Detection Of Aflatoxin B₁ Contamination In Pistachio Using Laser Induced Fluorescence Spectroscopy

Pistachio is one of the four largest tree nuts in the world.It is second only to chestnut and walnut in China and ranks third.Pistachio contains rich nutrients,and thus is easily infected by various fungi under poor weather,transportation,processing,and storage conditions before,during,and after harvest.The metabolic activities of molds can produce a variety of secondary products,of which mycotoxins seriously threaten human and animal health.Aflatoxin is the most toxic and very stable mycotoxin,including more than ten kinds of derivatives.Among them,aflatoxin B₁（AFB₁）is the most widely distributed and has the highest carcinogenicity and teratogenicity.The detection of pistachio contaminated with aflatoxin has attracted much attention from researchers for many years.However,there are still many problems and challenges that need to be resolved in actual industry analysis.Therefore,this work conducted experiments on problems such as weak target fluorescence intensity,difficult acquisition of characteristic signals,and high detection limit in early research.This study explored the fluorescence characteristics of AFB₁ and mold secondary metabolites like Kojic Acid（KA）and Kojic Acid Derivative（KAD）in solution,as well as signal sensitization methods;analyzed the similarities and differences of the fluorescence spectra of control and low concentration of aflatoxin contaminated samples under laser irradiation,and optimized the pattern recognition algorithm;designed one,two and three-probe detection system and compared the prediction results for AFB₁;set up an dynamic LIFS detection platform,and evaluated the applicability of the system to multi-variety samples.The purpose of this study is to provide a rapid nondestructive detection method for aflatoxin contaminated pistachios based on LIFS technology,and to provide a theoretical basis for the application of fluorescence spectroscopy in analysis of quality and safety of agricultural products.The main research contents and results of this work are as follows:（1）The absorption and fluorescence spectrum characteristics of AFB₁,KA and KAD in solution were explored,and the optimal excitation wavelength was determined.Results indicated that KA has no obvious absorption peaks in UV-Visible region.The positions of main characteristic absorption peaks of AFB₁ showed blue shift with the decrease of solvent polarity,and were at 364,360,and 356 nm in water,methanol and acetonitrile respectively.The position of KAD absorption peak was at378 nm.The presence of high concentration of KAD would affect the identification of characteristic absorption peaks of AFB₁,causing red shift for the position of main aflatoxin peak.Three-dimensional fluorescence spectroscopy showed that there was different for fluorescence peaks of AFB₁ and KAD.The positions of AFB₁fluorescence peaks under the excitation of three wavelengths including 234,265,and365 nm were all at 430 nm,and the optimal excitation wavelength was 365 nm.KAD only had one fluorescence peak at 493 nm under 375 nm excitation,and did not emit fluorescence under the excitation of 234 and 265 nm.Compared with KAD,the fluorescence peak of AFB₁ was more susceptible to influence of solvent type,and its signal intensity in 80%methanol was much higher than that of KAD.Therefore,it can reduce or eliminate interference of KAD on AFB₁ fluorescence signal in high concentration methanol under the excitation of deep UV light source.（2）A enhancement of AFB₁ fluorescence signal method based on molecular inclusion technology was proposed,and the mechanism was explained.Three kinds of cyclodextrin（CD）were selected for further analysis in this study.Results showed thatα-,β-andγ-CD can significantly increased AFB₁ fluorescence intensity,but had no significant effect on KAD signal intensity.Among them,β-CD had the best improvement effect,and the multiplication value for AFB₁ was 7.96.KAD could inhibit fluorescence enhancement effect of CD on AFB₁ to some extent.However,the multiplication value for AFB₁ was still high（4.68）,which was 3.76 times of KAD.The concentration of methanol had no significant effect on fluorescence intensity of AFB₁-β-CD,but high concentration of acetonitrile would reduce its signal intensity.The hydrophobic cavities ofα-,β-andγ-CD could form 1:1 inclusion complex with AFB₁.AFB₁-β-CD had the highest stability,and the value of log K was2.49.Molecular simulation results showed thatβ-CD had the highest affinity with AFB₁,and the binding group was the furan ring of AFB₁.（3）A LIFS system was constructed,and the fluorescence spectra of pistachios contaminated with AFB₁ were studied,and the optimal classification models were obtained.Using a 360 nm laser as excitation light source,the static LIFS system was set up to detect two varieties of pistachios（“Zaoshu”and“Kerman”）contaminated with AFB₁ at 5,10,20 and 50 ppb levels.Results indicated that the average fluorescence signal intensity of the samples from two species in the uncontaminated group（no acetonitrile and AFB₁ was added onto the samples’surface）was slightly higher than that in contaminated group between 400 and 800 nm.Establishing Linear Discriminant Analysis（LDA）models based on different bands,it was found that the contribution rate of 400～610 nm to correctly identify contaminated samples was high.Compared with Logistic Regression Classification（LRC）,the Support Vector Machine（SVM）models established with Standard Normal Variate（SNV）had higher classification accuracy rates.Among them,the accuracy rates of“Zaoshu”and“Kerman”models were≥92.3%and≥93.8%respectively,and the mixed variety models were≥98.4%,especially there was no false negative.（4）A multiplexing optical fiber LIFS detection method was proposed,and the optimal detection mode was determined.In order to collect the fluorescence signals of stimulated“Kerman”pistachios from different emission directions,a set of multiplexing optical fiber detection device was designed.Results showed that the order of fluorescence intensity obtained in different detection modes was:three-probe>two-probe>one-probe,and the difference between average spectra of control group（only added with acetonitrile）and contaminated group samples was the largest using three-probe mode.Overall,the SVM model based on second-order differential spectral data in 390～660 nm had the best performance.Three-probe mode had the highest accuracy rate（97.1%）for low concentration of AFB₁（5 ppb）contaminated samples,followed by two-probe mode（91.2%）,and one-probe mode was the lowest（88.2%）.The correlation between fluorescence spectra data obtained by one-,two-,and three-probe modes and AFB₁ concentrations were relatively high.Root Mean Square Error of Prediction（RMSEP）of Stepwise Multiple Linear Regression（SMLR）models for one,two,and three-probe modes were 6.7,5.7,and 4.4 ppb respectively,and the Limit of Quantification（LOQ）for three-probe mode was the lowest.（5）A set of dynamic LIFS detection system was designed,and the applicability for different varieties of AFB₁ contaminated pistachios was evaluated.Pistachio could move linearly controlled by a linear motion,and then passed through a three-probe detection unit in turn.This dynamic system automatically completed spectral collection by means of a photoelectric switch.The spectra of“Zaoshu”and“Kerman”samples contaminated with AFB₁ were collected in this work.Principal Component Analysis（PCA）results showed that there was a certain separation trend between control,uncontaminated and contaminated samples for each variety and mixed variety.The LDA model could well distinguish contaminated groups for single variety,and the discrimination accuracy rates were≥91.2%,but the performance of mixed variety models were slightly low（≥85.1%）.Screening the wavelengths within 174～1100 nm based on Competitive Adaptive Reweighted Sampling（CARS）,the prediction precision of Partial Least Squares Regression（PLSR）model for AFB₁ concentration in samples from single and mixed variety were the highest,and the values of RMSEP were all<10.0 ppb,especially the number of selected variables were<70.Results of principal component spectra and correlation analysis showed that three wavelengths of 428,520,and 741 nm had the highest correlation with AFB₁ contamination.