL) in the presence of NaBH 4 coated AuNPs (BH 4 -AuNPs) and citrate-capped AuNPs (Cit-AuNPs) in alkaline solution and luminol-H 2 O 2 enhanced chemiluminescence (CL) in the presence of citrate coated AgNPs (Cit-AgNPs) and thioglycolic acid (TGA) functionalized cadmium telluride (CdTe) quantum dots (TGA-CdTe QDs) in alkaline solution as sensing elements. Combining nanomaterial-assisted CL probes with the sensor array design led to a selective and powerful approach for the discrimination of important biothiols. The developed CL sensor array produced distinct response patterns for the analytes of interest in a wide concentration range. Figure 1 shows a schematic diagram of the experimental setup. For discrimination purposes, chemometric methods, including linear discriminant analysis (LDA), and hierarchical cluster analysis (HCA), were employed.

Result and Discussion
Enhancement effect of nanoparticles on luminol CL. In a preliminary study, the enhancement effect of different nanoparticles on the luminol-H 2 O 2 and luminol-AgNO 3 CL was investigated within a flow injection system. As shown in Fig. 2, the results indicated that the CL signal of the luminol-H 2 O 2 system was significantly enhanced by AgNPs and CdTe QDs, while the CL intensity of luminol-AgNO 3 reaction was significantly enhanced by BH 4 -AuNPs and Cit-AuNPs. Therefore, we used these four nanoparticle-enhanced CL systems as sensor elements to generate response patterns. It is well-known and has been widely studied, that metal ions can catalyze the decomposition of H 2 O 2 and therefore catalyze luminol-H 2 O 2 CL systems. AgNPs are also able to efficiently catalyze the H 2 O 2 decomposition. As shown in Fig. S1 H 2 O 2 was decomposed by AgNPs to • OH radicals, which further reacted with H 2 O 2 to generate superoxide anions (O 2 •− ) and luminol radicals. Furthermore, the mechanism of luminol-H 2 O 2 CL system was reported due to the reaction of luminol radical with superoxide anion 2,13 . Silver nitrate is capable of oxidizing luminol but it doesn't produce a strong CL signal under general conditions. In the presence of gold nanoparticles, luminol could react with AgNO 3 and produce strong CL signal. As shown in Fig. S2, AgNO 3 was reduced to Ag atoms and luminol was oxidized to luminol radical which were catalyzed by gold nanoparticles and happened on the surface of gold nanoparticles. Luminol radical was then followed by reaction with the dissolved oxygen, giving rise to light emission 12,39-41 . The possible mechanism of the enhanced luminol-H 2 O 2 CL reaction induced by CdTe QDs was shown in Fig. S3 In the present system, we suppose that the H 2 O 2 oxidant can oxidize CdTe QDs, producing cationic radicals R +• , or oxidized species (CdTe QDs) +• , which can act as enhancers for the luminol-H 2 O 2 CL system to generate strong CL radiation 42,43 . Optimization of experimental conditions. The experimental conditions were optimized for the luminol-H 2 O 2 -Cit-AgNPs and luminol-AgNO 3 -Cit-AuNPs CL systems. Considering the CL intensity, standard deviation, and the amount of reagent consumption, the optimal conditions for CL systems were utilized for further experiments. As shown in Figs S4

Inhibition and enhancement effects of biothiols on nanoparticle-based chemiluminescence.
Organic compounds containing amino (NH 2 ), hydroxyl (OH), and mercapto (SH) groups have been found to interact with nanoparticles 2 . The results demonstrate that such compounds interact with nanoparticles in the CL reaction leading to a change in CL intensity. In the present work, the effects of these biomolecules on the luminol-AgNO 3 -AuNPs and the luminol-H 2 O 2 -(AgNPs/CdTe QDs) CL systems has been investigated. Biothiols  in wide concentration ranges were exposed to different types of NPs and the signals were recorded. The results are illustrated in Figs 3, S6, S7, and S8. As shown in Fig. 3, by increasing GSH concentration, the CL intensity of luminol-H 2 O 2 -QDs and luminol-H 2 O 2 -AgNPs increased linearly, while the CL intensity of luminol-AgNO 3 -(Cit-, BH 4 -) AuNPs decreased. As shown in Fig. S6, the results indicated that no significant changes in CL intensity of luminol-H 2 O 2 -AgNPs was observed by the addition of GSSG. It could be seen that GSSG has an inhibition effect on luminol-AgNO 3 -(Cit-, BH 4 -) AuNPs. Fig. S8(a1, b1, c1) clearly indicates that in luminol-H 2 O 2 -QDs CL system, the most intensive CL signal was obtained from GSH. Figure S7 shows that an opposite effect was observed for Cys: when the Cys concentration was lower than 35 μ M, the CL intensity of luminol-H 2 O 2 -BH 4 -AuNPs increased with increasing Cys concentration; when Cys concentration was higher than 35 μ M, the CL intensity decreased with increasing Cys concentration. A similar effect was observed for luminol-H 2 O 2 -Cit-AuNPs. Based on previous reports 41, [44][45][46][47][48] we suggested that the enhancement effect of low concentrations of Cys on luminol-AgNO 3 -AuNPs reaction may be due to the change of the gold nanoparticles surface-charge density. It is obvious that the negative charge density of AuNPs surface decreases after aggregation which occurs in the presence of Cys. On the other hand, luminol anion which is the molecular form of luminol in alkali media, can not easily interact with anionic AuNPs because of the electrostatic repulsion. Albeit it can more easily interact with aggregated AuNPs, causing a higher catalytic effect on the CL reaction. Thus, we reasoned that the enhanced catalysis of AuNPs occurring in low concentrations of Cys resulted from the decrease in the surface negative charge density of AuNPs. In contrast, in higher concentrations of Cys, an opposite effect was observed which is in agreement with previous reports stating that organic compounds such as Cys interact with gold nanoparticles in the CL reaction leading to an inhibition of CL intensity 2,13,49 . According to the proposed mechanism, reducing groups such as SH

Application of NP-enhanced CL system for biothiol discrimination. The responses of various CL
systems consisting four types of NPs and reagents to three different biothiols are shown in Fig. S8. The results revealed that the three biothoils with different molecular structures provided different binding capabilities toward nanoparticles, leading to a meaningful differences in the values of CL signals. CL signals are distinct, suggesting the possibility of biothiols discrimination using such a sensor array. The proposed sensor array was made of four sensor elements, and each of them shows different CL signal changes after interaction with biothiols. The Δ I patterns were subjected to linear discriminant analysis (LDA). These patterns were transformed to canonical scores which were visualized as a well-clustered two-dimensional plot as shown Fig. 4. We successfully discriminated biothiols solutions, for GSSG and GSH in the concentration range of 5.0-800.0 μ M and for Cys in the concentration range of 25.0-100.0 μ M by utilizing the training matrix generated from CL patterns. Three replicates were tested for each biothiol sample. They appeared exactly to their respective groups. Our system provided a sensitive way to differentiate biothiols at wide concentration ranges. The LDA method classified the biothiols into three distinct classes, revealing that nanoparticle-based chemiluminescence sensor array was capable of discriminating various biothiols.
By implementing hierarchical cluster analysis (HCA), a statistical classification method, the similarity of biothiols was analyzed. As shown in Fig. 5, the three biothiols appeared exactly to their respective groups.
Additionally, we achieved successful discrimination of individual thiols within their mixtures. Binary and ternary mixtures of thiols (at concentrations of 75.0 μ M for each thiol) were tested. As shown in Fig. 6, LDA analysis on the mixtures of biothiols demonstrated that the canonical response patterns of the individual thiols and their mixtures were clearly clustered into seven groups.

Fingerprints of biothiols generated via the CL sensor array based on various NPs. Fingerprints
of Cys, GSH, and GSSG based on Δ I values were obtained from four nanoparticle coupled CL systems of luminol-H 2 O 2 -CdTe QDs, luminol-H 2 O 2 -Cit-AgNPs, luminol-AgNO 3 -Cit-AuNPs, and luminol-AgNO 3 -BH 4 -AuNPs. Δ I = I − I 0 , where I and I 0 refer to the value of CL intensity in the presence and absence of biothiols, respectively. As shown in Fig. 7, the results confirmed that different biothiols affect CL responses with different degrees, leading to characteristic response patterns. We also employed the heat maps to visualize the ability of that this sensor in generating characteristic fingerprints for each of the three biothiols. A heat map is a 2D representation of data in which individual values in a matrix are represented as colors. As a visual tool, heat maps can be very powerful because it provides an immediate visual summary of information of data values that would be much harder to understand if presented numerically. The color key (right side of the Fig. 8) is a continuous range. In the color scale, a dark blue is for the lowest value and a dark red is for the highest value. A color range moves from blue to red as the values increase, giving the appearance of getting hotter. The heat map depicted in  Blood plasma analysis. In order to verify the applicability of the developed method, the proposed procedure was applied for the analysis of analytes in real sample. Human plasma sample was spiked with different concentrations of analytes. As shown in Fig. 9, LDA analysis demonstrated that the proposed array could successfully discriminate between cysteine, glutathione and glutathione disulfide in the concentration range of 5.

Conclusions
In summary, we have designed a chemiluminescence sensor array on the basis of distinct CL responses patterns generated by different nanoparticles in order to discriminate biothiols for the first time. The CL sensor array can successfully discriminate among three selected biothiols in a wide concentration range. The developed sensor array based on various NPs as sensor elements, opens a new window to the use of nanoparticle-based CL systems constructing various sensing arrays. We believe by varying the CL reagents or NPs as sensor elements, the nanoparticle-based chemiluminescence sensor array can emerge new opportunities for various biomolecules discrimination.
Instrumentation. UV-Vis spectra were measured and recorded with a Lambda (Perkin Elmer, USA) spectrophotometer with the use of 1.0 cm cell. The transmission electron microscopy (TEM) images were taken using a (PHILIPS MC 10 TH microscope at an acceleration voltage of 100 kV). Size distributions of the particles were measured using Zetasizer Viscotec 802 at ambient temperature.

Synthesis of different NPs. Synthesis of citrate-coated AuNPs.
Cit-AuNPs with a concentration of 8.16 nM were prepared according to Turkevich method 50 . In regard to the modified Turkevich method, after bringing 50.0 mL of HAuCl 4 solution (1.0 mM) to boil, 5.0 mL of sodium citrate (38.8 mM) was added. The mixture was refluxed for 30 min. Then ice-cold water was quickly used for cooling it and finally stored at 4 °C temperature. The synthesized AuNPs were characterized by TEM, DLS, and UV-vis spectroscopy, for which the results are shown in Figs S9a, S10a and S11a.
Synthesis of NaBH 4 coated AuNPs. BH 4 -AuNPs with a concentration of 1524 nM were prepared as follow 51 . Briefly, 100.0 μ L 50.0 mM mixture solution of HAuCl 4 and HCl was added, with vigorous stirring, to 9.6 mL DI water. Then, 400.0 μ L 50.0 mM mixture solution of NaBH 4 and NaOH was added and stirred for 15 min. The synthesized AuNPs were characterized by TEM, DLS, and UV-vis spectroscopy, and the results are shown in Figs S9b, S10b and S11b.
Synthesis of citrate-capped AgNPs. Cit-AgNPs with a concentration of 1.67 nM were synthesized by the reduction of AgNO 3 solution using citrate according to a previous report 52 . Typically, 250.0 μ L 100.0 mM AgNO 3 solution was added to 100.0 mL DI water. Then, 250.0 μ L 100.0 mM sodium citrate solution and then 1.0 mL 5.0 mM sodium borohydride solution were respectively added to the solution and it was stirred for 30 min. The resulting mixture was stored in the dark for 24 h before use. The synthesized AgNPs were characterized by DLS, and UVvis spectroscopy, and the results are shown in Figs S10d and S11c.  Fig. S10c. Fig. 1 illustrates the laboratory-built flow injection CL system employed in this work. The NPs solution was injected into the ultrapure water as a carrier stream through a 100.0 μ L loop-valve injector, mixed with luminol, and H 2 O 2 (or AgNO 3 ) solutions through three-way pieces. Then, the mixed solution moved to a spiral-shaped flow cell which was located in front of a photomultiplier tube (PMT). The PMT signal was imported to the computer and the CL signal was monitored.

Procedure of detection. The diagram shown in
Fabrication of array-based sensing system for biothiols discrimination. For discrimination of biothiols, three biothiols with concentrations ranging from 5.0 to 800.0 μ M were tested against the four NPs. Three replicates were tested for each of them. The variations in CL intensity (Δ I = I − I 0 ) were used to generate the response pattern, where I and I 0 are in order the value of CL signal in the presence and absence of the biothiols.
The stock solutions of three biothiols were prepared in DI water. For each biothiol, AgNPs solution (50.0 μ L) was added to 100.0 μ L biothiol solution and TGA-CdTe QDs solution (30.0 μ L) was added to 90.0 μ L biothiol solution. Either AgNPs or QDs-biothiols solutions were injected into the carrier stream (ultrapure water) through a 100.0 μ L loop-valve injector, mixed with H 2 O 2 and luminol solutions through three-way pieces. Then the mixed solution flowed via a spiral-shaped flow cell positioned in front of a photomultiplier tube. In addition, the BH 4 -AuNPs solution (20.0 μ L) was added to biothiol solution (140.0 μ L) and Cit-AuNPs solution (50.0 μ L) was added to biothiol solution (100.0 μ L). The AuNPs-biothiols solutions were injected into the carrier stream (ultrapure water) through a 100.0 μ L loop-valve injector, mixed with AgNO 3 and luminol solutions through three-way pieces.
Preparation of human plasma samples. Trichloroacetic acid (TCA) was added to the freshly thawed human plasma sample and the mixture was stirred vigorously to precipitate proteins. After centrifuging at ~10,000 rpm for 10 min at room temperature, the clear supernatant was diluted 20 times and the subsequent was subjected to assay under the optimized conditions.