以自動化固相萃取搭配極致液相層析/串聯式質譜儀分析食品與飲用水中全氟碳化物

Abstract

全氟碳化物 (Perfluorinated chemicals, PFCs) 由於具有防水抗油的特性，在過去五十年中已被廣泛使用於多種生活用品與工業製程中，其於環境中分布廣泛且具有持久性與生物累積性，長期暴露對人體健康具有潛在的危害。目前已有研究證實自來水處理廠的淨水過程中無法有效移除全氟碳化物。而人類主要會經由食品、飲用水與家中灰塵暴露到全氟碳化物。 本研究的目的是開發以自動化固相萃取搭配極致液相層析/串聯式質譜儀(ultra-high performance liquid chromatography/tandem mass spectrometry, UHPLC- MS/MS)分析飲用水與食品中十種全氟碳化物的分析方法。每種化合物最佳之偵測母離子(precursor ion, [M-H]-)及兩個分別做為定量和定性子離子(product ion)之質譜/質譜參數、電灑游離（electrospray ionization, ESI）離子源參數以及液相層析條件皆已最適化。液相層析使用Kinetex C18管柱(2.1 x 50 mm, 2.6 μm)，在每分鐘0.9 mL的流速下，整體層析時間只需5.6分鐘。層析移動相選用甲醇與10 mM 甲基嗎啡琳的組合，層析溫度40℃，使用梯度流析。儀器偵測極限(instrumental detection limit, IDL)為0.03 – 0.16 pg，儀器定量極限(instrumental quantitation limit, IQL)為0.08 – 0.60 pg。由於許多實驗室物品含有全氟碳化合物成分，容易形成背景干擾。本研究除了盡量避免使用含有全氟碳化合物的實驗器具之外，對於一些所使用的器皿和裝置也分別進行了背景干擾的測試，結果發現主要的背景訊號來源為UHPLC-MS/MS系統管線，絕對質量在十種待測物中以全氟己酸與全氟辛酸背景值較高，分別為0.96 pg及1.68 pg。 在前處理部分，鮮乳樣本以0.5 N氫氧化鉀水溶液消化並稀釋，在調整至pH值3.5以及過濾後，以Atlantic HLB disk搭配自動化固相萃取器進行萃取。飲用水樣本調整pH值至3.5後，直接進行固相萃取。固態食物樣本均質化後以0.5 N氫氧化鉀甲醇溶液進行消化化兩小時，再以3,000 rpm (1,410 xg) 離心30分鐘，之後將上清液以去離子水稀釋100倍，調整pH值至3.5以及過濾後，使用Atlantic HLB disk進行固相萃取。固相萃取步驟為，先以甲醇與去離子水活化吸附劑，在通過水樣之後，以40%甲醇水溶液進行淨化流析，以洗去雜質，沖提溶劑選用100%甲醇添加0.1% 氨水。由於全氟辛烷磺胺與N-甲基辛烷磺胺會於濃縮過程中揮發而損失，不能濃縮至近乾，因此萃取液最後濃縮至1 mL。 整體來說，多數待測物的離子抑制效應小於50% (-29-49%)，且在飲用水與魚肉樣本之離子抑制效應較小。在各種基質中，多數待測物的樣品前處理回收率皆大於50% (52-121%)。多數待測物在飲用水中的方法偵測極限皆小於1 ng/L(0.29-0.85 ng/L)，在鮮乳樣本中則為1.8-11 ng/L，固態食品樣本則低於1 ng/g(0.15-0.50 ng/g)。由於全氟辛烷磺胺與N-甲基辛烷磺胺兩者的離子抑制效應較其他待測物嚴重，且在前處理過程中容易揮發流失，偵測極限在各種基質中皆大於其他待測物。本研究使用同位素稀釋技術定量，並比較兩種定量方法，分別使用六個以及兩個穩定同位素標定內標準品來定量十個待測物。六個同位素標定內標準品是選用與待測物相同結構或只差一個碳數且結構相似者來對其定量，結果多數待測物的精密度與準確性遠優於僅使用兩個內標準品之情形，因此每個待測物使用與自己較相近碳數的同位素內標定量，會較同一類待測物只選用一個同位素內標定量來得準確。 本研究所開發的分析方法可以簡化流程，具有良好的偵測靈敏度並能節省時間與勞力；而針對每個待測物選用適當的同位素標定內標準品定量，可得到很好的準確性與再現性，並可適用於多種環境與食品樣本中全氟碳化物的分析。

Perfluorinated chemicals (PFCs) are persistent, ubiquitous, and bioaccumulative in the environment, and are potentially harmful to human health. Because of their lipid and water repellent characteristics, they have been widely used in various products for more than fifty years. The processes of drinking water treatment are ineffective in removing these chemicals. Humans are primarily exposed to PFCs via drinking water, food and household dust. This study developed a method to determine 10 PFCs in drinking water, milk, fish, beef and liver by automated solid-phase extraction (SPE) and ultra-high performance liquid chromatography/tandem mass spectrometry (UHPLC-MS/MS). The 10 PFCs were separated on a Kinetex C18 column (2.1 x 50 mm, 2.6 μm) at 40℃ and the flow rate was 0.9 mL/min; the total chromatographic time was 5.6 minutes. The mobile phase was composed of methanol and 10 mM N-methylmorpholine. Trace amount of perfluorohexanoic acid (PFHxA) (0.96 pg) and perfluorooctanoic acid (PFOA) (1.68 pg) was observed in backgrounds and the major contamination source was identified as the lines in the instrument of UHPLC-MS/MS. Milk was digested with 0.5-N potassium hydroxide in Milli-Q wate, after adjusted to pH 3.5 and filtration, the sample was extracted with an Atlantic HLB disk by automated SPE. Drinking water (adjusted to pH 3.5) was directly extracted with the Atlantic HLB disk. Solid food samples were homogenized and digested by 0.5-N potassium hydroxide in methanol for two hours. After centrifugation at 3,000 rpm (1,410 xg) for 30 minutes, the supernatant of the samples were diluted with 100-fold Milli-Q water then was extracted with the Atlantic HLB disk. After loading the samples, the disks were washed with 40% methanol/60% water, and then were eluted with 0.1% ammonium hydroxide in methanol. In the concentration step, perfluorooctanesulfonamide (PFOSA) and N-methylperfluorooctanesulfonamide (N-MeFOSA) were found to be evaporated when the eluent was concentrated to barely dry; therefore the extracts were only concentrated down to one milliliter. Ion suppression of most analytes was below fifty percentages (-29-49%), and was generally lower in fish and drinking water but was higher in liver. Recoveries of sample preparation of most analytes were higher than 50% (52-121%) in five matrixes, but only small portions of PFOSA and N-MeFOSA remained after sample preparation (1.8-34%). The limits of detection (LODs) for most analytes were sub-ng/L (0.29-0.85 ng/L) in drinking water, and were from 1.8 to 11 ng/L in milk. LODs of most analytes were 0.15-0.50 ng/g net weight of solid food samples. LODs of PFOSA and N-MeFOSA were higher than other analytes because they suffered higher ion suppression and loss at the concentration step. This study compared the accuracy and precision in five matrixes between using two and six isotope-labeled internal standards to quantify the ten analytes. Quantitative accuracy and precision on almost all analytes were better by using all the six internal standards than those using only two of them. The method was simple, bettered the detection sensitivity, and saved time and labor. Use of suitable isotope-labeled internal standards for each analyte was crucial for the quantitative precision and accuracy. This method can be applied to measure these chemicals in a variety of food samples.