Due to the extremely strong toxicity of aflatoxins, necessary cautions and essential protections are needed throughout the whole experiments according to strict guidelines. Anti-poison respirators, isolation gowns and goggles were used. After being contaminated with aflatoxins, the containers and consumables were dipped in 10% sodium hypochlorite in a specified holder for at least 24 h. The lab garbage was disposed in an appropriate way.

The IAC-ToxinFast^®-Aflatoxins Test IACs were supplied by Huaan Magnech Bio-Tech Co., Ltd. (Beijing, China). The TGL 16M centrifuge was bought from Hunan Kaida Scientific Instrument Co., Ltd. (Changsha, China) and PB-10 pH meter from Sartorius Intec (Hamburg, Germany). HPLC grade methanol and acetonitrile were purchased from Fisher Scientific (Fair Lawn, NJ, United States). Other reagents and chemicals of analytical grade were obtained from Beijing Chemical Works (Beijing, China). Deionized water was purified by the Milli-Q Plus ultra-pure water system (Millipore, Billerica, MA, United States).

Phosphate buffer saline (PBS) was prepared by dissolving 0.2 g KCl, 0.2 g KH₂PO₄, 8.0 g NaCl, and 1.2 g Na₂HPO₄ in 1000 mL of pure water at pH 7.4. 0.1 M NaH₂PO₄ solution was set as Solution A, which was obtained by dissolving 1.56 g NaH₂PO₄⋅2H₂O in 100 mL of pure water, while Solution B was obtained by dispersing 35.8 g Na₂HPO₄⋅12H₂O in 1000 mL of pure water. 0.1 M Phosphate buffer (PB), encompassing 60 mL of Solution A and 940 mL of Solution B, was homogeneously mixed and afterward the pH value was adjusted to 8.0 with 0.1 M HCl. 0.5%, 1%, and 2% PBST (PBS with addition of Tween-20) solutions were got by adding 5, 10 and 20 mL of Tween-20 into 1000 mL PBS (pH 7.4), respectively.

All analyses were performed on a Shimadzu LC-20AT HPLC system (Shimadzu, Kyoto, Japan) consisting of two LC-20 AT pumps, an SIL-20A autosampler, a CTO-20A column oven, a CMB-20A controller, the post-column PCD reactor and an RF-20AXL fluorescence detector. The PCD reactor with a mercury lamp (λ = 254 nm) and a knitted reactor coil of 0.74 mL (15 m × 0.25 mm) was bought from AURA Industries (New York, NY, United States). The eluate was monitored by using a fluorescence detector with an excitation wavelength of 360 nm and an emission wavelength of 450 nm. The chromatographic separation of tested aflatoxins was conducted on a CAPCELL PAK C18 column (4.6 mm ID × 150 mm, 5 μm) at constant 30°C. HPLC separation of four AFs was carried out by using methanol and acetonitrile (40: 18, v/v) with the water phase (pure water) as the mobile phase at isocratic elution. Injection volume was 20 μL and the flow rate was set at 1.2 μL/min. The total run time was 30 min.

Aflatoxins standard including 2 μg AFB₁, 2 μg AFG₁, 0.5 μg AFB₂, 0.5 μg AFG₂ in 1 mL of acetonitrile were bought from Pribolab (Singapore). Working standard solution was prepared by diluting the standards with methanol to get 200 ng/mL of AFB₁, AFG₁ and 25 ng/mL of AFB₂, AFG₂. All the standard solutions were stored at -20°C in the dark before analysis.

Twenty-six batches of H. sabdariffa samples including 6 batches of ground powder (S1–S6), 11 batches of crude materials (S7–S17) and 9 kinds of scented teas (S18–S26) were randomly collected from Guangxi, Sichuan, Anhui, Guangdong and Yunnan provinces, Beijing city of China and Thailand. 500 g sample was used for the preparation of sample solution. A positive control (S28) was obtained by inoculating the sample from Beijing city with A. flavus spores while another fraction of this sample (S27) was kept at the same incubation conditions without A. flavus. All crude samples and scented teas were identified by Prof. Meihua Yang (Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences, Peking Union Medical College). All the samples were thoroughly triturated to obtain consistent particle size and homogenized samples. The sub-sample and scented teas were powdered and sieved through a 24-mesh sieve. All samples were classified and labeled clearly in lock bags while stored in a cool dry place before use.

The resulting 40 mL of filtrate was passed through an immunoaffinity column (IAC-ToxinFast^®-Aflatoxins Test) at a flow rate of approximately 3 mL/min at room temperature for at least 30 min. When the filtrate nearly passed through the column, 20 mL of ultra-pure water was added to wash the column at a flow rate of 3 mL/min. Finally, all the residual liquid was excluded with clean air. Next, 1.5 mL of methanol was added onto the IAC for elution of the aflatoxins. The final volume of the eluate was brought to 2 mL with methanol. After being mixed uniformly, the final solution was filtered through a 0.22-μm glass fiber filter paper.

Sensitive quantitation of four AFs in H. sabdariffa samples in the form of ground powder, crude herbs and scented teas was carried out by measuring the peak area responses in the HPLC-FLD chromatograms at eachg aflatoxin retention time and comparing them with the linear equation of a calibration graph. The peak area (y-axis) of each aflatoxin was plotted against the concentration (ng/mL, x-axis) and expressed as a linear equation y = ax+ b. By determining the values of slope (a) and y-intercept (b), the equations were established and the content of each aflatoxin in the tested sample was calculated.

For accurate determination of four target aflatoxins in H. sabdariffa samples with high acidity, the established HPLC-PCD-FLD method was validated according to EU Regulation No. 401/2006 (The Commission of the European Communities, 2006) in terms of linearity, limit of detection (LOD) and limit of quantitation (LOQ), accuracy, precision, stability, selectivity and robustness.

Each experiment was performed in triplicate and all the data were represented as mean ± standard error (SE) and analyzed by ANOVA using SPSS software (version 19.0). Means were separated by Tukey’s multi-range tests when ANOVA results were significant (P < 0.05).

Results of preliminary experiments using the recommended method (Chinese Pharmacopoeia Commission, 2015) for detection of AFs in H. sabdariffa showed that the four AFs (AFB₁, AFB₂, AFG₁, AFG₂) were not adequately separated and recovered. Unfavorable detection was due to complex matrix components and the high acidity of this TCM. To overcome these issues, we modified and improved sample pretreatment and cleanup for sensitive HPLC-FLD analysis. First of all, necessary work was reasonably performed on different dilution ratios of pure water. The sample extract was diluted with pure water at dilution ratios of 1:4 and 1:8 (v/v). These dilutions were inadequate to recoveries of AFs at the required levels of 70–110%. Therefore, other highly effective dilution reagents were taken into consideration.

The optimal pH value for IACs is in a range of 6.0–8.0. The mean pH value of the above-mentioned diluent in water at 1:8 ratio was around 2.4, which will weaken the stability of antibodies and their target affinity. Few immune reactants would be out of function under this condition. Diluting the sample just with water to provide buffer capacity and minimize the matrix was not sufficient to overcome the strongly acidic matrix of H. sabdariffa (Da-Costa-Rocha et al., 2014).

To resolve the problem, PBS (Ip and Che, 2006) was introduced to adjust the pH of the test solution before application onto the IACs. Unfortunately, it was found that the recoveries of AFG₁ and AFG₂ still did not meet the official requirements. To address this issue, Tween-20 (Zhang et al., 2017) was added into the buffer, and three levels (0.5, 1 and 2%, v/v) of Tween-20 were compared to screen the most effective buffer dilution solution. Increasing the volume of Tween-20, the recoveries of AFs went up mildly (Figure 1). However, even with 2% Tween-20 in PBS (pH 7.4) the recovery of AFG₂ did not meet the requirement.

For extracts of H. sabdariffa with such high acidity (around pH3.0), 0.1 M PB (pH 8.0), which has a greater buffering capacity, could be used to successfully adjust the pH to the appropriate range (pH > 6.0) for IAC. When 0.1 M PB (pH 8.0) was used for dilution, the recoveries of four aflatoxins in H. sabdariffa increased sharply to acceptable levels (Figure 1).

By diluting samples with 0.1 M PB (pH 8.0), matrix interferences were minimized and AFs recoveries were within the standard range. Therefore, to ensure that the concentrations of AFs in the final eluates could successfully reach the LODs and to decrease useless hands-on time, a dilution ratio of 1:8 (v/v) was finally chosen.

Using a suitable sample extraction technique plays an important role in trace detection for effectively extracting aflatoxins in H. sabdariffa without increasing impurities. A homogeneous extraction technique has been recommended (Chinese Pharmacopoeia Commission, 2015). Here, two simple and commonly used techniques including homogeneous extraction for 3 min at 11,000 rpm with a homogenizer and ultrasonication-assisted extraction in an ultrasonic bath at 500 W for 20 min were both tested to compare their efficacy in extracting AFs from spiked H. sabdariffa samples with 5.0 μg/kg of AFB₁ and AFG₁, and 1.25 μg/kg of AFB₂ and AFG₂. Recoveries of the four AFs in spiked H. sabdariffa samples were all over 80%, but the values obtained from ultrasonication-assisted extraction were all higher than that from homogeneous extraction (P < 0.05) (Figure 2). We found that after homogeneous treatment, a large quantity of the sample solution was adhered to the surface of the stirring arm. In addition, with this method, samples were processed sequentially and the machinery had to be washed carefully between testing, resulting in wasted analytes and time. Therefore, ultrasonication-assisted extraction using 70% methanol was chosen for sample preparation. With the application of modifications, the recoveries of four aflatoxins met the trace analysis requirements.

The method was validated for linearity, LOD and LOQ, accuracy, precision, stability, selectivity and robustness described as follows.

To confirm the suitability and versatility of our optimized pretreatment conditions including ultrasonication-assisted extraction and IAC cleanup, the validated HPLC-FLD analytical procedure after online PCD was applied for simultaneous determination of four AFs in 28 H. sabdariffa samples including 6 batches of ground powders, 11 batches of crude materials and 9 kinds of scented teas purchased from different markets, herbal shops and bazaars in China and Thailand, and two man-made moldy samples. The HPLC-FLD chromatograms showed that the four AFs were well separated with satisfactory responses under the present chromatographic conditions (Figure 3). No interference peaks were detected at the retention time of each aflatoxin in the normal (AFs-free) and moldy samples. The occurrence and contents of four AFs in different H. sabdariffa samples for this survey were summarized in Table 3. One artificial moldy control sample after being inoculated with A. flavus spores for 20 days was tested to be positive for AFB₁ at 3.11 μg/kg, which did not exceed the MLs set by China and other international organizations. Because of its important medicinal values and world-wide consumption, it is important to monitor the occurrence and levels of aflatoxins especially AFB₁ in H. sabdariffa products to ensure quality and safety.