2.1 Materials

All the reagents and solvents were commercially available and used as received unless otherwise specified purification. Ten batches of QZC were obtained from Shaanxi Buchang Pharmaceutical Co., Ltd. The batch numbers are 180101, 180102, 180103, 180104, 180105, 180106, 180107, 180108, 180109 and 180110. Ten batches of Shuanghuanglian (SHL) injection were provided by Henan Fusen Pharmaceutical Co., Ltd. The batch numbers are 2002241, 2002251, 2002261, 2002271, 2002281, 2002291, 2002301, 2002311, 2002321 and 2002331. Carboxylated azocalix[4]arene (CAC4A), sulfonated azocalix[4]arene (SAC4A), sulfonated azocalix[5]arene (SAC5A) and sulfonated azocalix[8]arene (SAC8A) were synthesized and purified according to the literature procedures (Zhang et al., 2020, Zhang et al., 2021, Yue et al., 2022). The syntheses and characterizations of quaternary-ammonium modified azocalix[4]arene (QAAC4A) and tertiary-ammonium modified azocalix[4]arene (TAAC4A) are shown in Supplementary Information. Al(III) phthalocyanine chloride tetrasulfonic acid (AlPcS₄) was obtained from Frontier Scientific. β-cyclodextrin (β-CD) and 2-hydroxypropyl-β-cyclodextrin (HP-β-CD) were purchased from Heowns Biochemical Technology. Cucurbit[7]uril (CB7) was purchased from Beijing Hwrk Chemicals. 8-anilino-1-naphthalenesulfonate (1,8-ANS) and 2-(p-toluidinyl)naphthalene-6-sulfonic acid (2,6-TNS) were purchased from Tokyo Chemical Industry. 4′,6-diamidino-2-phenylindole (DAPI) was purchased from Sigma-Aldrich. The phosphate buffered saline (PBS, pH = 7.4, 10 mM) was purchased from Macklin. Methanol (HPLC grade) was purchased from Aladdin. Calixarenes and AlPcS₄ were dissolved in ultrapure water prepared by Thermo Scientific purification system, except for CAC4A and TAAC4A which were dissolved in PBS.

2.2 Instruments

Steady-state fluorescence spectra were recorded in a conventional quartz cell (light path 10 mm) on a Varian Cary Eclipse spectrometer (Agilent Technologies Inc., USA) equipped with a Varian Cary single-cell peltier accessory to control temperature. The ultrasonic cleaner used to extract the ingredients is SW 6H of SONO SWISS. High speed refrigerated centrifuge was purchased from Sichuan Shuke Instrument Co., Ltd. QUINTIX35-1CN one hundred thousandth electronic balance was purchased from Sartorius Scientific Instruments (Beijing) Co., Ltd.

2.3 Extraction method

The powder in the capsules (500.0 mg) precisely weighed was transferred into a 25.0 mL volumetric flask and then 75% methanol aqueous solution was added to the flask until it reached the scale line. The mixture was ultrasonically extracted at 60 °C for 30 min and then cooled to room temperature. The extracted solution was centrifuged at 12700 rpm for 10 min to obtain the QZC extract supernatant at a concentration of 20.0 mg/mL, which was subsequently diluted for use in the analysis system.

2.4 Fluorescence spectroscopy

For CAC4A, SAC4A, SAC5A, SAC8A, QAAC4A, TAAC4A sensing elements, the QZC extract (2.5 μL) was diluted into water (2.5 mL) in a quartz cuvette firstly. Then, one of the calixarene solutions (0.2 mM CAC4A 10.0 μL, 1.0 mM SAC4A 2.5 μL, 1.0 mM SAC5A 5.0 μL, 1.0 mM SAC8A 5.0 μL, 0.5 mM QAAC4A 20.0 μL, and 0.5 mM TAAC4A 20.0 μL) was added and thoroughly mixed. The equilibrium time is three minutes for CAC4A and SACnAs, and ten minutes for QAAC4A and TAAC4A. I₀ and I are the fluorescence intensities of QZC extract before and after addition of calixarene at 403 nm under excitation at 325 nm.

For QAAC4A/AlPcS₄ and TAAC4A/AlPcS₄ sensing elements, the fluorescence signal of AlPcS₄ was used to indicate the concentration of compounds from QZC through the process of competitive complexation. Firstly, the fluorescence intensity of AlPcS₄ (1.0 μM, 2.5 mL) as I₀ (excitation wavelength at 610 nm and emission wavelength at 680 nm) was recorded. Then the calixarene solution (0.5 mM QAAC4A 5.0 μL or 0.5 mM TAAC4A 5.0 μL) and the QZC extract at 20.0 mg/mL (10.0 μL for QAAC4A/AlPcS₄ sensing or 5.0 μL for TAAC4A/AlPcS₄ sensing) were added. After incubating for ten minutes, the fluorescence intensity at equilibrium was measured as I. Fluorescence intensity ratios (I/I₀) were calculated and recorded.

2.5 Methodological evaluation

2.6 Discrimination of different concentrations

AlPcS₄ was added to a black 96-well plate to obtained the final concentration at 1.0 μM and fluorescence intensity (I₀) was measured on a microplate reader. Then QZC extract and QAAC4A or TAAC4A were introduced to give final concentrations from 22.5 to 90.0 μg/mL for QZC extract and 1.0 μM for the employed calixarenes, which were mixed and incubated for 10 min at 25 °C. The fluorescence intensity of each well was recorded (I). The obtained relative fluorescence intensities (I/I₀) were used as the response signals for array sensing analysis. Each set of target experiments was repeated six times. Finally, the raw data matrix was handled using linear discriminant analysis (LDA) in Past 3 program.

2.7 Fluorescence spectroscopy for SHL injection

β-CD and 2,6-TNS were diluted into water (2.5 mL) in a quartz cuvette to obtained the final concentrations at 800.0 μM and 10.0 μM, respectively. I₀ and I are the fluorescence intensities before and after 5.0 μL SHL injection addition to the solutions at 470 nm under excitation at 350 nm.

HP-β-CD and 2,6-TNS were diluted into water (2.5 mL) in a quartz cuvette to obtained the final concentrations at 1.0 mM and 10.0 μM, respectively. I₀ and I are the fluorescence intensities before and after 5.0 μL SHL injection addition to the solutions at 446 nm under excitation at 350 nm.

CB7 and 1,8-ANS were diluted into water (2.5 mL) in a quartz cuvette to obtained the final concentrations at 1.0 mM and 10.0 μM, respectively. I₀ and I are the fluorescence intensities before and after 4.0 μL SHL injection addition to the solutions at 490 nm under excitation at 350 nm.

CB7 and DAPI were diluted into water (2.5 mL) in a quartz cuvette to obtained the final concentrations at 2.0 μM and 1.0 μM, respectively. I₀ and I are the fluorescence intensities before and after 5.0 μL SHL injection addition to the solutions at 470 nm under excitation at 361 nm.

3.1 Principle of supramolecular sensor array

The underlying reason that the sensor array can evaluate TCM is that it can give specific fingerprints for qualified samples and unqualified samples. QZC extracts contain compounds that show fluorescence emission under the excitation wavelength of 325 nm (Fig. S6). If part of the fluorescence could be altered after the macrocyclic receptor complexes the fluorescent substance in the extract, this would lead to a changed output signal by constructing direct sensing. For qualified samples, the starting fluorescence signals (only the sample) were consistent, and when the complex system reached the equilibrium, the final output signal should also be consistent. To explain the discrimination principle of sensor arrays more clearly, we propose a theoretical model based on host−guest recognition. As shown in Table S1, we considered that the analyte (guest) formed a 1:1 host−analyte complex with a host at an association constant (K_a), which satisfied the respective law of mass action relating to the equilibrium concentrations of free host, free analyte, and host-analyte complex. The addition of macrocycles caused a change in the concentration of the free analyte, which in turn resulted in a change in the fluorescence signal of the system. The components or concentrations of unqualified products differ from those of qualified products. Therefore, we simulated unqualified products by assigning different K_a (simulating the case of different components) and analyte concentrations (simulating the case of different concentrations). According to the simulated calculations, if two analytes have K_a values of 1 × 10⁵ and 5 × 10⁵ M⁻¹ with the same macrocycle, the fluorescence change ratios they produce at the same concentration are 0.62 and 0.27; if the concentrations of specific analytes are 1.0 and 0.5 μM, then the fluorescence change ratios of the system after complexation are 0.62 and 0.56, respectively (Table S1). The well-defined binding constants between the supramolecular host and guest imply that the specific macrocycles can produce reliable output signals for the qualified samples under the same conditions. In addition, the inconsistent components or concentrations of qualified and unqualified products will inevitably result in different output signals.

Another regularly used method is indicator displacement assay (IDA). The competitive binding of the indicator and the analyte to a receptor results in the fluorescence change of the indicator, thereby enabling the translation of the microcosmic behavior of molecular recognition into the macroscopic signal (You et al., 2015). Compared with direct sensing, the fluorescence changes relying on the IDA strategy can better reflect the content information of analytes lacking chromophores in QZC, which is complementary and even more applicable to other TCM that do not emit fluorescence. Data simulations are presented in Table S2. Assuming that the dye concentration stays within the linear range of fluorescence intensity versus concentration, then the ratio of fluorescence intensity before and after adding the analyte can be predicted. Simulations show that if the binding constants of the two competitive guests (K_c) to the macrocycle are 1 × 10⁷ and 5 × 10⁶ M⁻¹, the IDA-based fluorescence response ratios are 2.54 and 2.07; if the concentrations of a specific analyte are 1.0 and 0.5 μM, the fluorescence response ratios are 2.54 and 1.91, respectively (Table S2). Both the host−dye and host−analyte have fixed binding constants, so that different concentrations of analytes result in varying degrees of dye displacement, leading to characteristic fluorescence response ratios. Although our data simulations focus on a single analyte, every guest in the complex system, QZC, follows the same binding pattern. There is a specific apparent binding constant between qualified QZC mixture and the host, and composition and concentration anomalies in unqualified products can lead to signal differences.

Due to the difference of binding abilities between the receptors and compounds from QZC, the degrees of complexation of the same sample by different receptors are different, which are reflected in the difference of fluorescence signal. Combining all sensing elements to form an array provides multi-dimensional information, which gives a composite response to the components of the TCM product. Overall, the sensor array could produce a fingerprint for each batch of QZC. If the qualified products have good batch-to-batch reproducibility, their compositions and contents are varied in an acceptable range, which leads to the consistent complexation equilibrium between the sensor array and compounds, and the reflected fluorescence signals must be consistent. Conversely, unqualified products will have abnormal fluorescence patterns (Fig. 1b).

As a kind of macrocyclic host, the molecular recognition of calixarenes in aqueous medium has attracted widespread attention (Pan et al., 2021, Wu et al., 2022) and has been widely used in the field of sensing (Zheng et al., 2018, Zheng et al., 2019, Yu et al., 2019). Calixarenes are macrocycles consisting of phenolic units connected by methylene groups at the 2- and 6-positions (as shown in Fig. S1). By modifying the 4-position of the phenol unit (i.e., the upper rim of calixarenes), it is possible to provide additional recognition sites to the calixarenes. In our previous research, azobenzene modification allowed calixarenes to be extended longitudinally, and the deep cavities endowed azocalixarenes good recognition ability for many guests and even neutral molecules (Geng et al., 2019, Li et al., 2022, Yue et al., 2022, Zhang et al., 2020). Considering the complex chemical compositions and rich compound structures in QZC extracts, we selected azocalixarenes with different upper rim modifications and skeleton sizes for constructing sensor array. Four negatively-charged (CAC4A, SAC4A, SAC5A, and SAC8A) and two positively-charged azocalixarenes (QAAC4A and TAAC4A) were used in this work (Fig. 1c). These calixarenes with different recognition properties can provide complementary and cross-reactive interactions for multiple target analytes. Fig. 2a−f showed that the fluorescence of QZC extracts were partly quenched by CAC4A, SAC4A, SAC5A, SAC8A, QAAC4A and TAAC4A, but in different extent. Besides, more sensing units were introduced based on IDA strategy to provide more available information. A fluorescent molecule, AlPcS₄ can be complexed by QAAC4A and TAAC4A, and its fluorescence can be significantly quenched at 680 nm. After adding the QZC extracts, the replace of dye from calixarene cavities resulted in the recovery of fluorescence. Essentially, the QAAC4A/AlPcS₄ and TAAC4A/AlPcS₄ reporter pairs allowed for a fluorescence “switch-on” sensing of compounds from QZC (Fig. 2g and h). The ratio of fluorescence intensity, I/I₀ (the black line represents I₀ and the red line represents I in Fig. 2), was used as the response signal of supramolecular fluorescence sensing. Therefore, an eight-element sensing system consisting of CAC4A, SAC4A, SAC5A, SAC8A, TAAC4A, QAAC4A, QAAC4A/AlPcS₄ and TAAC4A/AlPcS₄ was carefully constructed.

Close

Fig. 2

(a−f) Fluorescence quenching of QZC extracts by the host−guest complexation of CAC4A, SAC4A, SAC5A, SAC8A, QAAC4A and TAAC4A (λ_ex = 325 nm). [QZC] = 20.0 μg/mL, [CAC4A] = 0.8 μM, [SAC4A] = 1.0 μM, [SAC5A] = [SAC8A] = 2.0 μM, [QAAC4A] = [TAAC4A] = 4.0 μM. (g−h) Fluorescence quenching of AlPcS₄ (λ_ex = 610 nm) by the host−guest complexation of QAAC4A and TAAC4A, and reactivation by QZC extracts displacement. [QAAC4A] = [AlPcS₄] = 1.0 μM and [QZC] = 80.0 μg/mL, [TAAC4A] = [AlPcS₄] = 1.0 μM and [QZC] = 40.0 μg/mL.

Export to PPT

3.2 Method validation of supramolecular fluorescence sensing for QZC

The precision and repeatability of the established supramolecular fluorescence sensor array were further evaluated. As shown in Fig. 3a and Table S3, the repeated tests of the sensing method on the same day showed good stability, and the RSD of each sensing element for six repetitions was less than 3.0%, which indicated the good intraday precision of the method. Next, we used these sensing elements to perform repeated measurements on consecutive three days to investigate the variation of the measurement results. Each dot on Fig. 3b represented the average of six repetitions in one day, and the three dots for each sensing element represented tested values on consecutive three days. Table S4 showed that the RSD values of eight sensing elements were less than 2.5% which proved good interday precision. Subsequently, the repeatability of the supramolecular fluorescence sensing method was tested by preparing six extracted solutions in parallel according to the extraction method. Table S5 showed the detailed data. The RSD of each sensing element did not exceed 3.0%, which proved the good repeatability (Fig. 3c). In the methodological study, good precision of the instrument and good repeatability of the experimental method indicated that the established supramolecular fluorescence sensing method was reasonable and feasible for quality analysis of QZC.

Close

Fig. 3

(a) The fluorescence response (I/I₀) for evaluating intraday precision (n = 6), (b) interday precision (n = 3, each dot represents the average of six repeats on the same day), and (c) repeatability (n = 6) of the established method (For direct complexation, [QZC] = 20.0 μg/mL, [CAC4A] = 0.8 μM, [SAC4A] = 1.0 μM, [SAC5A] = [SAC8A] = 2.0 μM, [QAAC4A] = [TAAC4A] = 4.0 μM. For competitive complexation, [QAAC4A] = [AlPcS₄] = 1.0 μM and [QZC] = 80.0 μg/mL, [TAAC4A] = [AlPcS₄] = 1.0 μM and [QZC] = 40.0 μg/mL).

Export to PPT

3.3 Visual modes of the quality analysis and semi-quantitative analysis results for QZC

Ten batches of qualified samples provided by the manufacturer were used to establish the fingerprint. For each batch of QZC, two testing solutions were prepared according to the extraction method, and each sensing element was used to measure each testing solution for three times. The average of six test results was plotted on the radar chart and each axis represented one sensing element (Fig. S7). The fluorescence response values of CAC4A, SAC4A, SAC5A, SAC8A, QAAC4A, TAAC4A, QAAC4A/AlPcS₄ and TAAC4A/AlPcS₄ sensing elements, namely I/I₀, were around 0.66, 0.57, 0.66, 0.64, 0.65, 0.68, 0.52 and 0.58, respectively. The fluorescence intensity ratio caused by QZC extracts of ten batches were closely arranged on each axis, showing consistency and good batch-to-batch reproducibility on the radar chart. As shown in Table S6, RSD among ten batches of samples is less than 3.5% for each sensing element. The evaluation results of the supramolecular fluorescence sensing method are consistent with the fact that these ten batches of QZC are all qualified products. This radar chart could be used as a standard fingerprint for qualified samples of QZC. The selected eight supramolecular sensing elements detect and judge the samples from eight dimensions, showing more comprehensive and accurate characteristics. In order to intuitively display the trend of content distribution, we adopted a heat map, which was shown in Fig. 4a. The heat map reflected the signal fluctuation in different batches through the gradient color. The results showed that these products had essentially good batch consistency, but the color fluctuations in the TAAC4A/AlPcS₄ and QAAC4A/AlPcS₄ groups meant the detected differences in the content of some components among these ten batches of products. This is basically consistent with the quality evaluation results of QZC using ultra-performance liquid chromatography technology reported in the literature (Xue et al., 2022). In several batches of qualified samples provided by the manufacturer, the RSD values of different detected components were ranged from 4.2% to 19.1%. Compared with HPLC method, which takes tens of minutes to analyse sample, supramolecular fluorescence sensing is almost real-time response, and high-throughput screening can be achieved with the help of 96-well plates.

Close

Fig. 4

(a) The heat map of I/I₀ results (n = 6) for the quality evaluation of ten batches of QZC through the constructed eight sensing elements (For direct complexation, [QZC] = 20.0 μg/mL, [CAC4A] = 0.8 μM, [SAC4A] = 1.0 μM, [SAC5A] = [SAC8A] = 2.0 μM, [QAAC4A] = [TAAC4A] = 4.0 μM. For competitive complexation, [QAAC4A] = [AlPcS₄] = 1.0 μM and [QZC] = 80.0 μg/mL, [TAAC4A] = [AlPcS₄] = 1.0 μM and [QZC] = 40.0 μg/mL). (b) The scope of the standard QZC products on radar chart and the analysis results for three batches of unqualified products.

Export to PPT

Based on our analysis of ten batches of qualified samples, each with six replicates of the test performed for each sensing element, we assumed that the data followed a normal distribution. To obtain a 95% confidence interval, we defined the range of qualified products as “μ ± 2σ,” where μ represents the average and σ represents the standard deviation. The threshold of standard products area has been shown in Table S6 and Fig. 4b. For a quick analysis on the production line, if the unknown sample point falls within the range of the standard chart, it is considered a qualified product with good batch-to-batch reproducibility. However, if the unknown sample falls outside the green area of the radar plot (Fig. 4b), it is recommended for further rigorous inspection. That is, the quality of the QZC extract with unknown quality can be preliminarily judged by rapid fluorescence measurement. In order to demonstrate the applicability of the standard fingerprint in differentiating qualified and unqualified products, we used three batches of QZC samples that had expired or even severely agglomerated to simulate unqualified products. The test results of these three batches (Table S7 and Fig. 4b) show that almost all of the I/I₀ values are outside the green region, indicating significant deviation from the qualified product. This may be due to the moisture absorption and deterioration of the long-stored products, which leads to inconsistencies in composition and concentration compared to the qualified products. Such products inevitably lead to changes in the supramolecular complexation equilibria and differential fluorescence signals.

Next, we verified the ability of the supramolecular fluorescence sensing method to perform semi-quantitative analysis. We used QAAC4A/AlPcS₄ and TAAC4A/AlPcS₄ to construct a binary sensor array to distinguish four different concentrations of QZC extracts. Through pattern recognition method such as LDA, a versatile system to identify differences among the tested samples could be obtained. LDA is a statistical method, which minimizes intraclass variance and maximizes interclass variance to differentiate the response patterns from the samples. The experiment was carried out with a black 96-well plate. QZC sample solution obtained according to the extraction method was diluted to give final concentrations from 22.5 to 90.0 μg/mL, and two sensing elements were used to test them for six times (Table S8). As shown in Fig. 5a and b, the intensity fold change was calculated using I/I₀ and then further analyzed by LDA. The two most significant LDA factors have been used to generate the two-dimensional differentiation plot with 95% confidence ellipses, which clearly showed that all the four concentrations of QZC were distinct from each other. The jackknifed classification matrix with cross-validation revealed 100% accuracy which proved the reliability of this chemical sensor array. On the LDA plot, the first dominant factor accounted for 99.29% of total variance, and the distribution of factor 1 varied with the concentration of QZC. By converting the result into a scatter plot of factor 1 and concentration, a quantitative identification of samples could be achieved. The linearity indicated that the interactions between the sensing elements and analytes were homogeneous and stable (Fig. 5c). Therefore, valuable arrays can be constructed by combining appropriate supramolecular sensing elements, which can not only quantitatively analyze QZC, but also meet the needs of distinguishing different samples, such as unqualified products with abnormal concentrations of ingredients. Compared to other methods and techniques, our proposed method is well-suited for quality evaluation of commercial TCM products due to its advantages of simplicity and less time consumption (Table 1).

Close

Fig. 5

(a) Fluorescence response patterns of QAAC4A/AlPcS₄ and TAAC4A/AlPcS₄ sensing reporter pairs against various concentrations of QZC extracts ([QAAC4A] = [TAAC4A] = [AlPcS₄] = 1.0 μM). (b) Canonical score plot for the two factors of simplified fluorescence response patterns obtained from LDA with 95% confidence ellipses (n = 6). (c) Plot of the first discriminant factor vs. the concentrations of QZC extracts.

Export to PPT

3.4 Quality analysis for Shuanghuanglian (SHL) injection

To demonstrate the generalizability of the supramolecular fluorescence sensing method for quality evaluation of TCM, we further created a standard fingerprint for SHL injection. SHL injection, composed of the extract of Lonicerae Japonicae Flos, Forsythiae Fructus, and Scutellariae Radix, has the function of heat-clearing and detoxifying, as well as dispelling wind and relieving exterior symptoms (Wang et al., 2022). Here, other supramolecular macrocycles, cyclodextrins and cucurbit[7]uril were selected as recognition receptors, and four sensing elements, β-CD/2,6-TNS, HP-β-CD/2,6-TNS, CB7/1,8-ANS and CB7/DAPI were constructed (Fig. S8). The complexation of the used dyes with these three macrocycles sensitizes their fluorescence (Liu and You, 2001, Wagner et al., 2003, Miskolczy et al., 2009). Based on the IDA strategy, the addition of SHL injection breaks the original complexation equilibria and the displaced dyes regain their inherent weak fluorescence (Fig. S9). Table S9 was obtained by testing 4 sensing units × 10 batches of samples × 6 replicates and calculating the ratio of fluorescence intensity before and after the addition of SHL injection. The RSD values among the ten batches of samples were less than 4.0%. The fluorescence fingerprint was constructed by outputting the average fluorescence response data (I/I₀) on a radar plot (Fig. S10). These samples were all qualified products from the manufacturer, and their fluorescence responses were closely aligned on each axis. They did not show large deviations, except for fluctuations in a certain small range. The results demonstrate that the application of supramolecular fluorescence sensing method can be extended to other TCM.