Simple Boronic Acid-Appended Sensor Array for Saccharides by Linear Discriminant Analysis

Abstract

Saccharides play important roles in human health and life. However, detecting and differentiating the saccharide types using one probe is difficult even under optimized conditions due to the similar structures of different saccharides. In this study, only one sensor was used to construct the array for the discrimination of different types of saccharides on the basis of a dialdehyde-diboronic acid-functionalized tetraphenylethene (TPE-DABA) under different pH conditions. By integrating the fluorescence responses of the probe with linear discriminant analysis (LDA), 12 kinds of saccharides and the polyhydroxy compound sorbitol could be distinguished and identified by as few as one sensor probe. This study provides a reference for developing more simple and effective sensor arrays for saccharides.

1. Introduction

Saccharides, also known as sugars and carbohydrates, play a key role in human health and life. Detection of saccharides is crucial for food quality control, clinical diagnostics, and metabolic disorder monitoring, and has gained increasing attention in the field of fluorescence sensing and detection [1,2,3]. However, owing to the small differences among different saccharides, designing “lock-key” mode fluorescent probes with high specificity only for a certain analyte is difficult. Only a limited number of probes selective for certain sugars have been reported [2,4]. Usually, these high-performance probes typically require elaborate design and laborious synthesis. Sensor arrays, which differ from highly selective probes, have the feature of simultaneous cross-reactivity to several components [5,6,7,8]. They can act like artificial chemical noses. When principal component analysis (PCA) and linear discriminant analysis (LDA) are introduced to treat corresponding response data, analytes can be successfully distinguished from each other [7,9,10,11].

On the basis of the ability of arylboronic acids to selectively recognize cis-diol-containing molecules (cis-diols), fluorescent boronic acid-functionalized compounds and materials have been elaborately designed for the detection and discrimination of saccharides [4,12,13,14,15,16,17,18]. For example, in our group, several probes and arrays with special structures were developed to effectively differentiate fructose, glucose, ginsenosides, and polyglycosylated protopanaxadiols [5,19,20,21,22,23]. Wang et al. used fluorescent boronate affinity probes to discriminate cis-diol-containing molecules via PCA [9]. For the majority of reported sensor arrays, a minimum of three sensor molecules is typically needed.

The spatial orientation of the vicinal diol moiety modulates the binding process. Importantly, the existing form of phenylboronic acids and the interactions between phenylboronic acids and saccharides vary according to the pH of the environments [24,25]. The boron atom of arylboronic acid can be ionized by OH⁻ in alkaline aqueous solutions [4]. Simultaneously, a higher pH benefits covalent binding between boronic acids and polyhydroxy compounds. We envisioned that through the rational design of a fluorescent sensor and an adjustment of the pH of the detection environments, sensor arrays capable of identifying various monosaccharides and disaccharides may be constructed with only one probe.

Aggregation-induced emission (AIE) luminogens (AIEgens) have the advantages of facile functionalization and a high signal-to-noise ratio [26,27,28,29]. Their fluorescence is greatly affected by their existing state. AIEgens are weakly luminescent or nonemissive in solution but highly emissive in aggregate form. The interactions between saccharides and AIEgens may affect the solubility and aggregation state of AIEgens, further affecting the fluorescence signals of AIEgens. Based on the above, in this study, a novel AIE molecule, i.e., dialdehyde-diboronic acid-functionalized tetraphenylethene (TPE-DABA), was utilized for the discrimination of different saccharides. TPE-DABA displays a change in fluorescence in different pH buffers (Figure 1). The phenylboronic acid moiety in TPE-DABA could act as a receptor and reversibly interact with carbohydrates to form a boronate complex. TPE-DABA can yield different responses to various saccharides at different pH values. At neutral pH, saccharides interacting with TPE-DABA could increase the solubility and decrease the fluorescence of TPE-DABA. At a strongly alkaline condition, TPE-DABA can be transformed into a borate salt (Figure 1). The fluorescence of the probe–sugar complex (TPE-DABA-III in Figure 1) is stronger than that of the soluble borate salt (TPE-DABA-I) due to the restriction of intramolecular motions. At weak alkaline condition, red-shifted fluorescence may be observed due to the D-π-A structure in TPE-DABA-I and TPE-DABA-III (Figure 1). Overall, a three-channel fluorescence assay for saccharides was achieved by employing TPE-functionalized diboronic acid, resulting in a fingerprint-like response pattern. This work provides a simple and direct way to distinguish and identify different types of sugars using only one probe.

2. Materials and Methods

The dialdehyde-diboronate-functionalized AIE luminogen TPE-DABA was synthesized according to previously published procedures [21]. Britton–Robinson buffer solutions were prepared by adding different amounts of 0.2 M NaOH to a mixed solution of acids (phosphoric acid, 0.04 M; acetic acid, 0.04 M; boronic acid, 0.04 M). A stock solution of TPE-DABA at a concentration of 2.5 mM was prepared by dissolving TPE-DABA in DMSO, and the solution was stored in a refrigerator at 4 °C. Deionized water was used to prepare all aqueous solutions. Saccharides and sorbitol stock solutions were freshly prepared every time before use. pH 7.4, 9.25, and 10.5 were selected to represent neutral, weakly alkaline, and alkaline conditions, respectively. The mixtures of saccharides at different concentrations and 50 µM TPE-DABA in 0.1 M sodium phosphate buffers at pH 7.4, or 0.1 M carbonate buffers at pH 9.25 and 10.5 were allowed to stand for 30 min before fluorescence measurements. The solutions of TPE-DABA (50 µM) and analytes (sorbitol and 12 kinds of saccharides) at different concentrations contained 2 vol% dimethyl sulfoxide (DMSO). Fluorescence tests were performed on an FS5 spectrofluorometer (Edinburgh Instruments). The fluorescence spectra were measured in 1-cm × 1-cm quartz cells at an excitation wavelength of 380 nm. The fluorescence intensities at 495 nm at pH 7.4 and 595 nm at pH 9.25 and 10.5 were used for data treatment.

3. Results and Discussion

TPE-DABA comprising two phenylboronic acid moieties and two aldehyde groups was first synthesized by our group in six steps [21]. Phenylboronic acid in water exists as an equilibrium mixture of the nonionic trigonal boronic form PhB(OH)₂, and the ionic tetrahedral boronate form PhB(OH)₃⁻ [4,30]. We carried out pH titration experiments with the TPE-DABA sensor in Britton–Robinson buffer solutions. As depicted in Figure S1, the initial fluorescence of TPE-DABA is strong under acidic and neutral conditions but greatly decreases when the pH further increases. Under acidic and neutral conditions, the solubility of TPE-DABA is poor, and the fluorescence mechanism is mainly the restriction of intramolecular motions caused by aggregation. In contrast, TPE-DABA becomes a soluble ionic form (TPE-DABA-I in Figure 1) in buffers with relatively high pH values, resulting in a significant decrease in fluorescence emission.

Notably, the pH of the aqueous medium also strongly affects the interactions between phenylboronic acid and sugars [31]. Then, we used the fluorescent probe TPE-DABA to sense various sugars at different pH values. As discussed above, we envisioned that saccharides with a strong binding interaction with TPE-DABA could induce decreased fluorescence due to its effect on the solubilization of TPE-DABA at the neutral condition, increased fluorescence at the strongly alkaline condition, and redshift at the weak alkaline condition. The stronger the binding force, the greater the response amplitude.

We carried out titration experiments with different saccharides under neutral conditions (pH 7.4). As shown in Figure S2, the fluorescence response behaviors were different. The fluorescence spectra of the mixtures of most sugars and TPE-DABA partially overlap (Figure 2). This behavior occurs because the sugar complex with trigonal PhB(OH)₂ is highly susceptible to hydrolysis under neutral conditions, and the interactions between TPE-DABA and most sugars are weak. Only fructose and sorbitol can markedly quench the initial strong fluorescence of the sensor (Figure S3), which can be attributed to their strong binding affinity with tetraphenylethene-functionalized phenylboronic acids (TPE-DABA) and the dissolution enhancement for TPE-DABA. The increased solubility after binding with soluble fructose and sorbitol can decrease the degree of aggregation and the fluorescence intensity of the sensor based on the AIE effect.

On the basis of previous reports [4], in contrast to the pyranose form for monosaccharides, boronic acid prefers to bind with the five-membered furanose form to minimize the angle strain. The response selectivity for monosaccharides is closely correlated with the relative abundance of their furanose form, which contains a syn-periplanar adjacent hydroxyl pair in aqueous solutions. For example, approximately 25% of dissolved fructose exists in the reactive five-membered furanose form. In contrast, the proportions of the corresponding galactofuranose, mannofuranose, and glucofuranose in aqueous solutions are only 2.5%, 0.3%, and 0.14%, respectively [4]. Owing to the existence of a pair of cis-diol units in glucose, the corresponding diboronate complex may occur with a cis-1,2-diol unit or a cis-5,6-diol unit of glucose [30]. The highest selectivity towards sorbitol can be attributed to the flexibility of its non-cyclic structure. TPE-DABA also has a better fluorescence response to glucosamine and sialic acid, possibly because the nitrogen atoms of glucosamine and sialic acid donate electrons to boron [32]. This enables the conversion of nucleophilic sp²-hybridized boron into an sp³ configuration under neutral conditions. The corresponding proposed structures of the arylboronic acid/sugar complex are shown in Figure S4. Furthermore, the binding affinity between saccharides and TPE-DABA is not the only decisive parameter determining the fluorescence response. It may also be due to the better solubilizing effect of disaccharides than of monosaccharides, as some disaccharides can result in remarkable fluorescence responses.

pH is an indispensable factor for producing diverse fluorescence signals in response to different saccharides. Sugars generally form stable complexes with tetrahedral PhB(OH)₃⁻ in aqueous solutions. We then investigated the dose–response behaviors of the probe to sugars at pH 10.5. The sensor with a negatively charged tetrahedral form (TPE-DABA-III in Figure 1) not only has lower background fluorescence due to its better water solubility but also becomes more reactive to saccharides. The negatively charged boron- and electron-withdrawing aldehyde groups in the TPE core can create a D-π-A architecture (TPE-DABA-I and TPE-DABA-III). The corresponding fluorescence wavelength red-shifts relative to that at pH 7.4 (Figure 1). On the basis of boronate affinity interactions, the presence of bulky substituents can restrict intramolecular motions to a certain degree, resulting in fluorescence enhancement upon the addition of most testing saccharides (Figure S5). The relative fluorescence ratios are highly correlated with the binding force between saccharides and TPE-DABA (Figure 3). As expected, among all the analytes, fructose and sorbitol induced a significantly larger enhancement in fluorescence intensity than the other saccharides (Figure S6). The fluorescence responses follow the order: sorbitol, fructose > galactose, glucose, mannose, ribose > maltose, sialic acid, lactose, cellobiose > trehalose, sucrose, glucosamine. This sequence almost agrees with the trend in affinity reported for saccharide binding with phenylboronic acid [4].

Furthermore, we also performed similar dose–fluorescence response experiments with sorbitol and 12 saccharides as analytes at pH 9.25. Notably, certain rules exist behind the plausibly ruleless fluorescence responses (Figure S7). The fluorescence-response behaviors of the mixtures of TPE-DABA and various saccharides at pH 9.25 are not the same. The different wavelengths can be ascribed to efficient multivalent interactions with hydroxyl ions and polyhydroxy compounds. We propose a plausible mixed structure of TPE-DABA-I, TPE-DABA-II, and TPE-DABA-III, as shown in Figure 1. With an increasing concentration of sugars or a binding affinity with sugars, the fluorescence initially decreases with a redshift, and then increases at longer wavelengths (Figure 4 and Figure S7). For example, TPE-DABA gives rise to more prominent fluorescence at longer wavelengths in response to fructose and sorbitol. The fluorescence response amplitudes observed with various saccharides are consistent with those obtained at pH 10.5. The stronger the interaction between the sensor and saccharides is, the greater the redshift.

Overall, the pH of the aqueous medium markedly affects the fluorescence of TPE-DABA and the interactions between TPE-DABA and saccharides. For saccharides with strong binding interactions with TPE-DABA, the fluorescence spectrum decreases under neutral conditions, increases under strongly alkaline conditions, and redshifts under weakly alkaline conditions. Although it cannot be used to discriminate saccharides in a single pH environment, this cross-reactivity “weakness” can be converted into a sensor array. The integration of three fluorescence channels (I/I₀ @495 nm @pH 7.4, I/I₀ @595 nm @pH 9.25, and pH 10.5) provides fluorescence signal output for use in fluorescence-based sensor arrays and enables excellent discrimination of saccharides via LDA. A “fingerprint” spectrum was obtained with the concentration of analytes set at 6 mM. As shown in Figure 5, 65 data points, which belong to 13 analytes, were divided into 13 groups in the 2D canonical score plot. The first two canonical factors carried about 74.2% and 20.4% of the total variance, respectively. The well-separated two-dimensional LDA plot demonstrates that the sensor array based on three pH environments provides satisfactory discrimination of different types of sugars with similar structures (Figure 5). Interestingly, for common saccharides, the right-to-left sequence in Figure 5 follows: fructose > galactose, glucose, mannose > maltose, cellobiose, lactose, sucrose, trehalose. Similarly, the maximum emission wavelengths at pH 9.25 (Figure 4) are ordered from right to left as follows: fructose > galactose, glucose, mannose > maltose, lactose, cellobiose, trehalose, sucrose. The ranking also aligns with the value of I/I₀ @595 nm @pH 10.5 (Figure S6) minus I/I₀ @495 nm @pH 7.4 (Figure S3). The aforementioned sequences also resemble the order of the binding affinities with phenylboronic acids [4,24,33,34,35].

4. Conclusions

The detection and discrimination of saccharides is highly challenging due to the similar structures of different saccharides. In this work, we cleverly constructed a sensing array using only one phenylboronic acid-containing optical sensor, i.e., TPE-DABA, based on the fact that the interactions between phenylboronic acid and saccharides depend on the pH. Upon incubation with various saccharides at different pH values, the aggregation state and fluorescence of TPE-DABA change to various degrees. The changes in the fluorescence signals depend on the saccharides. When LDA was employed to analyze these fluorescence response data, TPE-DABA was found to identify sorbitol, seven monosaccharides, and five disaccharides with 100% accuracy. This work provides a way of thinking to construct multichannel arrays with fewer sensors only by changing the test environment. In the future, highly selective probes and effective sensor arrays for glycoproteins, glycolipids, and oligosaccharides will still be required so as to acquire the detailed distribution and functions of sugars and related biological macromolecules and further facilitate corresponding clinical diagnosis and diseases treatment.