. ‘ l. I. ‘ Ty“. Gr '1‘ c. 'vY . .!, AW? { 1 .Nuquls .. 3mg. HAWAQJ. v i , u.‘ x 2. :1 It! u :11”. I: ‘ A? :5... tabs". fl 1 .1 1- .I. 3w 3..me -. . 4‘ ." I.’.S ha: | I ' Sn. SW 633.). . . . 1:1... . . s “a .y , f. V. a} a}. .‘ 1 ll . . If‘uiiknuvvwcvM-‘wz. .. . [I v.3.Mnn: .dnflr. ‘ A: O . vJIa’,'. A VXI’,‘ I}. v. . ul.§.J..I ,. I.‘ u . .avllflufin .. . ‘ - I. tl t... . ! Ewan... x. . I. II 15.! gfiwmh . ‘ .4 x V 1‘ bull-i . . . i. I! (7‘ uril ‘;\—J.. V .L l ‘ ‘~ ' l . a fill Ir?! {1 In ! ti... .\¢.. 3.53%! i . ...r...l.1 .. r ,8. Van . 4 h ,o . Ina Kn. ts) . iii}! . I I .‘l.’ .A‘bli‘ Ill. ‘7" ‘l“ ‘ V. flu“ m? $flumm..wxmmufiwu a . .HFJmWUu . , . '~ .I, . v 4.3 n . ., u . 4.71.1.5" e.. z» , .11? 1.314.! c Elm-5&3, 6“? . .w 1 , I. : ‘a 3.2 .U lllllllllllllllllll”Hill“'IHIIIHHIHIlllllHlHlNlllll 193 02058 6560 This is to certify that the thesis entitled IDENTIFICATION AND QUANTITATION 0F NONYLPHENOL ETHOXYLATES AND NONYLPHENOL IN FISH TISSUES OF MICHIGAN, USA presented by Timothy L. Keith has been accepted towards fulfillment of the requirements for M.S. Zoology degree in - 9&1”me Major Mfessor Date {AZ/00 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution i LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. To AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 11100 acumen-peep.“ IDENTIFICATION AND QUANTITATION OF NONYLPHENOL ETHOXYLATES AND NONYLPHENOL IN FISH TISSUES OF MICHIGAN, USA By Timothy Lawrence Keith A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Zoology 2000 ABSTRACT IDENTIFICATION AND QUANTITATION OF NONYLPHENOL ETHOXYLATES AND NONYLPHENOL IN FISH TISSUES OF MICHIGAN, USA By Timothy Lawrence Keith Persistent metabolites of nonylphenol polyethoxylates (NPE, n=1-18), such as nonylphenol (NP), nonylphenol monoethoxylate (NPE1) and nonylphenol diethoxylate (NPEz) are weak estrogenic environmental contaminants and have been implicated in the disruption of endocrine function in wildlife. In order to evaluate bioaccumulation potential and to identify potential risks posed by these chemicals, concentrations of NP and NPE1, NPE2, and nonylphenol triethoxylate (NPE3) were determined in the tissues of fish inhabiting various waters in Michigan, USA. To measure these concentrations, a method was developed to extract samples using exhaustive steam distillation with concurrent liquid extraction. Concentrations of NP among all sites and species ranged from <3.3 to 29.1 nglg wet wt and varied little statistically among sites. NPE1 was detectable but at concentrations below the limit of quantitation. NPEz and NPE3 were below their method detection limits of 18.2 and 20.6 nglg wet wt, respectively. DEDICATION I would like to thank Dr. John P. Giesy for his guidance and support. I am grateful for the advice of my committee members, Dr. Robert Huggett and Dr. Susan Masten. I am indebted to Dr. Kurunthachalam Kannan and Dr. Shane Snyder of the Aquatic Toxicology Laboratory, Cheryl Summer from the Michigan Department of Environmental Quality, and my Laboratory Technician Adam Pitt for all their help. Thanks to my colleagues in the Aquatic Toxicology Laboratory and to all my family and friends who have made this possible. Most importantly, I thank my parents, Ray and Joan Keith, for their belief and support throughout the years. ACKNOWLEDGMENTS I would like to acknowledge the Alkylphenol Ethoxylate Research Council for providing funding and technical guidance for this project and the Michigan Department of Environmental Quality for their assistance in sample collection. TABLE OF CONTENTS LIST OF FIGURES LIST OF ABBREVIATIONS Introduction Bibliography CHAPTER 1: METHOD DEVELOPMENT Introduction... . . Glassware Preparation Test Matrices. SoxhletExtractioni...............::::::::::::::::::12:21:22..21:21:...12121221121223:ZZZ-...: Lipid Removal 1 Lipid Removal 2... Sample Concentration .. Liquid Chromatography and Detection Reverse Phase Liquid Chromatography Preliminary Results... Steam Distillation with Concurrent Liquid- LiqUId ExtractionT... Normal Phase Liquid Chromatography .. Cyano Column... Silica Gel Column Extraction. GIasswareCleanmé Final Extraction Method Attempted Derivatization...... ..I'.'..'.'.I'.'.I'.'.I'.'.I'.'.I'.'.TILT.IT.ITLCIZIIZIZIXTL Internal Standard.......................... .. Method Validation................................ Recovery and PreCISIon Conclusmn Bibliography CHAPTER 2 FIELD APPLICATION Introduction... .. Overview of stddy Area Fish and Sampling... Standards and Reagents... 2']. IT. vii viii ix 37 38 40 41 TABLE OF CONTENTS, CONTINUED Extraction... . Results and Discussmn .. Estimated Water Concentrations and Bioconcentration Factors. Range of Possible Concentrations...............................................”......: Conclusmn Bibliography APPENDICES................................ . 41 42 47 47 50 51 55 60 69 75 LIST OF TABLES Table 1. Analyte ions monitored for GC/MSISIM. Page 14. Table 2. Comparison in NP recovery between soiled and new extractors. Page 26. Table 3. Recoveries of spiked analytes. Page 33. Table 4. Instrumental and method detection limits. Page 34. Table 5. Detectable concentrations of NP in tissue across species. Page 44. Table 6. Range of NP tissue concentrations by site and specie. Page 45. vii Figure 1. Figure 2. Figure 3. Figure 4. Figure 5. Figure 6. Figure 7. Figure 8. LIST OF FIGURES HPLC chromatogram showing goldfish homogenate spike recovery extract using large concentration of analytes (approximately 1 pg NP/mL extract concentration). Page 15. HPLC chromatogram showing goldfish homogenate spike recovery extract using trace concentration of analytes (approximately 50 ng NP/mL extract concentration). Page 16. Fish weight to acid volume comparison. For the first row of numbers, the first value represents the mass of fish homogenate weight in g and the second value represents the volume of concentrated sulfuric acid in mL. Page 24. Normal phase HPLC chromatogram showing collected fraction. Page 30. GCIMS chromatogram showing standard solution of 4th, NP, NPE1 and NPEz. Page 32. Location of sampling sites. Page 39. Range of NP tissue concentration, nglg tissue, wet weight. 0 = All samples below MDL. Vertical bars represent sites that are not significantly different. See Figure 6 for site abbreviations. Page 46. NP tissue concentration using proxy values for non-detects. Species: WS = White sucker, RB = Rock bass, GS = Green sunfish, LNS = Longnose sucker, 66 = Bluegill sunfish, SMB = Smallmouth bass. See Figure 6 for site abbreviations. Page 49. viii LIST OF ABBREVIATIONS 4tb = 4-tert-butyl ortho-cresol ACN = acetonitrile AP = alkylphenol APE = alkylphenol ethoxylate CP = para-cumylphenol CV = coefficient of variation DCM = dichloromethane ECD = electron capture detector GC = gas chromatograph, gas chromatography GC/MS = gas chromatograph/mass spectrometer, gas chromatography/mass spectrometry GPC = gel permeation chromatography HP = Hewlett-Packard HPLC = high pressure liquid chromatograph high pressure liquid chromatography IDL = instrumental detection limit i-Ca = iso-octane LC = liquid chromatography mlz = mass-to-charge ratio MDL = method detection limit MeOH = methanol NP = 4-Nonylphenol NPE = nonylphenol ethoxylate NPE1 = nonylphenol monoethoxylate NPEz = nonylphenol diethoxylate NPE3 = nonylphenol triethoxylate NoP = normal phase OP = octylphenol OPE = octylphenol ethoxylates ppb = part per billion (nglg, ng/mL) ppm = part per million (pg/g, pglmL) ppt = part per trillion (pg/g, pg/mL) PE = Perkin Elmer PDMS = polydimethylsiloxane RP = reverse phase SIM = selected ion monitoring TFAA = trifluoroacetic acid WWTP = wastewater treatment plant w = wet weight INTRODUCTION Alkylphenols and alkylphenol ethoxylates (APEs) have numerous applications, including pesticide formulations, petroleum production, cleaning products, pulp and paper manufacturing, and plastics manufacturing (Metcalfe et al., 1996). Approximately 80% of the APEs used are nonylphenol ethoxylates (NPEs), while the remaining 20% are almost entirely octylphenol ethoxylates (OPEs). The hydrophilic moiety is an ethoxylate chain ranging from 1 to 20 ethoxy units while the hydrophobic moiety is a branched alkyl group typically with 8 or 9 carbons. These compounds enter wastewater treatment plants (WWl'Ps) where they may undergo degradation. The degradation intermediates, nonylphenol and its mono, di and triethoxylates tend to be more persistent, so are often the dominant NPE species in WWTP effluents (Ahel et al., 1993a; McLeese et al., 1981). This degradation resistance is a result of the alkyl branching and the presence of the aromatic ring. They are lipophilic and tend to adsorb to organic surfaces (Metcalfe et al., 1996). Wastewater treatment in the United States generally removes around 95% of all the NPEs entering the plant (Naylor, 1995). It has been shown that NP is an estrogen mimic and can interfere with the reproduction of fish. NP at low pg/L concentrations can induce the production of vitellogenin in cultured rainbow trout hepatocytes (Jobling and Sumpter, 1993). NP is also capable of inducing cell proliferation in the estrogen-sensitive MCF-7 human breast tumor cells (Soto et al., 1991). The estrogenic potency of NPEs appear to decrease with increasing ethoxylate chain length. However, the water solubility of NPE increases with increasing ethoxylate chain length (Ahel and Giger, 1993b). The first objective of this study was to develop a reliable, cost-effective, and simple method to sensitively detect and quantify NPE1-3 and NP in the tissues of fish using commonly available equipment. The second objective was to measure concentrations of NPE1-3 and NP in fish from various rivers of mid-Michigan and the Great Lakes, USA. BIBLIOGRAPHY Bibliography Ahel, M; McEvoy, J; Giger, W.; 1993a. Bioaccumulation of the lipophilic metabolites of nonionic surfactants in freshwater organisms. Environ. Poll. 79, 243-248. Ahel, M. and Giger, W., 1993b. Aqueous solubility of alkylphenols and alkylphenol polyethoxylates. Chemosphere 26, 1461-1470. Huntsman Corporation, Austin, TX, 1994. Method of test and standard operating procedure for determination of nonylphenol and nonylphenoxyethanol in environmental water by steam distillation and high performance liquid chromatography. Method No. ST-38.34-94. Jobling, S. and Sumpter, JP, 1993. Detergent components in sewage effluent are weakly oestrogenic to fish: An in vitro study using rainbow trout (Oncorhynchus mykiss) hepatocytes. Aquatic Toxicol. 27, 361 -372. McLeese, D. W.; Zitko, V.; Sergent, D. B.; Burridge, L.; Metcalfe, D. D., 1981. Lethality and accumulation of alkylphenols in aquatic fauna. Chemosphere 10, 723-730 Metcalfe, C.; Hoover, L.; Sang, S., 1996. Nonylphenol ethoxylates and their use in Canada. World Wildlife Fund Canada, 90 Eglinton Avenue East, Suite 504, Toronto, Ontario M4P 227, 33+ p. Naylor, CG, 1995. Environmental fate and safety of nonylphenol ethoxylates. Text. Chem. Color. 27, 29-33. Soto, A.M.; Justica, H.; Wiay, J.W.; Sonnenschein, C., 1991. P-Nonyl-phenol: An estrogenic xenobiotic released from ‘modified' polystyrene. Environ. Health Perspectives 92, 167-193. Veith, GP. and Kiwus, L.M., 1977. An exhaustive steam distillation and solvent extraction unit for pesticides and industrial chemicals. Bull. Environ. Contam. Toxicol. 17, 631-636. Chapter 1 METHOD DEVELOPMENT Introduction When faced with the challenge to develop a reliable, cost-effective and simple method to sensitively detect and quantify nonylphenol (NP), nonylphenol monoethoxylate (NPE1), nonylphenol diethoxylate (NPEz), and nonylphenol triethoxylate (NPE3) in the fish tissues, several obstacles had to be overcome. One was to develop or adapt a pre-existing procedure to prepare the sample for extraction and to extract the analytes from the tissue matrix. The second and most difficult was separation of lipid interferences from the sample extract. The third was to develop a procedure for quantifying the analytes. Glassware Preparation All glassware was washed using a quality commercial glassware detergent with deionized water, rinsed three times with high purity acetone, followed by three rinses with high purity hexane and allowed to dry before use. Test Matrices Two different matrices facilitated the method development. The sample matrix used for extraction development was homogenated laboratory raised whole body goldfish (Carassius auratus). Goldfish were removed from a -20 °C freezer and allowed to thaw. Goldfish were quartered and homogenized using a Sorvall OmniMixer blender in 400 mL OmniMixer blender cups. Homogenate was stored in glass jars at -20 °C until needed. The homogenate was spiked with external standards of NPE1.3 and NP to represent an environmentally exposed sample. The second matrix used to develop clean up and subsequent quantification was corn oil. Corn oil was used as a surrogate for fish lipids in an attempt to reduce time in sample preparation. To simulate actual samples, a percent lipid of 15% w/w was assumed and the corn oil spiked appropriately. This percentage was chosen since the extracted goldfish homogenate contained 11 to 25% lipid ww. Soxhlet Extraction Soxhlet extraction was the first method explored. Ten 9 of the pre-homogenated tissue sample was placed into an OmniMixer cup with 50 g anhydrous sodium sulfate (Na2804) and blended for 1 min. The mixture was placed in a —10 °C freezer for 5 min to assist in desiccation. Two more additions of 50 g of sodium sulfate were added to the mixture and homogenized as previously described. The sample was then spiked with 50 uL of a mixture containing 100 pglmL each of octylphenol (OP) and NP in acetonitrile (ACN) for a total concentration of 500 nglg. The sample was Soxhlet extracted overnight with 350 mL of dichloromethane (DCM). The extract was then cooled and stored out of light until lipid removal could be performed. The total time for this extraction procedure varied in total from 18-24 hr. Lipid mass was determined by removing 100 III. of the extract and placing it into a pre-weighed aluminum weigh boat. The sample was then placed into a drying oven at 60 °C until the solvent had evaporated. The sample was weighed again and the percentage lipid calculated. Goldfish tissues ranged from 11 to 25% lipid or 1.1 to 2.5 g per 10 9 sample. Among the advantages of this Soxhlet extraction method were its common availability to analytical laboraton'es and ease of operation. Samples could remain relatively unattended for the majority of their extraction. Until a lipid clean-up method was developed, this extraction procedure was considered adequate. Lipid Removal Method 1 Following the Soxhlet extraction, each sample was concentrated using a Labconco TurboVap at 30° C to a final volume of approximately 5 mL. The concentrated sample was transferred with DCM to a 15 mL centrifuge tube and evaporated under nitrogen to a final volume of 6 mL. Several gel permeation chromatography (GPC) elution profiles were tested. The GPC system consisted of a Rheodyne 7725i injector (Cotati, CA) with a 5 mL sample loop, a quaternary high-pressure liquid chromatography (HPLC) pump (Perkin Elmer, Series 410, Norwalk, CT) and an electronic fraction collector (lSCO Foxy 200, Lincoln, NE). Mass loading, mobile phase flow rates and the number and type of columns were investigated. The largest mass of interferences that could be loaded into the GPC system while maintaining acceptable lipid removal was determined to be 0.5 9. Therefore, three separate injections of 2 mL of the 6 mL extract were introduced through a Rheodyne injection port into an isocratic HPLC system. A Phenomenex Envirosep-ABC 350 mm X 21.20 mm column followed by a Phenomenex Phenogel 5A 300 mm X 21.20 mm column was used to obtain the greatest lipid removal. DCM was used as the mobile phase at a flow rate of 5 mL/min. Ten fractions were collected every 2 min starting at 10 min to determine where the APs would elute. Fractions were evaporated under nitrogen and solvent exchanged into ACN. Extracts were analyzed by reverse phase HPLC using fluorescence detection. Recovery of NP was 71.0% +l- 8%. During the solvent exchange to ACN, it was observed that residual lipids would precipitate and could be removed by centrifugation. While some lipid did remain in the ACN solution, the addition of 0.5 mL of water resulted in a greater lipid precipitation (removal). This led to an alternative lipid removal procedure summarized in the following section. Lipid Removal Method 2 Preliminary testing of this method used corn oil as a surrogate for fish lipids in an attempt to reduce time in sample preparation. Experiments were conducted with various mixtures of ACN, ACN/water, methanol (MeOH), and MeOH/water. The following describes some of the more successful trials. In a 15 mL centrifuge tube 4 mL DCM, 1 mL corn oil and 50 IIL of a 100 ppm alkylphenol (AP) mixture in ACN were vortexed for approximately 1 min. Two mL MeOH were added and vortexed for 5 min. The mixture was evaporated under nitrogen to 3 mL and vortexed again. The sample was centrifuged at 3500 RPM for 5 min and the supernatant transferred into another 15 mL centrifuge tube. Two 2 mL of MeOH added to sample, vortexed and centrifuged and supernatant transferred to the same tube. This was repeated. The supernatant was concentrated under nitrogen to 1 mL in ACN and quantitated using the same HPLC method described in Method 1. This method was then tested with actual fish lipids using Soxhlet extracts from 10 9 fish tissue. The extracts were concentrated by rotary and nitrogen evaporation to 5 mL and Lipid Removal Method 2 applied. NP recovery in corn oil was 84.1% +/- 9% and 66.8% +/- 5% in fish homogenate. Sample Concentration As all the aforementioned techniques require varying degrees of solvent concentration, different techniques were investigated. For bulk concentration, the Buchi RotoVap and the Labconco RapidVap N2 Evaporation System were 10 tested and the Organomation 25 position and 100 position N-Evap for fine solvent concentration. Logistically, the Labconco RapidVap N2 Evaporation System is superior to the Buchi Rotary Evaporator. The RapidVap is programmable to regulate temperature, vortex speed and evaporation time. The RapidVap sample container is engineered to eliminate the risk of complete solvent evaporation that may cause loss of volatile analytes. In addition, six samples may be simultaneously concentrated with no need for monitoring during operation. Conversely, the Buchi Rotary Evaporator requires monitoring during operation and accommodates only one sample but the concentration occurs much faster. For example, 100 mL of MeOH evaporated to approximately 2 mL at 30 °C requires 2 hr for the RapidVap and only 0.5 hr for the RotoVap. Aside from logistics, recoveries of analytes were also investigated. Spike recoveries of NP spiked into 100 mL DCM were approximately 20% greater using the RapidVap than the RotoVap (97.3% to 78.5%). From this information, it was decided to employ the Labconco RapidVap N2 Evaporation System for bulk solvent concentration. Spike recoveries comparing the 25 position and the 100 position Organomation N-Evap were nearly identical. This is expected as both these systems utilize the same evaporation technique of gentle streaming nitrogen and a heated water 11 bath. However, beside the apparent advantage of increased sample throughput using the 100 position unit over the 25 position unit, the 100 position unit also allowed for greater control of nitrogen flow. While the 25 position unit uses one nitrogen flow control per sample, the 100 position unit uses one flow control per 10 samples. This ensures that the 10 samples controlled by the same flow control receive the same flow of nitrogen, decreasing sample variability. For this reason, the Organomation 100 position N-Evap was chosen for fine solvent concentration. To test this model’s reproducibility, eight spiked samples containing 10 pg of NP in ACN were concentrated from 6mL to 1mL. The average recovery of NP was 95% with a coefficient of variation (CV) less than 10%. Liquid Chromatography and Detection Despite its lower resolution and some coelution problems described below, liquid chromatography (LC) is a very versatile analytical technique suitable for compounds with a wide range of polarity, volatility and molecular weights such as the NPEs, as well as the isomers of NP. As the NPEs and NP have a strong chromophore (phenyl group), one of the most sensitive and selective detection methods is fluorescence. As the NPEs and NP have their excitation and emission spectra maxima at approximately 230 nm and 310 nm respectively, researchers have often used these wavelengths for detection of NPEs and NP. 12 Another important consideration to the LC method development was the ability to separate NP from OP, another common surfactant found in the environment. If the LC method could not resolve OP from NP, the calculated NP concentration would be suspect as a summed concentration of OP and NP. HPLC coupled to a programmable fluorescence detector was initially investigated. This HPLC system consisted of a Perkin Elmer (PE) (Nomalk, CT) series 200 autosampler, a PE series 200 binary pump, a Hewlett Packard (HP, Palo Alto, CA) 1046A fluorescence detector, a Degasys (Tokyo, Japan) electronic vacuum degasser, a PE Nelson series 900 interface and PE TurboChrome 4.0 data software package. Reverse Phase Liquid Chromatography As all the NPE oligomers have the same hydrophobic moiety, they elute as a single peak using the non-polar solid phase octadecylsilica. As such, this would require an additional analytical procedure to analyze the various oligomers of NPE. Gas chromatography/mass spectrometry (GC/MS) using electron impact ionization was proposed as a possible means to confirm the identification of NP and to identify and quantify the NPEs. The instrumental profile for the RP-HPLC method that best separated the APs is described below. Elutlon solvents were reagent water and ACN delivered at a 13 constant flow rate of 1 mL/min. The elution profile was a 20 min gradient (curve = -2) from 50% water/ACN to 2% water/ACN followed by a 10 min isocratic ACN purge. The column was returned to initial conditions by a 10 min isocratic flow of 50% water/ACN. The HPLC injection volume was 10 pL. For all compounds of interest, the fluorescence detector settings were 229 nm excitation and 310 nm emission. An HP 5890 Series II Plus GC and an HP 5972 electron impact ionization mass spectrometer (MS) were used for further separation and identification. The GC was operated using a 30 m DB-17 column (J&W Scientific, Folsom, CA), starting temperature of 100 °C for 2 min, ramped to 300 °C for 10 min at 4 °Clmin. The GC injection volume was 4 pL. To detect the analytes, the MS was operated in selected ion monitoring (SIM) (T able 1). Table 1. Analyte ions monitored for GCIMS/SIM Analyte Ions (mlz) Monitored NP 107,135, 149 NPE1 135, 179, 193 NPE2 135,223, 237 NPE3 135,267,281 Preliminary Results This combined technique of RP-HPLC and GCIMS worked well for standard mixtures of APs. In addition, this profile was sufficient for separation and 14 identification for goldfish homogenate spikes when using large amounts of analytes (approximately 1ug NPImL extract concentration) (Figure 1). However, this method was insufficient for analytes that were spiked in trace amounts (resulting in an extract concentration of approximately 50 ng NPImL) (Figure 2). \OP Time (min) Figure 1. HPLC chromatogram showing goldfish homogenate spike recovery extract using large concentration of analytes (approximately 1pg NPImL extract concentration). The remaining interferences, even after lipid removal using the technique described above, caused a high signal-to-noise ratio that prevented any 15 identification of analytes in trace quantities. The problem was compounded by the HPLC operating in reverse-phase, allowing the nonpolar residual lipid to elute over the entire elution. Time (min) Figure 2. HPLC chromatogram showing goldfish homogenate spike recovery extract using trace concentration of analytes (approximately 50 ng NPImL extract concentration). To remedy this, a post-extraction clean-up method needed to be developed. Miniature silica gel columns were constructed using 0.5 g silica gel (100/200 16 mesh) in a 5% inch Pasteur pipette. The column was wetted with 3 mL hexane and the extract loaded. The extract was first eluted with 3 mL DCM, then 3 mL MeOH. This second fraction was again analyzed by GCIMS. Concentration of lipids remained too high and quantitation by GCIMS was determined unsuitable. An additional approach was required. Steam Distillation with Concunent Liquid-Liquid Extraction Steam distillation separates chemicals based on vapor pressure differences over water (Veith and Kiwus, 1977). Veith and Kiwus developed a modified steam distillation apparatus that allows steam distillation with concurrent solvent extraction. This allows an exhaustive and continuous extraction of the matrix of interest. The flask containing the sample is heated to produce a vigorous boil. Steam distillate passes through an inner tube and condenses on the walls of the cooling jacket. Condensate flows down the condenser and through a layer of non-polar solvent of lower density than water. The water passes through this solvent layer and is recycled through an overflow tube back into the boiling flask and the cycle repeats. Once the desired amount of time has passed, the solvent can be removed through a side arm on the base of the condenser. This technique allows direct analysis of most extracts without additional concentration and clean up while using very little solvent. Although this method has utility for water, sediment, and sludge samples, few researchers have 17 investigated this method for biological matrices. One group reported successful attempts using this technique (Ahel et al., 1993). Aside from the benefit of no projected extract clean-up, this method offers several other advantages. Homogenization efficiency is greatly increased since this method uses water as the primary solvent. As the matrix is suspended in water causing the tissue to remain in continuous contact with the blender blades, dry homogenization requires frequent pauses to remove the sample from the container walls and back into the blender blades. In addition, there is no need for sample drying. This method is also environmentally friendly since it utilizes only tens of milliliters of non-aqueous solvent. Experiments as to the suitability of normal phase liquid chromatography (NoP- LC) were conducted since the extract would be solvated in a non-polar solvent and with the complications noted earlier using RPLC. Normal Phase Liquid Chromatography Since NPE oligomers differ from each other by the length of their polyethoxylate chain, they are best separated by NoP-LC. For NoP-LC, the order of elution is a function of number of ethoxylate groups: the greater the number of ethoxylates the longer the residence on the LC column. 18 Two columns were investigated: a Phenomenex Phenosphere 5p CN 80A (250 mm x 4.6 mm) and a Phenomenex Luna 5p silica (250 mm x 4.6 mm). Mixtures containing APs and APEs were prepared. These mixtures contained OP, NP and NPE1 at varying concentrations. Cyano Column Initially, a generic NoP-HPLC method using a cyano column was employed on spiked fish extracts. The elution profile was a 30 min linear curve from hexane to MeOH at 1 mL/min. This method proved to be hindered by matrix interferences and was unable to resolve OP from NP and NPE1. The complete resolution of OP was considered an important requirement. Along with NP, OP is another common environmental contaminant. If OP was not resolved from NP, any calculated concentration of NP had the possibility to be the summed concentration of NP and OP. A less aggressive solvent elution was necessary to allow the lipids to elute before the APs and APEs. Elutlon solvents considered were hexane combined with various percentages of 30% 2-proponal/MeOH, 20% DCM/MeOH, and 20% MeOH/DCM using curves ranging from -2 to 2 and flow rates between 0.5 to 1.5mUmin. These methods met with limited success: APs and APEs could be separated from the lipid interferences but couldn’t be resolved from each other, or APs could be somewhat resolved from APEs but were hindered by matrix interferences. As the method became less aggressive in order to provide adequate separation, the NP peak began to split into two peaks and eluted over a long period of time (approximately 3-4 min). 19 OP could not be completely resolved from NP. Standards of OP and NP were injected independently and when chromatograms of the two were superimposed, the column showed complete resolution. Injection of mixtures using the same method resulted in incomplete resolution. It was necessary to resolve OP from NP; however, complete resolution was never achieved using this column deeming it unsuitable for this study. Silica Gel Column Knowing silica gel has a much higher binding ability than cyano, it was hoped that silica gel could provide sharp, resolved peaks of all compounds and lipid interferences. The approach to using the silica column was the same as the cyano: slow flow rates and non-aggressive solvents to allow the APs and APEs adequate time to separate from the lipids. In early methods, the first 3 min were pure hexane to bleed the lipids off the column, then ramped up at various flow rates and solvent mixtures to elute the APs and APEs. This approach provided much sharper peaks with a higher detector response than did the cyano, but could not resolve OP from NP from NPE1. Again, when OP and NP were injected independently, superimposed chromatograms showed complete resolution. Mixtures of OP and NP at varying concentrations were injected and resulted in a single peak, roughly the average of the two individual retention times and sum of the individual areas. 20 It was discovered that when the silica column is initially filled with hexane, APs could not be resolved from the APEs. When the column is initially filled with a mixture of hexane and 20% MeOHIDCM and eluted isocratically, APs could be sufficiently resolved from the APEs. Still, OP could not be resolved from NP. It was decided that resolution of OP from NP was not necessary for this study. Efforts were redirected to optimize a method using silica gel to resolve NP and NPE1 from fish lipid. The elution profile that was found best to resolve NP from NPE1 with no interference from the fish matrix was an isocratic elution of 12% 1:4 MeOHIDCM and 88% hexane at a flow rate of 0.65 mL/min with fluorescence detection at 229 nm excitation and 310 nm emission. Extraction Experiments were initially designed to reproduce the method from Huntsman Corp. (method #ST-38.34-94, Austin, TX) which reported 100% recoveries of NP from spiked water with relative standard deviation of 2.5% (n=4). The method states that it is applicable to waste water and river water as well as “solid matrices such as sediments, sludge and biological tissues". When applying this technique to fish, it was believed that the extracts would not require additional clean up as the higher molecular weight lipids will not distill, therefore not contaminating the extraction solvent. 21 Experimental parameters for this method called for 20 g sodium chloride (NaCl), 1 L water, boiling chips, 2 mL hexane and cooling jacket temperature at 5 °C. Ten 9 of pre-homogenated sample, NP and NPE standard mixtures in hexane were spiked directly into the 1 L of water. The temperature-controlling rheostat was set to 100% heating. This method produced severe foaming which caused excess material to foam over and accumulate in the organic solvent rendering it unusable for analysis. Careful temperature regulation could prevent the sample from foaming over but would be impractical for a high-throughput method. Spike recoveries of NP and NPE1 were 65.1% and 60.8%, respectively, for those extractions that did not foam significantly. To address the problem of sample foam-over, varying amounts of polydimethylsiloxane (PDMS), an anti-foaming agent, were added at the beginning of extraction. While PDMS did suppress sample foaming, recoveries of NP fell to 53.3%, a decrease of 11.8%. Attempts to contact Dr. Ahel regarding the 100% recoveries of NP from fish tissues he reported in 1993 were unsuccessful (Ahel et al., 1993). Dr. Veith was contacted in mid-October 1998 and recommended the addition of small amounts (1 — 5 mL) of concentrated sulfuric acid to suppress the sample foaming. Initial 22 trials using sulfuric acid indicated that foaming was eliminated and recoveries improved to 68.0%. The addition of acid caused severe bumping despite the addition of various boiling chips, including Quartz, glass beads, crushed glass, and graphite chips. The bumping problem was solved by continuous mixing using a magnetic stir bar in the boiling flask by a stir plate placed beneath the heating mantle. The stir bar suppressed the bumping well enough as to remove the need for boiling chips of any kind. With the method developed where recoveries of NP were consistent at 65-70%, method improvement began by conducting a sensitivity analysis on the reagents and parameters. These included: 0 condenser temperature 0 heating temperature . extraction time a number of extractions o solvent type . solvent volume 0 sulfuric acid volume a sample mass . sodium chloride mass 23 From these sensitivity analyses, different heating temperature showed no significant differences (provided the water came to a boil). Twenty g of salt recovered the greatest amount of analyte. Recoveries differed with varying fish tissue mass to amount of acid ratio (Figure 3). The trend in Figure 3 shows the greatest recoveries are achieved with 1-1.5 mL acid per 10 9 fish homogenate. It could also be concluded that iso-octane (i-Ca) exhibited greater analyte recovery than hexane or cyclohexane. 100.0 - 90.0 80.0 I 70.0 ~ 60.0 a 50.0 ~ 40.0 - -- -~ 30.0 ~ 20.0 ~ 10.0 ~ 0'02 0,1 10,1 10,1 10,3 10,3 10,3 10,5 20,2 23,1 30,3 40,4 45,1 [RecoveryofNR'is 31.3 31.0 35.1 30.1 30.4 31.4 12.1 34.3 52.1 32.3 452 53.2 Recovery of NP, % Figure 3. Fish weight to acid volume comparison. For the first row of numbers, the first value represents the mass of fish homogenate weight in g and the second value represents the volume of concentrated sulfuric acid in mL. Of particular concern was the cooling water temperature, as the laboratory ambient temperature of this water was 13-15 °C. It was unknown if this temperature was too great to sufficiently condense the steam before loss occurred through the apparatus top. A recirculating-chiller using an ethylene glycol/water mixture producing cooling water between -5 and 15 °C was tested to 24 see the effect on spike recoveries. Recoveries did not increase with cooler temperatures and in the extreme case actually decrease with cooling temperatures at -5 °C. This is believed to occur by condensing the steam so quickly as to not pass over the inner tube and the analytes never have the opportunity to flow through the organic solvent layer. In addition, two 1‘/a hr extractions showed greater recoveries than one 3 hr extraction. The sample that was extracted for 3 hr recovered 56.2% of the spiked NP, while the sample that was extracted twice for 1% hr recovered 49.6% first extraction and another 19.8% the second extraction. Glassware Cleaning Cleaning procedures were studied since the steam distillation glassware was in frequent use. The columns are rinsed with acetone and hexane after each extraction. With the problems of sample foaming described earlier, the inner tube become soiled. Upon arrival of two new steam distillation units, an experiment was designed to compare the recoveries of NP between a new unit and a heavily soiled unit (Table 2). 25 Table 2. Comparison in NP recovery between soiled and new extractors. Column Condition Percent Recovery of NP Percent Coefficient of Variation Soiled 38.6 Soiled 39.7 1 .7 Soiled 39.8 New 59.0 New 58.9 0.55 New 58.4 Please note the samples were only extracted once since the trend could readily be seen. This resulted in lower recoveries when compared to samples extracted twice. Soaking the columns for one hr in 17% nitric acid returned the column to near original condition. Extractions using the same parameters shown in Table 2 were run in triplicate after the columns were soaked in acid. The extractions returned a mean recovery of 84.8% for NP with a CV of 5.8%. For information on steam distillation apparatus cleaning, please refer to Appendix A. Final Extraction Method From the sensitivity analyses, an optimal extraction and quantitation method was believed to be found. Twenty g of fish homogenate is homogenated in 600 mL deionized water and transferred to a 2 L boiling flask. Four hundred additional mL of deionized water is used to quantitatively transfer the remaining fish 26 homogenate from the blender cup to the boiling flask. Then, 20 g sodium chloride, 3 mL concentrated sulfuric acid, a small amount of Quartz boiling chips and a Teflon coated magnetic stir bar are added to the flask. Two to three mL of deionized water are added through the top of the steam distillation apparatus followed by 10 mL high-purity i-Ca and the column capped with aluminum foil. The 2 L boiling flask containing the sample homogenate and reagents are attached to the apparatus and placed in the heating mantle. Using the stir plate, a gentle vortex (approximately 50% maximum stir) is created, the heating mantle powered to HIGH and the cooling water flow is set to maximum flow. The sample is extracted twice at 1% hr each extraction. The extract is then concentrated, fractionated by NoP-HPLC and quantitated by GCIMS. For further information, please refer to Appendix B. To achieve greater sample throughput, six steam distillation extractors were operated simultaneously. Cooling water temperature was measured to lessen approximately 1 °C after flowing through one extractor operating at maximum temperature. To reduce the variability between different extractors, two separate cooling water feeds supplied three extractors each. Six replicate spike recovery experiments indicated that sample variability was not significantly greater for six extractors operating simultaneously than one extractor operated alone six times. 27 Attempted Derivatization Derivatization of the analytes was investigated for additional quantitation. To test the suitability of derivatized samples detected by an electron capture detector (ECD), NP was acylated by trifluoroacetic acid (T FAA) converting the active hydrogen into a fluorinated ester. Standards of derivatized NP showed increased detectability using ECD than GCIMS. Experiments were then conducted to derivatize environmental extracts. One approach was to attempt derivatization during extraction by adding TFAA into the organic solvent layer before extraction. Another approach was to derivatize the sample after extraction. The TFAA derivatized lipid interferences in both experiments and resulted in an extract unsuitable for ECD analysis. lntemal Standard lntemal standards were employed to correct for extraction variability and improve data quality. Upon suggested by Carter Naylor of Huntsman, Corp., p- cumylphenol (CP) was investigated as a surrogate standard for NP. Recoveries of p-cumylphenol to measure extraction recovery were 89.0% mean recovery with an 11.3% CV as measured by NoP-HPLC with fluorescence detection. This allowed for correction of NP concentrations due to loss by extraction, concentration or injection into the HPLC. 28 Method Validation As a test for the methods ability to analyze environmental samples, common carp (Cyprinus carpio) from Lake Mead, Nevada were obtained and processed. While using a matrix of goldfish homogenate in laboratory spikes proved the method reliable in separating the compounds of interest from the matrix interferences, an analysis of these carp samples proved to the contrary. That is, the type and/or amount of interferences from one species of fish are not necessarily the same interferences from another. Thus, further steps had to be taken to assure that the analytes could reliably be separated and quantitated. An alternative method of quantitation was implemented to resolve this problem. Using the original method, quantitation was achieved using NoP-HPLC with fluorescence detection using an extract that underwent no clean-up. With the improved method, the NoP-HPLC is now used as a fractionating step removing the bulk of the interferences from the compounds of interest while GCIMS separates and quantitates the analytes. While this added step required approximately two additional hours laboratory processing per sample, the identification of analytes is now accomplished by not only retention time matching but also mass spectral identification. This means of identification is also beneficial since the presence of any polyaromatic hydrocarbons, which have the possibility to interfere with analysis using only 29 fluorescence detection, will not complicate data interpretation. For further information, please see Appendix B. With the method now employing NoP-HPLC for fractionation and GCIMS for identification and quantitation, further method development was required. With the previous method, NoP-HPLC quantitated NP and CP and GCIMS quantitated the NPEs. The new method required the GCIMS to separate and quantify CP as well. A fraction from the HPLC collected between 7—16 min would not collect the bulk lipids while collecting NP, NPEs and CP (Figure 4). I—— Collected Fraction fi' l Lipids 0 2 4 6 810121416 Time (min) Figure 4. Normal phase HPLC chromatogram showing collected fraction. “Lb-—-—-—-——--—-——— rip—-‘fiTfi—-——--—--- 1 ——J I ...2 ‘ —4 —4 30 As this work progressed, however, it became apparent that CP could not be adequately separated from NP by GCIMS. The GCIMS method worked well for NPE and NP quantitation. The inability to separate CP from NP was overcome using an additional internal standard and not SIM the CP. This lntemal standard would measure the loss of analyte due to solvent concentration and subsequent injection into the GC. For this internal standard, 4-tert-butyl ortho-cresol (4tb) was investigated. Spike recoveries of 4th using the same volume of solvent that would elute for a fractionation to simulate an actual sample were conducted. Recoveries of the lntemal standard 4tb are a 92.2% mean recovery with a 10.9% CV. See Figure 5 for GCIMS chromatogram showing 4tb, NP, NPE1 and NPE2. 31 i \4tb .. - 444_J_4__4A___4_L_444 .4444 1.. 4A. 4344;4__4444L_..___ L4; .4444 44.4.4441». ‘ 4 A A l A A A A J A A L I I NPE " I I i —‘ NPE2 I I :— 4 II I ii I II I II I I II 1 I ? I I I l 3 I I .I II I I I} II II III I III III I ii I 1 Ike wllllllllld TTTleTDT IT\T4T6TOTTTI TM? l5llelTTTTllldTOl | i I Time (in I n) Figure 5. GCIMS chromatogram showing standard solution of 4th, NP, NPE1 and NPE2. 32 \Mth the improved separation and detection method, NP and NPE1 were detected in the Lake Mead carp at average concentrations of 184 +/- 4 nglg and 242 +/- 9 nglg, wet weight, respectively. NPE2-3 were not detected in any carp collected at Lake Mead. Recovery and Precision Recoveries of NP and NPE1-2 were greater than 70% with CVs less than 20%. NPE3 was found not to have sufficient volatility to have adequate recovery (Table 3). Table 3. Recoveries of spiked analytes. Analyte Percent Recovery Percent Coefficient of Variation NP 78.1 9.2 NPE1 76.1 9.8 NPEz 69.4 13.0 NPEs 17.0 20.1 Instrumental detection limits (lDLs) were determined by analyzing dilute standards (near the estimated IDL) and calculating the signal-to-noise for each standard concentration. Linear regression of the signal to noise ratios against concentrations was then used to determine the IDL (signal to noise ratio = 3). Seven replicates of homogenized goldfish tissues were spiked with NP and NPE1-3 at the estimated method detection limits (MDLs) to determine recovery 33 and precision (Table 4). MDLs were calculated by multiplying the standard deviation of the recovered concentrations by a t-value of 3.1427 (for n=7 replicates). Table 4. Instrumental and Method Detection Limits. Analyte IDL (ng) MDL (pg/kg) NP 5.1 3.3 NPE1 15.5 16.8 NPEz 17.3 18.2 NPEs 112 20.6 Conclusion This method was shown reliable to identify and quantify NP and NPE” in trace amounts and NPE3 if present in sufficient quantity in fish tissues. With a reliable method now developed, the next facet of the study was to apply this method to determine concentrations of NP and NPE”, in fish tissues in Michigan rivers and lakes. BIBLIOGRAPHY 35 Bibliography Ahel, M; McEvoy, J; Giger, W.; 1993. Bioaccumulation of the lipophilic metabolites of nonionic surfactants in freshwater organisms. Environ. Poll. 79, 243-248. Huntsman Corporation, Austin, TX, 1994. Method of test and standard operating procedure for determination of nonylphenol and nonylphenoxyethanol in environmental water by steam distillation and high performance liquid chromatography. Method No. ST-38.34-94. Veith, GP. and Kiwus, L.M., 1977. An exhaustive steam distillation and solvent extraction unit for pesticides and industrial chemicals. Bull. Environ. Contam. Toxicol. 17, 631 -636. 36 Chapter 2 FIELD APPLICATION Introduction The most comprehensive survey of nonylphenol from the United States reported water and sediment concentrations of NP and NPEs from 30 rivers that are influenced by municipal or industrial wastewater effluents (Naylor et al., 1992). That study found that 60-75% of water samples had no detectable levels of NP, NPE1, or NPE; while 30% of sediments were non-detect. However, reports on the concentrations of APs and APEs in fish in the US. waters are scarce. Recently, NP and NPE have been detected in water of the Las Vegas Bay of Lake Mead, Nevada (Snyder et al., 2000a). While information on the occurrence of NP in water and sediments is available, due to the lack of suitable analytical techniques, very few studies have examined the occurrence of nonylphenol in fish. Monitoring of NP in fish is important to assess the potential for dietary exposure of humans and wildlife. A method was developed for the analysis of NP and its ethoxylates in fish (Snyder et al., 2000b). In this study, the method was applied to determine concentrations of NP and NPE 1.3 in fish from Michigan waters. The objective of this study was to measure concentrations of NP and NPE”. in fish from various rivers in Mid-Michigan and the Great Lakes, USA. Since these 37 fish were not caged, samples from a particular “site" actually represent a segment of water that may range approximately 3 km up or downstream from the sampling location. 'Thus, analyte concentrations should be associated with a river segment, rather than a specific point. This information will prove valuable to guide further studies. Overview of Study Area Sampling sites were chosen to represent the ambient environmental concentrations of the compounds of interest (COI) in fish. Fish were collected with assistance from the Michigan Department of Environmental Quality from two major regions: the Kalamazoo River Basin, MI and Lake Michigan near the mouth of the Kalamazoo River (Figure 6). The Kalamazoo River flows through both urban areas and rural areas and receives secondary and tertiary WWTP effluent and industrial discharges including those of paper manufacturing facilities. For information on sample transfer from the Michigan Department of Environmental Quality, please see Appendix C. Sampling along the Kalamazoo River was conducted up and downstream of VWVTPs whenever possible. Fish were captured within approximately 40 m of the WWTP effluents. The Kalamazoo WWTP and the Battle Creek VWVI'P have tertiary treatment. The Portage, Allegan, Marshall, Gun Lake, Augusta, Albion and Otsego WWTPs employ secondary treatment. 38 Site Abbrevlation LG Looking Glass River AL Allegan MA Marshall BC Battle Creek AGUS Augusta upstream AGDS Augusta downstream PO Portage KC Kalamazoo PL Plainwell GR Gun River OT Otsego LA Lake Allegan SC Swan Creek LM Lake Michigan Figure 6. Location of sampling sites. 39 Fish and Sampling Fish species were selected for analysis based on several considerations. These include availability at sampling sites, size (weight), migratory behavior and placement in the food chain. Since the fish were meant to represent an area within a river, only less migrating species were preferred. Also, fishes that reside primarily in the middle depths of the water column were preferred to best represent the exposure to dissolved analytes and not those bound to sediments. Further, species that are classified as game fish were preferred. Species analyzed include Rock bass (Ambloplites rupestn‘s), Bluegill sunfish (Lepomis macrochirus), Green sunfish (Lepomis cyanellus), Smallmouth bass (Micropterus dolomieur), White suckers (Catostomus commersonr), Longnose suckers (Maxostoma macrolepidotum), and Rainbow smelt (Osmerus mordax). Fish were collected by electroshocking on three occasions between late-July and early-November 1999 and stored at —20 °C until analysis. Since the objective of this study was to detect NP and NPE in fish tissue, the area of the fish where NP was likely to accumulate was chosen. Fish were cut at the mid-section, which comprised almost entirely liver and gut and this portion used for the analysis. This section was utilized since the chosen as NP concentrates in the digestive/excretory system (Liber et al., 1999). For information on sample preparation and homogenization, please see Appendix D. 40 Standards and Reagents High purity standards (396% purity) of p-nonylphenol (NP), p-cumylphenol (CP), and 4-tert-butyl orthocresol (4tb) were obtained from Schenectady International (Freeport, TX). Standards of NPE1.3 were obtained from Huntsman Corporation (Austin, TX). High purity pesticide residue grade n-hexane, dichloromethane (DCM), and iso-octane were obtained from Burdick and Jackson (Muskegon, MI). Organic-free water was obtained by purification of reverse osmosis treated water followed by NanopureTM (Bamstead, Dubuque, IA) treatment. All glassware and stainless steel homogenization equipment was rinsed with organic-free water followed by high purity pesticide residue grade acetone and n-hexane. ACS reagent grade sodium chloride was obtained from JT Baker (Phillipsburg, NJ). Reagent grade concentrated sulfuric acid was obtained from EM Science (Gibbstown, NJ). Extraction Extraction and quantitation methods of nonylphenolics are described in detail elsewhere (Snyder et al., 2000b). Briefly, a 20 g representative cross-section was removed from the sample and homogenized (Blender 700, Waring Corporation, New Hartford, CN). This homogenate was transferred to a boiling flask and 20 g sodium chloride and 3 mL concentrated sulfuric acid were added. This homogenate was then extracted using a Nielsen-Kryger improved version steam-distillation column (Ace Glass, Vineland, NJ) for 3 hrs. The resulting extract was concentrated to 1 mL in iso-octane using a Nitrogen Evaporator 41 (Organomation Associates, Inc., Berlin, MA). To further remove lipids from the sample extract, a Perkin-Elmer (Nonlvalk, CT) series 200 autosampler and binary pump and a Hewlett Packard (HP) (Palo Alto, CA) 1046A fluorescence detector was employed to separate lipids from the compounds of interest. Eight hundred DL of the iso-octane extract was separated using a Phenomenex Luna 5 pm silica column (250 mm x 4.6 mm, Torrance, CA) by a 0.65 mUmin isocratic elution using 12% 1:4 MeOHzDCM and 88% hexane. Fluorescence detection was used to determine surrogate recovery during this fractionation. A fraction of HPLC effluent was collected between 7 and 16 min, 3.0 pg 4-tert—butyl orthocresol added as an internal standard and concentrated under nitrogen to 100 uL iso-octane. Compounds of interest were identified and quantified using a HP 5890 Series II Plus GC and a HP 5972 MSD. Separation was accomplished using a 30 m DB-17MS capillary column (0.25 mm ID, 0.15 um film, J&W Scientific, Folsom, CA). The GC was held at 100 °C for 2 min and ramped to 300 °C for 10 min at 4 °C/min. The MSD was operated in selected ion monitoring (SIM) mode with 3 ions monitored for each compound of interest. lntemal standard recoveries were greater than 80 %. Matrix spike recoveries were greater than 70 % for NP and NPE1-2, 17 % for NPE3. MDLs for NP, NPE1, NPE2, and NP53 were 3.3, 16.8, 18.2 and 20.6 nglg, ww, respectively. Results and Discussion Concentrations of NP greater than the MDL were found in 75 of 197 (38%) samples across all sites and species with a mean concentration, excluding non- 42 detects, of 12.0 ng NP/g, wet weight (w) with a range of 3.3 ng NPlg, w to 29.1 ng NP/g, ww (Figure 7). If non-detects are included, the mean NP concentration across all sites and species is 4.0 ng NP/g, ww. NPE1 was found in 21 samples but all at concentrations below the calculated MDL (16.8 ng NPE1lg, ww). NPE2 and NPE3 were not detected in any of the samples. Five of the seven species contained detectable concentrations of NP. Rock bass (Ambloplites mpestn's) contained the greatest average detectable NP concentration, 8.1 ng NP/g, ww, while the rainbow smelt (Osmerus mordax) exhibited the second greatest detectable tissue concentration at 7.7 ng NP/g, ww (Table 5). Longnose sucker (Maxostoma macrolepidotum) and green sunfish (Lepomis cyanellus) contained no detectable concentrations of NP. There were no significant differences in concentrations of NP among species so the samples were pooled for comparison among sites (ANOVA with a Type I error of 0.1 followed by a Tukey’s Studentized Range Test). However, there were significant differences in concentrations of NP in fish among sampling sites (Figure 7). The greatest concentrations were found in fish from the section of the river near Kalamazoo and Battle Creek (Table 6). While these cities employ advanced wastewater treatment technologies, industrial discharges, either through the 43 municipal WWTP or direct effluent discharge may have contributed to greater concentrations of nonylphenolic compounds in these cities. Table 5. Detectable Concentrations of NP in Tissue Across Species. Species Mean tissue Tissue Number of concentration concentration Samples (nglg. ww) range (nglg.