The congener PCB fractions (Fl) were analyzed by GC/ECD on a 60-m DB-5 column; the concentra- 

 tions of 80 congeners were quantified. The mono-ortho PCB fractions (F2) were analyzed by GC-ECD 

 on a 3-m DB-1 phase column. 



For the non-ortho PCB fractions (F3), the analyses were done by capillary gas chromatography inter- 

 faced to high resolution mass spectrometry (GC/HRMS). Samples were injected by cool on-column 

 technique onto a retention gap connected to an Ultra- 1 (DB-1 equivalent) 50-m capillary column. The 

 MS system resolution was tuned to 10,000. Selected ion monitoring of two mass windows was done 

 for Cl 3 and Cl 4 biphenyls, and C1 5 -C1 6 biphenyls. 



The PCDD/PCDF fractions (F4) went through a final cleanup step on activated basic alumina to re- 

 move possible chlorinated co-contaminants. The fractions were then analyzed by capillary GC coupled 

 to HRMS. The column used was a 50-m Ultra-2 (Hewlett-Packard DB-5 equivalent) capillary column. 

 The MS system resolution was tuned to 10,000. Eighteen compounds were detected by selected ion 

 monitoring with five mass windows to measure Cli-Clg PCDDs and PCDFs. 



The H4IIE rat hepatoma cell bioassay was performed with extracts of the samples from the same 20 

 samples characterized in the chemical analyses. The induction of cytochrome P450 in the whole ex- 

 tract (Fl) was measured following methods of Tillitt et al. (1991). Also, the toxicity of six fractions of 

 the whole extract was determined in each sample: a PAH fraction (F5); a dioxin/furan fraction (F12); a 

 combined PCB fraction (Fl 1); and three planar/co-planar PCB fractions (F7, F8, F9). 



Data Analyses. Results of the toxicity tests performed with the amphipods and bivalve embryos were 

 arcsin-square root transformed and compared with the controls with one-tailed, unpaired, t-tests to 

 determine significant differences (n=5 replicates, alpha = 0.05). The tests were conducted in 10 batches, 

 the control sediment was tested along with the environmental samples in each batch, and the results 

 from each test of the control were used in the statistical analyses for each batch. To determine if the 

 mean percent survival at any sites (n=3) were significantly different from mean survival in controls, the 

 untransformed data were evaluated with one-tailed t-tests (alpha=0.05). 



The Microtox tm test data were analyzed using a linear interpolation technique to determine concentra- 

 tions of the extract that inhibited luminescence by 50%. This value (expressed as wL of extract per mL 

 of Microtox tm exposure volume) was then converted to mg/mL using the wet weight of sediment in the 

 original extract. To determine differences from controls, a pairwise comparison was made between 

 test samples and LIS controls, using analysis of covariance (ANCOVA). Both the concentrations and 

 response data were log-transformed prior to the analysis to linearize the data. The ANCOVA was first 

 used to determine if the two lines had equal slopes (alpha=0.05), and if they did, it was used to check 

 for equal Y-intercepts (alpha=0.05). To determine which sites were significantly different from con- 

 trols, the three EC50 values for each site were compared to the control values with a one-way t-test 

 (alpha=0.05). 



The relationships between measures of toxicity and the concentrations of physical-chemical variables 

 in the samples were determined in several steps. First, simple, non-parametric, Spearman-rank corre- 

 lations were performed (Statview 4.0 software). Where the correlations appeared to be significant, the 

 data were examined in bivariate scatterplots to confirm the distribution pattern. Next, to determine 

 which chemicals were most elevated in concentration in the toxic samples, the average concentrations 

 in both toxic and nontoxic samples were compared. Finally, to determine which, if any, toxicants were 

 sufficiently elevated in concentration to cause or contribute to toxicity, the average concentrations in 

 the toxic samples were compared with applicable, effects-based sediment guideline values. 



27 



