Environ. Sci. Technol. 2004, 38, 2574-2583
Sources and Dechlorination of Polychlorinated Biphenyl Congeners in the Sediments of Fox River, Wisconsin IPEK IMAMOGLU,* KAI LI,† ERIK R. CHRISTENSEN, AND JULIE K. MCMULLIN‡ Department of Civil Engineering and Mechanics and Center for Great Lakes Studies, University of WisconsinsMilwaukee, Milwaukee, Wisconsin 53211
Polychlorinated biphenyl (PCB) congeners were analyzed in eight deep, dated sediment cores collected from the immediate upstream and downstream of DePere dam in Fox River, Wisconsin. The average time span of the cores is about 100 yr, except for one core (FR-9) which is influenced by mixing or covers a short time period (2 yr). The total PCB concentrations have a range of 0.2-6.8 ppm for the upstream and 0.3-17.6 ppm for the downstream cores. The PCB data obtained from the sampling were analyzed as upstream and downstream data, using a factor analysis (FA) model with nonnegative constraints to identify PCB sources and congener patterns. The factor loadings obtained from the FA model were interpreted in terms of the presence of possible environmental degradation mechanisms. In addition, a recently developed model, which is used to identify and quantify possible pathways of anaerobic dechlorination of PCBs in the sediments, was validated on the basis of in situ data from the literature, and then applied to the congener patterns obtained from the FA model. The major PCB source to the Fox River sediments is identified as Aroclor 1242, for both the upstream and the downstream sediments. Loss of di- and trichlorobiphenyls (e.g., 2-4, 25-2) from the sediments suggests desorption from the sediments. On the other hand, observation of elevated amounts of certain congeners such as 24-3 and 25-3 indicates the presence of anaerobic dechlorination activity. The anaerobic dechlorination model demonstrates significant similarities between the Aroclor 1242 profile altered according to dechlorination processes Q (upstream) and H′ (downstream) and the dominant congener patterns obtained from the FA model.
Introduction Sediments of Fox River, Wisconsin, are known to be contaminated with polychlorinated biphenyls (PCBs) from a large number of paper mills located along the banks of the * Corresponding author present address: Department of Environmental Engineering, Middle East Technical University, Inonu Bulvari 06531, Ankara, Turkey; phone: +90 312 210 5861; fax: +90 312 210 1260; e-mail:
[email protected]. † Present address: Pharmaceutical Products Development, Inc., 8500 Research Way, Middleton, WI 53562. ‡ Present address: Brown and Caldwell, 250 E. Wisconsin Avenue, Suite 1525, Milwaukee, WI 53202. 2574
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 9, 2004
river (1-3). PCBs were used for carbonless copy paper manufacturing between 1954 and 1971, and due to severe contamination, the area was included in the National Priority List of the United States Environmental Protection Agency (1). The total PCB mass in Fox River is estimated to be 30000 kg, with more than 22000 kg stored downstream of DePere dam (4). Therefore, the PCB pollution particularly in this section of Fox River is likely to serve as a major long-term source of PCBs to Green Bay and Lake Michigan (2, 5). Despite their stability, PCBs have been shown to be affected by degradation mechanisms in the environment, including physicochemical (6-9) and biological (10-12) processes. In general, the more soluble and volatile congeners (lowly chlorinated PCBs) are likely to be preferentially removed from contaminated sediments and transported downstream (6, 13, 14), within the sediment depth (9, 15, 16), or into the atmosphere after solubilization (17, 18). The environmental implications of the two processes differ significantly from each other. While physicochemical weathering redistributes the congeners into different environmental compartments without eliminating the contamination, biological processes (particularly anaerobic dechlorination) result in the conversion of highly chlorinated congeners to lower chlorinated and possibly less toxic, more biodegradable ones (19). The knowledge and understanding of the prevailing environmental degradation mechanisms is thus important for making informed decisions concerning the fate and remediation of contaminated sediments. PCBs in Fox River sediments were analyzed by the Wisconsin Department of Natural Resources in 1990 as a part of the Green Bay Mass Balance Study (4). Congenerspecific PCB data from this study were confined largely to the top 0-35 cm layers. These data were used in our previous study to investigate Fox River sediment PCBs (20). Indications of anaerobic dechlorination of PCBs were detected; however, only a qualitative discussion was made. A model simulating anaerobic dechlorination was developed and previously applied successfully to PCBs in the sediments of Ashtabula River, Ohio (3). In that study, model validation was presented for literature laboratory data only. The goal of the present study is to further investigate environmental degradation processes of PCBs in sediments of Fox River, Wisconsin. PCBs were analyzed in the deep (0 to ∼2 m), dated sediment vibra cores collected from eight locations along the Fox River (Figure 1) in 1999. The major PCB sources and/or congener patterns affecting the sediments were determined by a factor analysis (FA) model with nonnegative constraints (20, 21). The validation of the recently developed anaerobic dechlorination model with in situ data from the literature, as well as the application of this model to Fox River congener patterns obtained from the FA model, is presented for the first time in this paper. Thus, environmental degradation of PCBs in the Fox River sediments is investigated using both the FA model and the newly developed anaerobic dechlorination model, and a quantitative account of the congener pattern changes observed in the sediments is made.
Materials and Methods Sediment Sampling and Dating. Eight vibra cores were collected from the Fox River in May 1999, with the help of the USEPA R/V Mudpuppy. Five cores were collected from downstream of DePere dam, and the rest from immediately upstream of the dam (Figure 1). The locations of the cores downstream of DePere dam were chosen so that they would 10.1021/es035165x CCC: $27.50
2004 American Chemical Society Published on Web 04/01/2004
concentrator to about 5 mL, and then further reduced by nitrogen to 1 mL. PCBs were analyzed on a Hewlett-Packard 5890 series II gas chromatograph equipped with an electron capture detector (ECD) and a DB-5 column (30 m × 0.32 mm × 0.25 µm, J & W Scientific Inc.). Helium was used as the carrier gas. The initial oven temperature was set at 50 °C, held for 2 min, then increased to 290 °C at a rate of 3.0 °C/ min, and kept at 290 °C for 10 min. The temperatures of the injector and detector were 250 and 325 °C, respectively. The injection volume was 1 µL in the splitless mode. PCBs were identified by matching the peak retention times relative to that of the internal standard to the relative retention times of known standards. The concentrations of individual compounds in the sample were calculated on the basis of the ratios of the peak areas to that of the internal standard (pentachloronitrobenzene), which was added before injection. Hewlett-Packard G1701BA MSD ChemStation software was used for this purpose.
FIGURE 1. Map of the Fox River sampling locations. be in the vicinity of the two major PCB point sources, Nicolet Paper Co. and Fort James Paper Co. The cores varied in length from 124.5 to 231 cm, and following collection, they were frozen at -18 °C for several days, and later sliced into 15 equal subsections. All sediment samples were dated using 210Pb and 137Cs (22). The details of the sediment dating procedures are given by Van Camp (23). Porosity and loss on ignition (organic content) were also determined for each section. Material for PCB analysis for each section was transferred into amber glass bottles and stored at -18 °C. Dissolved oxygen concentrations of the interstitial water were e2 mg/L below the upper 6 mm of the sediment, both upstream and downstream of DePere dam (23). PCB Analysis. Measurement of PCBs in sediments was performed on the basis of procedures established in our laboratory (24), and EPA guidelines SW846 (25) and 8082 (26), and summarized below. All PCB congeners, the surrogate standards (dibutylchlorendate + tetrachloro-m-xylene), and the internal standard (pentachloronitrobenzene) were purchased from AccuStandard Inc. (New Haven, CT). Solvents used in analysis were obtained from Fisher Scientific (Chicago, IL) and were ACS grade or higher. Samples stored at -18 °C were thawed at room temperature and thoroughly mixed. A portion of each sample was withdrawn and freezedried overnight (-45 °C, 100 µmHg). The sample was then homogenized by grinding and stored in a desiccator at room temperature. An aliquot of 5 g of the dried sediment was placed in the Soxhlet thimble along with surrogate standards and extracted in the Soxhlet apparatus using 150 mL of hexane/acetone (1:1 v/v) for 24 h. The volume of the extract was reduced in a Kuderna-Danish (K-D) concentrator (500 mL flask, three-ball Snyder column, 10 mL concentrator tube) to about 5 mL, and the solvent was exchanged by addition of about 50 mL of hexane. The volume was reduced again to about 5 mL in the same K-D apparatus. Then a gentle stream of nitrogen was used to bring the volume of the extract down to about 2 mL. The cleanup of the sample extracts was accomplished by passing the extracts through a silica gel chromatographic column (11 × 300 mm) which was packed with HCl-rinsed copper at the bottom to absorb elemental sulfur, 10 g of 1% H2O deactivated silica gel, and about 5 g of anhydrous sodium sulfate on top to absorb residual water. The extract was eluted with 50 mL of hexane to recover the PCB fraction. The volume of solvent was reduced by the K-D
Among the 31 PCB congeners used in the creation of standard curves, 26 peaks could be resolved. These were IUPAC Nos. 1, 4/10, 8, 18, 19, 22, 25, 26, 27, 28/31, 33/53, 47, 52, 66/95, 70, 92, 101, 110, 118, 138, 139/149, 153, 180, 187, 201, and 203. Total PCBs refer to the sum of the concentrations of these congeners. The PCB standard stock solution was prepared by mixing 31 individual PCB congeners, and diluting with hexane to a concentration of 1000 ppb for each congener. A five-level calibration curve was then generated using standards with concentrations ranging from 1000 to 5 ppb. The average correlation coefficient (R2) was 0.993. To ensure the reliability of the analysis, tests on method blanks, a standard reference sample (NIST 1941a, Organics in Marine Sediment), sample replicates, and matrix spike replicates were performed prior to analysis of real sediment samples. The average recovery of the surrogate standard was 91.7%, ranging from 49.1% to 132.0%. The measured PCB concentrations in the sediment cores were corrected by surrogate recovery from individual samples. The certified marine sediment sample (NIST 1941a) was analyzed with the same procedures, and the average recovery was 95.2% ( 9.2%, ranging from 50.9% to 146.0%. The effect of freeze-drying on the recovery of PCBs was also evaluated using the NIST 1941a sediment. An average recovery of 93.5% was achieved after re-freeze-drying, on the basis of the nine congeners (IUPAC Nos. 52, 66, 95, 101, 110, 118, 138, 149, and 153) whose certified values are available from the certificate of analysis. Congeners 8 and 18 were not included. Although there is a possibility that loss of less chlorinated congeners such as 8 and 18 occurs during freeze-drying and multiple concentration steps due to their volatility, our analytical results indicated that such a loss was not significant. The average recovery from sediment spiked with 31 congeners including congeners 8 and 18 was 94.6% for direct extraction. The recoveries of congeners 8 and 18 were satisfactory (>80%). If the spiked sediment was resuspended in water and re-freeze-dried prior to extraction, the average recovery was 91.4%. This is comparable to the average recovery of 93.5% from NIST 1941a undergoing similar treatment. It was, therefore, concluded that freezedrying did not cause a significant loss of PCBs and the distribution of 31 congeners was not changed. From here onward, the congener designation used by Bedard and May (27) will be adopted, where the position of the chlorine atoms on each ring of biphenyl is given by numbers, and a hyphen represents separation of the rings. FA Model with Nonnegative Constraints. The FA model used in this study is described in detail by Imamoglu and Christensen (20), and Imamoglu (28), and briefly summarized here. The fundamental equation underlying the principal VOL. 38, NO. 9, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2575
component analysis based model is
D ) CR
(1)
(m × r) ) (m × n)(n × r) The data matrix D is factored into its components, the factor loading matrix C and the factor score matrix R, where the former represents source compositions and the latter represents source contributions. Additionally, m, n, and r are the number of congeners, sources, and samples, respectively. Among the 26 groups of congeners analyzed, 2-2/26- was never detected in any of the samples, and was not included in the analysis. Congener 2-, which has a very low response factor in the gas chromatograph detector, was only detected in some samples from cores FR-1 and FR-2. This congener may not be accurately quantified by the chlorine-sensitive ECD (29) and hence not included in the analysis due to the possibility of including an erroneous variable. Subsequently, a total of 24 variables were used. Missing values in the raw data set were replaced by the detection limits (29). Samples that had nondetects in more than three congener groups (peaks), i.e., approximately >10% of the variables, were excluded from the data matrix. The use of PCBs in the Fox River area started around 1954 (30); therefore, many of the deep, old samples had very low PCB content. This resulted in many nondetects, and overall, 36 samples (12 from the upstream and 24 from the downstream data set) were eliminated. The PCB data from Fox River were investigated in two parts, upstream of DePere dam (referred to as “upstream Fox River”) and downstream of DePere dam (referred to as “downstream Fox River”). The dimensions of the resulting data matrixes were 24 variables as rows and 33 samples as columns for the upstream and 24 variables as rows and 51 samples as columns for the downstream Fox River data set. Prior to factor analysis, the units of the data matrix were converted from nanograms per gram to nanomoles per gram. The data matrix was scaled by dividing the concentrations of congeners in each layer by their respective overall average concentrations for all layers (in all cores). Factor analysis, including diagnostics tests and backscaling, was then performed as by Imamoglu et al. (3). Anaerobic Dechlorination Model. A new model to be used in combination with the FA model was developed to identify possible patterns of anaerobic dechlorination of PCBs in the sediments, and to quantify the relevant dechlorination pathways. The model is described in detail by Imamoglu (28) and Imamoglu et al. (3), and briefly summarized here. This model uses a least-squares method to alter an original Aroclor profile according to six major dechlorination processes (M, Q, H, H′, P, N) identified in the literature (12, 19). The term “dechlorination process” as used here refers to a particular set of dechlorination reactions that indicates which congeners are substrates and which chlorines from indicated positions are removed from each (12). For example, in dechlorination process N, chlorines from flanked meta positions (i.e., where another chlorine is present in an adjacent position) are removed from penta- to nonachlorobiphenyls, whereas for process H, flanked p-chlorines and the m-chlorine from the -234 position are removed from tetra- to heptachlorobiphenyls (12, 19, 27). No detailed order for the removal of chlorines during anaerobic dechlorination is specified in the literature. The aim of the anaerobic dechlorination model is to alter an original Aroclor profile according to a specific dechlorination process to obtain a congener profile similar to that of a given sample (or congener pattern). If the similarity between the resulting altered Aroclor profile and the sample 2576
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 9, 2004
profile is improved significantly, we may conclude that the PCBs in the sample have undergone anaerobic dechlorination. The goodness of fit of the altered Aroclor profile to a given congener pattern was evaluated by the multiple coefficient of determination R2, the average mole per thousand error σ, and the percent improvement in the sum of squares of differences S between the given congener pattern and the altered Aroclor profile (28). The value of S was calculated from m
S)
∑(yˆ - x ) j
j
2
(2)
j)1
where yˆj is the concentration of altered congener j (mol ‰), xj is the concentration of sample congener j (mol ‰), and m is the number of marker congeners. Marker congeners are chosen to simplify the calculations. Congeners that are most abundant and/or important for the dechlorination may be selected as markers. When a complete congener analysis is not available (such as the case for Fox River samples), only the measured congeners are used as markers and included in the evaluation of the sums. However, all known congeners in the Aroclor profile are considered during simulation of dechlorination reactions, as long as at least either the product or the reactant PCB is a marker congener. If a complete congener analysis is available, then all can be regarded as markers, which would lower the uncertainty of quantification of the pathways. However, then all the possible dechlorination pathways applicable to each congener should be specified for the model. Validation of the Dechlorination Model. The validation of the anaerobic dechlorination model was performed using both laboratory and in situ anaerobic dechlorination data from the literature (28). The model validation using laboratory data is given by Imamoglu et al. (3), and the validation based on sediment PCB data obtained from an in situ PCB dechlorination study in Woods Pond, Massachusetts (27), is presented here. In Bedard and May’s (27) study, a quantitative mass balance by looking at the accumulation of key daughter and the depletion of key mother congeners (17 groups of congeners, due to coelution) relative to Aroclor 1260 is presented. The same key congeners used in the quantification are selected here as marker congeners in model validation, as well as the same Aroclor 1260 profile, so that a fair comparison can be made with the resulting altered Aroclor profiles. Consequently, two sample profiles, one representing a relatively less altered profile, A-35-2, and another which is a substantially altered profile, A-34-1, are used, together with the Aroclor 1260 profile as given in that study (27), as input for the model. Both single and combinations of dechlorination processes N, H, and P (applicable for highly chlorinated PCBs) are used in the model (12, 19). Even though a small contribution from Aroclor 1254 is noted for Woods Pond sediments, the quantification for the key congeners is carried out considering only Aroclor 1260 by Bedard and May (27), and hence only Aroclor 1260 is taken as the source in model validation calculations here.
Results and Discussion Sediment Dating and PCB Congener Profiles. Plots of 137Cs activity versus depth for all the cores are shown in Figure A, Supporting Information. This radionuclide comes from aboveground nuclear testing between 1954 and 1964, with a maximum in 1963. The peak, alone or in conjunction with those for 210Pb activities (22), is used to determine sedimentation rates. The plots for cores FR-2, FR-6, FR-8, and FR-10 (Supporting Information, Figure A) show definite peaks,
FIGURE 2. (A) Average congener profile with 1 standard deviation of the mean of upstream and downstream Fox River samples. (B) Congener profiles of major Aroclors (31). whereas the peaks for FR-1 and FR-5 are too close to the sediment-water interface, indicating a slow sedimentation rate, and FR-9 does not show a peak at all, which indicates mixing or a very high sedimentation rate. FR-3, on the other hand, shows a peak that extends over nearly 20 cm. This may be due to a change in the sedimentation pattern in connection with river flow alterations caused by construction or modification of the DePere dam. Sedimentation rates for cores FR-2, -6, and -8 were determined from 137Cs. For cores FR-3, -5, and -10, both 137Cs and 210Pb were used, while 210Pb alone was the basis of the sedimentation rate for core FR-1. Among the upstream Fox River sediments, FR-1 has the lowest (0.99 cm/yr) and FR-2 has the highest (1.89 cm/yr) sedimentation rate. For the cores collected downstream of DePere dam, FR-9 has possibly the highest sedimentation rate, showing a time span of only 2 yr, if not influenced by mixing. The other downstream cores have sedimentation rates between 0.36 cm/yr (FR-5) and 2.00 cm/yr (FR-10). The fluvial characteristics and the PCB sources of the upstream and downstream samples are expected to be different; therefore, they were handled separately in this study. The average congener profiles for the two sections, together with 1 standard deviation of their means, are given in Figure 2. The congener profiles of the four major Aroclors (31) are also included in Figure 2. Overall, upstream and downstream average congener profiles do not seem to show significant differences from each other. When compared to the possible source profiles, the sample average congener profiles resemble mostly the Aroclor 1242 profile, with slight deviations. This supports the fact that Aroclor 1242 was the main PCB emulsion used in carbonless copy paper industries along Fox River (30). Downstream samples exhibit higher total PCB concentrations (0.3-17.6 ppm) when compared to upstream samples (0.2-6.8 ppm); see Figures 3 and 4. FR-8 and FR-10 show the highest total PCB concentrations, as expected since FR-8 is located very close to one of the pollution sources, Fort James
Paper Co., and FR-10 is downstream from both PCB sources (Figure 1). FR-9 is also downstream from both of the major pollution sources; however, mixing or a high sedimentation rate prevents the observation of a PCB peak. The total PCB peaks (∼1970) are slightly higher (younger) than the 137Cs peaks (∼1963) for most of the cores. PCB Congener Patterns from Factor Analysis. The diagnostic tools used to determine the number of significant factors indicate two factors to be adequate for the upstream data set, and three for the downstream set (28). For the upstream Fox River data, 94% of the variance is explained by the first two factors, and congener-specific goodness of fit criteria (coefficient of determination, percent variance by congener) show no significant improvement for any of the congeners with the addition of a third factor. For the downstream Fox River data, a satisfactory fit for all the variables was obtained by three factors. An overall good fit is indicated by the value of the Exner function (0.32), and by the fact that 92% of the variance in the data set was explained with three sources. The factor loading and score plots obtained from the FA model for the upstream Fox River sediment samples, together with the total PCB profiles, are shown in Figure 3. The major congener pattern for upstream sediments, pattern U1, is one that is very similar to the profile of Aroclor 1242, with relatively less abundance of 2-4 and 25-2. Pattern U2, on the other hand, does not resemble any of the Aroclor profiles closely. This profile is dominated by the extremely high abundance of 25-2, and also 25-25. Congeners 24-3, 25-3, and 26-3, which are normally found in very low concentrations in Aroclors (Figure 2), are also elevated in this profile. In contrast, all highly chlorinated congeners (beyond 25-34) are in very low proportions. The factor scores indicate that samples with high total PCBs mostly exhibit pattern U1, whereas an increasing trend with depth is observed for the contribution of pattern U2 to all (especially FR-1 and FR-3) sediment cores. VOL. 38, NO. 9, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2577
FIGURE 3. (A) Factor loading plot and (B) factor score plot for upstream Fox River samples together with total PCB profiles.
FIGURE 4. (A) Factor loading plot and (B) factor score plot for downstream Fox River samples together with total PCB profiles. The three factor loadings and scores obtained from the FA model for the downstream Fox River samples are 2578
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 9, 2004
presented in Figure 4, together with the total PCB profiles. All three factor loadings (patterns D1, D2, and D3) resemble
FIGURE 5. (A) Original Aroclor 1260 profile with profiles of samples A-34-1 and A-35-2, from Bedard and May (27). Congener profiles of altered Aroclor 1260 from the anaerobic dechlorination model with 1 standard deviation and the altered Aroclor 1260 profile as calculated by Bedard and May (27), in comparison to (B) the profile of sample A-34-1 and (C) the profile of sample A-35-2. the profile of Aroclor 1242 (Figure 2) to some extent, but have lower abundances of congeners 2-4 and 25-2, especially for factors 2 and 3. Pattern D1 is characterized by elevated amounts of congeners 24-3, 25-3, and 26-3, and diminished amounts of 34-2/25-26, 24-34/236-25, and 25-34, as compared to the profile of Aroclor 1242. The factor scores indicate contribution of this pattern to a large number of the downstream samples. Pattern D2 is similar to the first pattern, yet it is characterized by a low abundance of 2-4 and 25-2. This pattern seems to be exhibited exclusively by samples from FR-10, the extreme downstream point of all sampling sites (Figure 1). In pattern D3, the presence of 2346-24/236-245 indicates traces of Aroclor 1254 and/or 1260 (31). Two highly concentrated samples from core FR-8 as well as cores FR-5 and FR-6 have
high scores for this pattern (Figure 4B). Aroclor 1254 was detected in the discharges of Nicolet Paper Co. and DePere publicly owned treatment works (30) (Figure 1), which may explain the occurrence of these pentachlorobiphenyls. Application of the Dechlorination Model to Woods Pond Sediments. The original Aroclor 1260 and the two sample congener profiles are given in Figure 5A, and the altered Aroclor 1260 profiles obtained from the model together with the respective sample profiles are shown in Figure 5B,C. Also shown in these plots are the altered Aroclor 1260 profiles, calculated from the quantitative mother-daughter analysis by Bedard and May (27). The first 17 congener groups (from 25-25 to 2345-234) are the marker congeners (the key congeners in Bedard and May’s (27) study) selected in the calculation of altered Aroclor 1260 profiles. Both plots show VOL. 38, NO. 9, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2579
TABLE 1. Quantification of the Most Probable Anaerobic Dechlorination Pathways for Two Samples from Woods Pond (27) for Anaerobic Dechlorination Model Validation sample A-34-1 conversion (mol ‰)
sample A-35-2 conversion (mol ‰)
congener dechlorination pathwaya,b
avgc
SDd
avgc
SDd
153 (245-245) f 99 (245-24) 99 (245-24) f 47 (24-24) 101 (245-25) f 49 (24-25) 138 (234-245) f 99 (245-24) 101 (245-25) f 52 (25-25) 85 (234-24) f 47 (24-24) 153 (245-245) f 101 (245-25) 180 (2345-245) f 141 (2345-25) 84 (236-23) f 46 (23-26) 179 (2356-236) f 152 (2356-26) 138 (234-245) f 85 (234-24) 141 (2345-25) f 101 (245-25) 87 (234-25) f 49 (24-25) 180 (2345-245) f 137 (2345-24) 97 (245-23) f 42 (23-24) 180 (2345-245) f 146 (235-245) 128 (234-234) f 85 (234-24) 172 (2345-235) f 130 (234-235) 146 (235-245) f 90 (235-24) 137 (2345-24) f 85 (234-24) 137 (2345-24) f 99 (245-24) 170 (2345-234) f 130 (234-235) 130 (234-235) f 83 (23-235) 59 (236-3) f 27 (26-3) 87 (234-25) f 44 (23-25) 170 (2345-234) f 137 (2345-24) 146 (235-245) f 92 (235-25)
68.2 65.0 33.2 28.0 24.8 24.7 20.5 19.3 18.2 17.0 17.0 15.7 15.4 14.9 14.8 11.5 10.8 8.20 7.31 7.17 6.95 6.64 6.50 4.48 3.85 3.83 3.78
21.5 9.85 14.8 25.5 2.12 11.0 16.7 16.4 10.6 9.68 11.0 15.1 12.2 10.6 5.84 7.14 4.59 5.74 6.47 6.28 6.70 5.43 3.73 4.27 2.85 3.72 2.83
22.3 7.8 7.8 15.0 17.7 8.1 11.4
10.23 4.60 5.35 13.38 1.10 4.67 9.52
7.6 9.4
5.49 8.09
17.1 8.9
13.54 6.09
6.5 14.4 7.3 7.4
4.72 6.51 3.74 6.01
x x
5.6
4.72
x x
ref 27 dechlorination process x x x x x x x x x x
x
x x
N N N N, H P, H N, H P, H (P), H N N N N N, H N N (P), H N, H N N N N (P), H P (N), M P N, H P, H
a
The chlorine removal takes place from the italic positions (up to two chlorines). A total of 64 dechlorination pathways exist; only the major ones are listed. b Marker congeners are indicated by boldface type. c Average of the 100 conversion values obtained after 100 shuffles of the sequence of pathways. d Standard deviation of the 100 conversion values obtained after shuffling of the sequence of pathways.
that the anaerobic dechlorination model is successful in obtaining a congener profile that is close to that of the sample profiles, for both marker and nonmarker congeners. The anaerobic dechlorination model provides scenarios of dechlorination and quantifies the reactions or pathways where the known original Aroclor profile is altered so that a profile similar to an observed congener profile can be produced. However, the quantification of the dechlorination associated with the pathways can depend on their sequence, and adequate information regarding the priority of one pathway over another is not available in the literature. Therefore, in this study, 100 random shuffles of the sequence of pathways are performed. At each new sequence, the dechlorination yield for each pathway is calculated. This calculation for a given pathway is carried out to an extent that will minimize the sum of squares of differences between measured and calculated amounts of marker compounds (eq 2). The measured value here is the sample profile (i.e., A-34-1, Figure 5) or the factor loading (pattern D1, Figure 4), and the calculated value is the modified Aroclor profile. When the dechlorination yields for all the pathways are calculated, the calculations are repeated for the same sequence four more times, since congeners can appear in more than one pathway. Estimated yields in a previous run are used as start values in a given run. This was found to be sufficient to create stable dechlorination yields. Then, the sequence of pathways is shuffled, and the calculations are carried out starting with the original sample and Aroclor profiles. When 100 shuffles are carried out, the average and standard deviation of the 100 dechlorination yields obtained for each pathway are presented. Dechlorination along pathways with standard deviations larger than their averages may, therefore, be relatively unlikely to take place, unless certain favorable sequences occur. The shuffling is carried out not to mimic 2580
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 9, 2004
anaerobic dechlorination necessarily, but to provide a probabilistic approach to the prioritization of pathways. The combined activities of N and P give the best fit of the altered Aroclor 1260 profiles to the sample profiles from Woods Pond sediments (27). A total of 64 possible pathways (from N and P) are combined to obtain altered Aroclor 1260 profiles; therefore, a variety of alternatives exist for the dechlorination of a given congener. And since 17 congener groups are used as markers, the nonmarker congeners tend to have higher standard deviations, because there is less information available about them in the program. This is apparent from the standard deviation bars in Figure 5B,C for congeners beyond 2345-234 (i.e., nonmarker congeners). The quantification of the most probable dechlorination pathways acting on the two samples analyzed here is given in Table 1. Bedard and May (27) proposed the major pathways of dechlorination for hexa- and heptachlorobiphenyls with no quantification; these are indicated by check marks in column six of Table 1. Dechlorination processes in parentheses would only be classified as the indicated type if the definitions of ref 32 were expanded, for example, to include heptasubstrates for process P. The fact that we identified the same dechlorination processes (N and P) as those proposed by Bedard and May (27), and find many possible pathways, demonstrates that the anaerobic dechlorination model can identify the important dechlorination processes and provide quantitative information on their occurrence. Accordingly, the model should be able to provide information on the possibility, type, e.g., H′, Q, etc., and extent of anaerobic dechlorination in a given environmental sample. From Table 1 we also see that dechlorination may not follow a strict sequence, e.g., first doubly flanked meta, followed by singly flanked meta, followed by unflanked meta on a di- or trisubstituted ring, etc., as assumed by Klasson and Just (33), but rather occurs according to a probabilistic
TABLE 2. Quantification of the Anaerobic Dechlorination Pathways for Pattern U2 from Upstream and Pattern D1 from Downstream Fox River Samples, As Obtained from the Model pattern U2 conversion (mol ‰) congener dechlorination pathwaya,b 8 (2-4) f 1 (2-) 28 (24-4) f 8 (2-4) 22 (23-4) f 8 (2-4) 44 (23-25) f 18 (25-2) 48 (245-2) f 18 (25-2) 49 (24-25) f 18 (2-25) 45 (236-2) f 19 (26-2) 46 (23-26) f 19 (26-2) 84 (236-23) f 19 (26-2) 22 (23-4) f 1 (2-) 22 (23-4) f 5 (23-) 60 (234-4) f 22 (23-4) 25 (24-3) f 6 (2-3) 55 (234-3) f 25 (24-3) 66 (24-34) f 25 (24-3) 67 (245-3) f 26 (25-3) 70 (25-34) f 26 (25-3) 59 (236-3) f 27 (26-3) 71 (26-34) f 27 (26-3) 28 (24-4) f 1 (2-) 28 (24-4) f 7 (24-) 60 (234-4) f 28 (24-4) 31 (25-4) f 9 (25-) 74 (245-4) f 31 (25-4) 33 (34-2) f 6 (2-3) 56 (23-34) f 33 (34-2) 66 (24-34) f 33 (34-2) 85 (234-24) f 47 (24-24) 101 (245-25) f 52 (25-25) 66 (24-34) f 6 (2-3) 105 (234-34) f 66 (24-34) 118 (245-34) f 70 (25-34)
avgc
SDd
74.3 1.8 0.70 18.6 3.00 7.7 0 0
4.14 4.31 1.22 9.65 3.62 8.47 0 0
6.6 2.7 3.6 2.9 0.2 14.6 0 15.9 3.4 9.5 22.2 19.7 0.90 29.1 0.80 12.3
6.34 2.98 3.53 1.87 0.29 2.18 0 0.17 0 0 26.09 24.74 2.10 25.59 1.86 3.08
1.7
2.87
10.6
pattern D1 conversion (mol ‰) avgc
SDd
0 0 0
0 0 0
1.80 0.98 0.61
1.55 1.20 0.65
0.82 17.5 1.36 21.0 3.37 9.46
0 0 0 0 0 0
8.61
2.06
14.5 2.03 1.65
2.07 1.41 1.39
2.93 5.59
0 0
4.23 3.36
0 0
5.62
dechlorination process Q Q Q, H′ Q, H′ Q, H′, H Q (Q, H′), N Q, H′ (H′), M, N Q Q (Q), P Q Q, H′, H Q, H′, H Q, H′, H Q, H′, H (Q, H′), M Q, H′, H Q Q Q, H′, H Q Q, H′, H Q, H′ H′ Q H′, H H′, H Q H′, H H′, H
a The chlorine removal takes place from the italic positions (up to two chlorines). b Marker congeners are indicated by boldface type. c Average of the 100 conversion values obtained after 100 shuffles of the sequence of pathways. d Standard deviation of the 100 conversion values obtained after shuffling of the sequence of pathways.
model where dechlorination from various chlorine positions occurs simultaneously and depends on other factors such as the symmetry of the product structure. For example, both 99 (245-24) f 47 (24-24) and 85 (234-24) f 47 (24-24) are more likely than para dechlorination, while 101 (245-25) f 52 (2525) may take precedence over 101 (245-25) f 49 (24-25) (for sample A-35-2). Also, apparently unrelated to symmetry considerations, 153 (245-245) f 99 (245-24) has a larger yield than 153 (245-245) f 101 (245-25), indicating a preference for singly flanked meta over singly flanked para dechlorination. However, the same does not hold true between 101 (245-25) f 49 (24-25) and 101 (245-25) f 52 (25-25) with sample A-35-2. The results mostly agree with literature information that doubly flanked chlorines are more susceptible to dechlorination (12), yet no specific preference could be observed for doubly flanked chlorine removal from para over meta positions. The anaerobic dechlorination model is able to quantify dechlorination pathways while maintaining overall congener balance, and also providing a fit to observed congener profiles. Note that this fit generally only can be obtained for certain types of dechlorination activities, such as N and P, for the Woods Pond sediments. Previously, pathways were proposed, but only sums of losses and gains for key (marker) congeners were shown to balance without quantification of individual processes (27). Dechlorination of Upstream Fox River Sediments. The first congener pattern obtained from the FA model (Figure 3) for the upstream sediments shows signs of physicochemical removal of the least hydrophobic di and trichlorobiphenyls 2-4 and 25-2 (34), possibly via desorption and solubilization
(6, 13). Loss of these congeners during analytical procedures is unlikely since satisfactory recoveries (>80%) are found for extraction and freeze-drying. The second congener pattern, which has an unusual congener profile unlike that of any Aroclor, is mostly exhibited by deep sediments marking years predating PCB use in the tributary. This suggests downward migration of congeners and/or physical mixing of sediments (9). Downward migration of PCBs (9) may be enhanced by microbial activity (16) and mostly affects congeners with low hydrophobicity, with the potential to result in a congener pattern such as the one found here. Congeners that were transported from upstream in the water column could also have been introduced into these sediment layers through interstitial water. The presence of suspected dechlorination products 24-3 and 25-3 (35) may indicate mild dechlorination activity (36) following physical mixing. Studies carried out further upstream of DePere dam point to the presence of anaerobic dechlorination activity in highly contaminated (total PCB concentrations >50 ppm) sediments (37). The quantification of the dechlorination pathways (Table 2) shows that the pathways with the highest dechlorination yields are 8 (2-4) f 1 (2-), 31 (25-4) f 9 (25-), and 28 (24-4) f 1 (2-). Since none of the product congeners for these pathways are measured in the sediments, and these lowly chlorinated congeners are susceptible to weathering processes in the environment, it is difficult to differentiate the impact of different environmental degradation mechanisms on the PCB contamination. However, regardless of other degradation mechanisms, application of the anaerobic dechlorination model to Aroclor 1242 with pattern U2 as a target shows that VOL. 38, NO. 9, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2581
FIGURE 6. Congener profiles of pattern U2 from upstream Fox River samples and Aroclor 1242 (A) before and (B) after the application of the anaerobic dechlorination model (process Q). Congener profiles of pattern D1 from downstream Fox River samples and Aroclor 1242 (C) before and (D) after the application of the anaerobic dechlorination model (process H′).
TABLE 3. Improvement in the Goodness of Fit Measures between the Factor Loading and Degraded Aroclor 1242 Profiles with Application of the Anaerobic Dechlorination Model
parametera
R2 error,σ (mol ‰) improvement in S (%)
initial final initial final
upstreamb
downstreamb
pattern U2
pattern pattern pattern D1 D2 D3
0.51 0.96 44 13 92
0.90 0.97 19 10 73
0.75 0.84 28 23 33
0.74 0.84 33 26 40
a
R 2 is the multiple coefficient of determination, σ is the average mole per thousand error between the two profiles, and S is the sum of squares of the differences between the two profiles. Initial and final refer to before and after simulation of anaerobic dechlorination. b The dechlorination activity in the upstream sediments is identified as process Q, whereas it is process H′ in downstream sediments.
it is possible to explain a significant portion of the degradation by process Q (Table 3, Figure 6A,B). Dechlorination of Downstream Fox River Sediments. The first congener pattern (pattern D1) obtained by the FA model for downstream Fox River sediments (Figure 4A) seems to be the one that is least affected by physicochemical weathering, and it contributes to most of the downstream samples (Figure 4B). The Aroclor 1242 profile (31) was altered by applying processes P, H, H′, N, M, and Q (12, 19) using the anaerobic dechlorination model to simulate the profile depicted by pattern D1 (Figure 4A). Application of the reactions of process H′ to Aroclor 1242 resulted in the best fit to pattern D1 (Figure 6C,D). Note that if process Q is allowed instead, one can also obtain a good 2582
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 9, 2004
fit of the altered Aroclor profile to pattern D1. However, the model results show that the gain of congener 24-4 from 234-4 f 24-4 is roughly balanced by its loss through 24-4 f 2-, 24-4 f 24-, and 24-4 f 2-4. This is contrary to laboratory observations of process Q (38). If process Q is present, there should be clear evidence of loss of unflanked p-chlorines through the aforementioned reactions or through 25-4 f 25-. If process Q were implied, rather than process H′, one would obtain a somewhat better fit of congener 2-4 by loss of an unflanked p-chlorine through 2-4 f 2-. However, as mentioned earlier, congener 2-4 is likely to be affected by desorption from the sediments. The similarity between the resulting altered Aroclor 1242 profile and pattern D1 suggests the presence of in situ anaerobic dechlorination. The congeners in both profiles are normalized to 1000 nmol. Although the two profiles shown in Figure 6C are similar to begin with, some detectable differences exist between the two. The resulting altered Aroclor 1242 profile (Figure 6D) shows a remarkable correspondence with the pattern. The R 2 value increases from 0.90 to 0.97, by including dechlorination, which indicates a very close match between the two profiles (Table 3). We conclude, therefore, that the altered Aroclor profile of pattern D1 is generated by anaerobic dechlorination according to process H′. A quantification of the dechlorination reactions responsible for these changes is given in Table 2. Note that reactions involving m-Cl removal from -236 at one time was considered to be possible under H′, but now are classified as either M or N (32). Since the associated yields are small, this distinction does not seem to be important here. The low standard deviations in the fifth column indicate that all the reactions listed in Table 2 are likely to occur regardless of their
sequence. Diminished amounts of 25-34 and 24-34 in pattern D1 are explained by the corresponding production of 25-3 and 24-3, respectively. Low proportions of congeners 2-4 and 25-2 cannot be explained by dechlorination process H′, indicating their loss due to weathering processes (Figure 6C,D). Patterns D2 and D3 are characterized by low abundances of congeners 2-4 and 25-2. The application of the anaerobic dechlorination model to these patterns indicates the presence of mild anaerobic dechlorination activity. However, the similarity of the altered Aroclor 1242 profile to patterns D2 and D3 is less than that of pattern D1 as evidenced, for example, by the final mole per thousand errors (23 and 26 mol ‰ vs 10 mol ‰).
Acknowledgments This research was supported by U.S. National Science Foundation Grants BES-9725068 and BES-0107402.
Supporting Information Available 137Cs activity versus depth figures. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) U.S. EPA State, Tribal, and Site Identification Center. NPL Site Narrative at Listing: Fox River NRDA/PCB Releases Green Bay, Wisconsin. Web site http://www.epa.gov/superfund/, 1998. (2) Manchester-Neesvig, J. B.; Andren, A. W.; Edgington, D. N. J. Great Lakes Res. 1996, 22, 444-462. (3) Imamoglu, I.; Li, K.; Christensen, E. R. Environ. Toxicol. Chem. 2002, 21, 2283-2291. (4) Velleux, M.; Endicott, D. J. Great Lakes Res. 1994, 20, 416-434. (5) U.S. EPA. Moving Mud: Remediating Great Lakes Sediments. A Report on the Sediment Assessment and Remediation Program in the Great Lakes Basin. Web site http://www.epa.gov/glnpo/ sediment/movemud/mud.html, 2000. (6) Sanders, G.; Hamilton-Taylor, J.; Jones, K. C. Environ. Sci. Technol. 1996, 30, 2958-2966. (7) Muir, D. C. G.; Omelchenko, A.; Grift, N. P.; Savoie, D. A.; Lockhart, W. L.; Wilkinson, P.; Brunskill, G. J. Environ. Sci. Technol. 1996, 30, 3609-3617. (8) Eisenreich, S. J. The chemical limnology of nonpolar organic contaminants: polychlorinated biphenyls in Lake Superior. In Sources and Fates of Aquatic Pollutants; Eisenreich, S. J., Ed.; American Chemical Society: Washington, DC, 1987. (9) Gevao, B.; Hamilton-Taylor, J.; Murdoch, C.; Jones, K. C.; Kelly, M.; Tabner, B. J. Environ. Sci. Technol. 1997, 31, 3274-3280. (10) Abramowicz, D. A. Crit. Rev. Biotechnol. 1990, 10, 241-251. (11) Brown, J. F. J.; Bedard, D. L.; Brennan, M. J.; Carnahan, J. C.; Feng, H.; Wagner, R. E. Science 1987, 238, 709-712. (12) Bedard, D. L.; Quensen, J. F., III. Microbial reductive dechlorination of polychlorinated biphenyls. In Microbial transformation and degradation of toxic organic chemicals; Cerniglia, C. E., Ed.; Wiley-Liss Inc.: New York, 1995; pp 127-216. (13) Bergen, B. J.; Rahn, K. A.; Nelson, W. G. Environ. Sci. Technol. 1998, 32, 3496-3501. (14) Willman, E. J.; Manchester-Neesvig, J. B.; Armstrong, D. E. Environ. Sci. Technol. 1997, 31, 3712-3718. (15) Sanders, G.; Jones, K. C.; Hamilton-Taylor, J.; Dorr, H. Environ. Sci. Technol. 1992, 26, 1815-1821. (16) Elder, J. F.; James, R. V.; Steuer, J. J. J. Great Lakes Res. 1996, 22, 697-706. (17) Chiarenzelli, J. R.; Scrudato, R. J.; Wunderlich, M. L. Environ. Sci. Technol. 1997, 31, 597-602. (18) Achman, D. R.; Hornbuckle, K. C.; Eisenreich, S. J. Environ. Sci. Technol. 1993, 27, 75-87.
(19) Bedard, D. L. Microbial dechlorination of PCBs in aquatic sediments. In Recent Advances in the Environmental Toxicology and Health Effects of PCBs; Robertson, L. W., Ed.; University Press of Kentucky: Lexington, KY, 2001. (20) Imamoglu, I.; Christensen, E. R. Water Res. 2002, 36, 34493462. (21) Rachdawong, P.; Christensen, E. R. Environ. Sci. Technol. 1997, 31, 2686-2691. (22) McMullin, J. K. M.S. Thesis, Department of Civil Engineering and Mechanics, University of WisconsinsMilwaukee, Milwaukee, WI, 2000. (23) Van Camp, R. P. M.S. Thesis, Department of Civil Engineering and Mechanics, University of WisconsinsMilwaukee, Milwaukee, WI, 1999. (24) Li, A.; Ab Razak, I.; Christensen, E. R. Toxic organic contaminants in the sediments of the Milwaukee Harbor Estuary, Phase III, Kinnickinnic River sediments; Final Report to the U.S. Army Corps of Engineers; University of WisconsinsMilwaukee: Milwaukee, WI, 1995. (25) USEPA Test Methods for Evaluating Solid Waste, Physical/ Chemical Methods, SW846, Revision 2; Office of Solid Waste and Emergency Response: Washington, DC, 1994. (26) USEPA Test Methods for Evaluating Solid Waste, Physical/ Chemical Methods, SW846, Method 8082 Revision 0; Office of Solid Waste and Emergency Response: Washington, DC, 1996. (27) Bedard, D. L.; May, R. J. Environ. Sci. Technol. 1996, 30, 237245. (28) Imamoglu, I. Ph.D. Dissertation, Department of Civil Engineering and Mechanics, University of WisconsinsMilwaukee, Milwaukee, WI, 2001. (29) Li, K.; McMullin, J.; Imamoglu, I.; Christensen, E. R. Quantification of polychlorinated biphenyls (PCBs) in the sediments of Fox River and Green Bay; Progress Report No. 1; Department of Civil Engineering and Mechanics, University of Wisconsins Milwaukee: Milwaukee, WI, 2000. (30) Patterson, D.; Velleux, M.; Witthuhn, J. Compilation and estimation of historical discharges of Total Suspended Solids and PCBs from Lower Fox River point sources; Technical Memorandum 2d; Wisconsin Department of Natural Resources: Madison, WI, 1999. (31) Frame, G. M.; Cochran, J. W.; Bowadt, S. S. J. High Resolut. Chromatogr. 1996, 19, 657-668. (32) Bedard, D. L. Polychlorinated bephenyls in aquatic sediments: Environmental fate and outlook for biological treatment. In Dechlorination: Microbial Processes and Environmental Applications; Bossert, I. D., Ed.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2003; pp 443-465. (33) Klasson, K. T.; Just, E. M. Biorem. J. 2002, 6, 77-85. (34) Mackay, D.; Shiu, W. Y.; Ma, K. C. Illustrated Handbook of Physical-Chemical Properties and Environmental Fate for Organic Chemicals: Monoaromatic Hydrocarbons, Chlorobenzenes, and PCBs; Lewis Publishers: Boca Raton, FL, 1992. (35) Brown, J. F. J.; Wagner, R. E.; Feng, H.; Bedard, D. L.; Brennan, M. J.; Carnahan, J. C.; May, R. J. Environ. Toxicol. Chem. 1987, 6, 579-593. (36) Quensen, J. F., III; Tiedje, J. M. Evaluation of PCB dechlorination in sediments. In Methods in Biotechnology, Vol.2: Bioremediation Protocols; Sheenan, D., Ed.; Humana Press: Totawa, NJ, 1997; pp 257-273. (37) McLaughlin, D. B. Ph.D. Dissertation, Land Resources, University of WisconsinsMadison, Madison, WI, 1994. (38) Quensen, J. F., III; Tiedje, J. M.; Boyd, S. A. Science 1988, 242, 752-754.
Received for review October 20, 2003. Revised manuscript received February 18, 2004. Accepted February 19, 2004. ES035165X
VOL. 38, NO. 9, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2583
