|Abstract: ||Abstract:This study developed and validated a method of measuring the feminizing chemicals 4-tert-octylphenol, 4-nonylphenol, nonylphenol monoethoxycarboxylate (NP1EC), nonylphenol monoethoxylate (NP1EO), nonylphenol diethoxylate (NP2EO), estrone, 17β-estradiol, estriol, 17α-ethinyl estradiol and bisphenol A in river water, sediment, and tissue using ultra-high performance liquid chromatography/tandem mass spectrometry (UHPLC/MS/MS) and isotope-dilution techniques. Water samples were pretreated using disk-type automated solid-phase extraction (SPE). Solid samples of sediment, fish, and clams were treated with matrix solid-phase dispersion (MSPD) using C8 adsorbent. Eluents were directly passed following alumina cartridges for cleanup. The signal intensity of analytes on electrospray ionization (ESI) was compared with that of atmospheric pressure photoionization (APPI). The analytes were separated on a UHPLC C18 column with aqueous 10-mM ammonium acetate for NPEOs and aqueous 10-mM N-methylmorpholine for the other compounds. On-line cleanup was evaluated using two-dimensional liquid chromatography (2-D LC).
ESI could provide satisfactory response for all of the analytes. Though APPI did not offer suitable response for NP1EO, NP2EO and NP1EC, it provided better signal intensities for the steroid estrogens (1.0-2.4 times) and the phenols (3.2–4.4 times) than ESI. UHPLC shortened chromatographic time to less than 10 min. Disk-type automated SPE and MSPD dramatically increased the throughput of sample preparation. The extraction efficiency on surface water samples ranged from 10% to 91%. The extraction efficiency of MSPD on sediment, fish, and clams was 51-101%, 36-109%, and 30-111%, respectively. Acidic alumina cleanup was essential for the analysis of the tissue sample, and reduced matrix effects better than 2-D LC on-line cleanup. The limits of detection (LODs) in water ranged from 0.81 ng/L to 89.9 ng/L. The LODs in sediment and tissue ranged from tens of pg/g wet weight to only a few ng/g wet weight. This method proved to be accurate and reproducible, as both quantitative biases and relative deviations remained smaller than 20% at three spiked levels.