應用螢光光譜指紋技術檢測食用生菜污染之研究

Abstract

食品與飲用水中之微生物可能造成的衛生安全問題，過量的微生物或菌株造成食物中毒的現象時有所聞。衛生單位傳統之檢驗方法，需採樣食品後進行菌株的培養，檢驗其中所含的菌株與數量，然而菌株的分離與培養耗時且過程繁複，培養至肉眼可分辨之菌落至少需要一天到兩天的時間，無法立即檢驗食物是否符合衛生標準，給予消費者保障，若檢驗結果為菌量超標或有病原菌，對於已經把食物吃下肚的消費者也是於事無補。有鑑於此，生物技術演進發展出其他檢測微生物的方法，依據不同種類菌株中特有的酵素或基因，設計適當的實驗使檢驗之微生物產生特殊反應，利用間接的方式檢測其反應，可立即獲得菌種與菌量等資訊，這些方法不但快速、具備專一性，且非常靈敏，感度甚至可以定量至單一菌株，然而其所需技術門檻、成本較高，缺點是間接處理過程仍會對樣本進行破壞，食品檢測完畢後無法供給食用，無法進行全檢。光譜檢測技術兼具快速、非破壞性之優點，雖然相較於上述兩種方法定量準確性稍差，足以用於檢驗食品中微生物，目前有許多相關研究利用光譜檢測水果或肉品之污染情形，然而用於蔬菜的研究與檢驗非常少。 因此檢驗生食用蔬菜之微生物為本研究之重點，其中指標性微生物是衛生單位判定食品衛生標準之依據。本研究以螢光光譜檢測紅蘿蔓萵苣，並以標準方法檢驗萵苣中生菌數與大腸桿菌數作為依據，建立螢光光譜預測蔬菜受菌污染之模型。以PCA投影的二維或三維空間中可區別未受污染與人工污染之萵苣樣本；PLS-DA預測污染與否之成功率相當不錯，選取兩個隱藏函數時校正組與驗證組正確率為84.0 %與82.9 %，選取六個隱藏函數時校正組與驗證組正確率為96.0 %與80.0 %。預測大腸桿菌數最好之模型為PLSR，校正組rc = 0.780、SEC = 1.603 (log CFU)，驗證組 rv = 0.701、SECV = 1.806 (log CFU)；預測生菌最好之模型為MLR挑選兩個特徵波長條件，校正組rc = 0.701、SEC = 1.384 (log CFU)，驗證組 rv = 0.670、SECV = 1.403 (log CFU)。

Fresh-to-eat vegetables become popular in our daily diet, however numerous outbreaks of microbial infections are related to food or water contamination. Conventional methods of microbial monitoring need to sample the food by swabbing or mincing into small particles. A series of procedures for sample preparation and microbe measurement is time-consuming and labor-extensive. Although rapid analysis methods using enzyme reactions and DNA technologies have been developed, these methods are sample-destructive and indirect measurements, and they also take time, money, professional skills and instruments to accomplish. Spectral methods with advantages of high-speed and non-destructive can be applied in microbe measurements. There are many studies related to fruits and meats contamination detection by spectral methods but few studies related to vegetables. As a result, a rapid and non-destructive method for vegetables is indispensable. In this study, nondestructive fluorescence spectroscopy was applied to the inspection of E. coli contamination on fresh-to-eat vegetables. A fluorescence spectrophotometer was used to measure the fluorescence intensity from the surface of red Romaine lettuce (Lactuca sativa L. var. longifolia) samples. The measurements provided a three-dimensional data cube, and could be used to study the biomaterial’s surface properties by both analyses of discrimination and quantification. An excitation Emission Matrix (EEM) was established, fluorescence fingerprints were analyzed by Mahalanobis distance, and fluorescence peak points were identified.The contaminated lettuce samples could be distinguished by PCA analysis. The successful rates of PLS-DA for better prediction models were 84% and 82.9% when selecting two LVs and 96.0% and 80.0% when selecting six LVs. The best model to predict E. coli contamination condition on lettuces was PLSR model. The correlation coefficients for calibration set rc = 0.780, and SEC = 1.603 (log CFU). For validation set, rv = 0.701 and SECV = 1.806 (log CFU). The best model to predict Aerobic Count Plates was MLR model. For calibration set rc = 0.701 and SEC = 1.384 (log CFU). For validation set, rv = 0.670 and SECV = 1.403 (log CFU). The work showed that fluorescence can be applied to rapid inspection of E. coli contamination on fresh vegetables to assure the food safety.