Identification of Structural Properties Associated ... - ACS Publications

Dec 21, 2009 - The C4.5 algorithm in Weka, which maximizes information gain while penalizing branches with relatively more nodes, allows nodes to appe...
0 downloads 0 Views 1MB Size
Environ. Sci. Technol. 2010, 44, 2842–2848

Identification of Structural Properties Associated with Polychlorinated Biphenyl Dechlorination Processes A M A N D A S . H U G H E S , †,‡ J E A N N E M . V A N B R I E S E N , * ,‡ A N D M I T C H E L L J . S M A L L †,‡ Department of Engineering & Public Policy, Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, Pennsylvania 15213 and Department of Civil & Environmental Engineering, Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, Pennsylvania 15213

Received July 14, 2009. Revised manuscript received November 30, 2009. Accepted December 2, 2009.

Polychlorinated biphenyl molecules can be biologically dechlorinated through sequential losses of a chlorine atom, following 840 pathways from higher chlorinated to lesserchlorinated congeners and biphenyl. Previously, eight recurring sets of pathways, herein referred to as explicitly reported pathways in dechlorination processes, have been identified through qualitative analysis of shifts in congener masses in field and laboratory studies. Dechlorination process generalizations were qualitatively extrapolated based on limited attributes of the congeners dechlorinated in the explicitly reported pathways. They are valuable because they allow comparisons of dechlorination patterns across laboratory experiments and contaminated sites. However, due to analytical limitations and a paucity of studies, the explicitly reported pathways in dechlorination processes likely do not represent all of the pathways that could occur at contaminated sites. This work presents an alternative, quantitative, and replicable approach to the identification of candidate pathways for inclusion in dechlorinationprocessgeneralizationsthroughuseofclassification trees. This method considers 46 structural and property attributes of dechlorination pathways. Trees fit for pathway inclusion in each of the eight dechlorination processes with alternative assumptions are compared in terms of critical congener attributes. The classification trees correctly classify explicitly reported pathways into dechlorination processes at rates of 0.90 to 0.99. While many of the attributes used in the original generalizations were also selected as predictors by the classification trees, the extra attributes allow identification of additional dechlorination pathways that can be considered as candidates for monitoring in future studies.

Introduction An estimated 1.3 million tons of polychlorinated biphenyls (PCBs) have been produced around the world (1) for use in * Corresponding author phone: (412) 268-4603; fax: (412) 2687813; e-mail: [email protected]. † Department of Engineering & Public Policy, Carnegie Mellon University. ‡ Department of Civil & Environmental Engineering, Carnegie Mellon University. 2842

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 8, 2009

capacitors, transformers, fire retardants, and plasticizers (2). After recognition of their environmental persistence and potential threat to human health, their production was phased out in the late 1970s. Since that time, policy concerns have shifted toward the management of PCB residues in contaminated river, lake, and estuary sediments. The management of such sediments is complex due to the nature of PCBs, which are composed of 209 distinct anthropogenic molecules called congeners. PCB congeners are composed of a biphenyl ring with 10 possible attachment sites for chlorines, labeled 2 through 6 on each ring. The range of congener structures results in a wide range of behaviors in the environment. Generally, highly chlorinated congeners sorb to organic matter and are eventually deposited in sediment, while less chlorinated congeners are more mobile in water and air. Further, some congeners selectively partition into the lipids of living organisms, causing bioaccumulation in the food chain. In sediment, PCBs can be transformed through physical and chemical means, but the most significant long-term mass reductions are achieved via biotransformation. Biotransformation occurs through two distinct activities: aerobic oxidative degradation targeting the carbon rings and anaerobic reductive dechlorination targeting attached chlorines (3). Aerobic degradation typically occurs in the top few centimeters of sediment, is active on the less chlorinated congeners, results in destruction of the biphenyl ring, and can be carried out by many microorganisms (3, 4). In contrast, anaerobic dechlorination is active on more highly chlorinated congeners, results in the transformation of one PCB congener to another, and is carried out by specialized populations of microorganisms (5). Dechlorination is important because it typically results in a reduction in risk because lesserchlorinated congeners biomagnify at lesser rates and to lesser extents, resulting in less human exposure to PCBs through the food chain (6, 7). Notably, reductions in risk due to dechlorination are not directly reflected by total PCB mass measurements because dechlorination can significantly reduce chlorination level without appreciably changing total mass. Nonetheless, sediment remedial policies focus heavily on total PCB concentrations, rather than congener concentrations. The transformation of a higher-chlorinated to a lesserchlorinated congener through the loss of a single chlorine atom is termed a dechlorination pathway. In the 1980s, dechlorination patterns resulting from subsets of the 840 possible pathways connecting the 209 congeners to each other and biphenyl, were recognized in sediments originating from multiple sites (8–10). The pathways involved in these patterns, termed dechlorination processes (DPs) (11–16), have since been formalized through reviews of published and unpublished congener-specific laboratory and environmental analyses (11–17). DPs are a valuable tool for researchers because they enable comparisons of dechlorination experiments and field observations. To date, eight single DPs have been identified on the basis of targeted parent congeners and chlorines, and subsequent daughter congeners: H, H′, LP, M, N, P, Q, and T (11–17). Of the 840 possible pathways, 108 are explicitly reported to belong to one or more DPs. It has been hypothesized that a single microorganism, or possibly consortia of microorganisms, is responsible for each DP (12, 18, 19). Hiraishi (20) presents a review of microbial species reported to show some PCB dechlorinating activity; all known species show some congener specificity. While PCB dechlorinating organisms have recently been identified (17, 21–26), full sequencing exists only for Dehalococcoides, 10.1021/es902109w

 2010 American Chemical Society

Published on Web 12/21/2009

and thus, putative dehalogenases have not been identified for PCBs. However, dechlorinating organisms that target chlorinated solvents are known to contain multiple putative dehalogenases. It is likely that PCB dechlorinators also contain multiple dehalogenases and that this accounts for their ability to dechlorinate multiple structurally dissimilar congeners. In addition, congener-specificities are known to be affected by concentrations of nearby PCBs, incubation or environmental conditions, and intermicrobial competition (12, 14, 27–29), suggesting either different microbial species dominate under different conditions or differential activation of different dehalogenases based on environmental conditions. Similarly, it has been suggested that DPs are associated with particular sediments or sediment conditions (11–14, 16, 18). Frequent reports of dechlorination that is similar to but never exactly like the observed pathways in DPs (for example 6, 30) further suggest the influence of sediment conditions. While pathway classification into DPs was a major step forward in understanding how different pathways are related to one another in the environment, its application was limited by the analytical methods to detect all congeners, and in the paucity of congener-specific studies. For instance, sediment from only one contaminated site was examined in order to identify the explicitly reported pathways in DP LP (13, 16, 17). Consequently, the explicitly reported pathways likely represent only a subset of the possible pathways that might be related to the responsible specific microorganisms or specific dehalogenases. To overcome this limitation, dechlorination process generalizations (DPGs) were qualitatively extrapolated from the explicitly reported pathways (12). DPGs relate targeted groups of congeners in a DP, include more pathways than were explicitly reported, and occasionally exclude explicitly reported pathways. They utilize chlorophenylgroups, which describe the chlorine structure on only one of the biphenyl rings. For example, chlorophenyl group 23has chlorines in positions 2 and 3 on one phenyl ring. Negative findings, including excluded homologues, dechlorination types (meta, para and ortho) and positions (flanked, doubly flanked), and chlorophenyl groups (12, 31–33), were considered in the construction of DPGs. Negative findings may be due to the enzymatic specificities of the responsible microorganism(s) and therefore represent important information about DP exclusivity. However, negative findings related to the biogeochemical conditions in which microorganism(s) existed or the limited time periods for which many of the experiments took place may have excluded possible pathways for reasons unrelated to microbial specificity. For these reasons, dechlorination pathways included in DPGs are likely a subset of the possible pathways. Despite this limitation, DPGs are commonly reported in the literature (for example, refs 34, 35) and have been widely used in the construction of retrospective dechlorination models (for example, refs 36, 37). Since our understanding of the suitability of any remedial action will rely on our ability to adequately characterize the natural or enhanced biodegradation potential and rate at a given site, it is critical that we consider congener-specific transformations and their likelihood in natural systems. Such a characterization has traditionally been problematic because the relevant PCB-dechlorinating microorganisms remain largely unidentified (5). With recent identification of PCB dechlorinating cultures and species (17, 21–26), it is possible that a full characterization of biodegradation potential and rate could be undertaken through comprehensive laboratory experiments; however, such studies would be costly and timeconsuming given the 209 congeners and multiple bacterial species involved. There exists an unmet need for dechlorination models that simulate the potential for all 840 pathways to occur, recognizing that the potential for some or many of

these pathways will be very low in many biogeochemical settings. This development will compliment laboratory work to elucidate the links between specific dehalogenase enzymes and processes observed in field sites. However, complete pathway and process identification is daunting as the problem spans 840 pathways (651 of which have parent congeners present in Aroclors (38)). This level of coverage is needed to provide improved retrospective identification of DPs and facilitate quantitative, systematic identification of new DPs. In this paper, we present a novel statistical approach that will aid in the achievement of this goal. It is not intended to replace traditional laboratory techniques but rather to compliment them. Improved knowledge of dechlorination processes and their retrospective identification in laboratory and field studies will enable elucidation of the links between biogeochemical conditions and dechlorination process activity.

Methods Classification trees provide a statistical method for finding the optimal combination of explanatory attributes to predict set membership. Previous applications include prediction of chemical persistence (39), an examination of the relationship between coral reef abundance and environmental conditions (40) and identification of environmental factors that predict fish species presence (41). These applications demonstrated the ability of CTs to handle discrete and continuous attributes and provide interpretable results (40). Here, CTs identify the missing sets of possible pathways from known DPs based on the structures of the parent and daughter congeners in the explicitly reported pathways: thereby adhering to the widely published structure-based premise that was established in the identification of DPGs (12). Furthermore, the CT approach is systematic and quantitative, where the former approach was more qualitative and difficult to replicate across experts. The resulting eight sets of explicitly reported plus missing dechlorination pathways are titled Classification Tree Dechlorination Process Generalizations (CTDPGs). A CT was developed for each DP using the C4.5 algorithm with standard pruning (42) in Weka software (43). Trees were created from sequential additions of nodes (rectangles in Figure 1), which represent predictive attributes of pathway inclusion or exclusion from a DP. A series of nodes terminating in classification into or out of a DP (diamonds and ovals in Figure 1) is termed a branch. The C4.5 algorithm in Weka, which maximizes information gain while penalizing branches with relatively more nodes, allows nodes to appear more than once throughout the tree (43). Final pruning excludes bifurcations that do not yield sufficient differentiation. Forty-six attributes of each of the 840 pathways were considered in this analysis. Note that all possible dechlorination pathways were included in our analysis for completeness. Three of the 46 attributes describe the loss of an ortho, meta, or para chlorine across the pathway, while the remaining attributes describe the pathway’s parent congener (Table 1). Included attributes may be functions of each other but have Pearson correlation coefficients less than 0.9. Potentially relevant attributes like diffusivity and free energies of formation, were excluded for having Pearson correlation coefficients greater than 0.9 with an included attribute. All attributes were standardized to avoid bias introduced by differences in the magnitude and spread of values. This method is suitable for application to DPs because it allows attributes to be predictive individually or in combination at different levels for subsets of the analyzed dechlorination pathways. For example, if a DPG states that the number of ortho flanked meta chlorines is a predictor of DP inclusion, then a CT is likely to identify sequential nodes relating the VOL. 44, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2843

FIGURE 1. Classification tree for dechlorination process M created using 41 nonhomologue-like attributes.

TABLE 1. Attributes Used in Classification Tree Creationa attribute description

source

10 attributes for chlorine presence at positions 1-10 on parent congener homologue of the parent congener* number of ortho chlorines present on parent congener number of meta chlorines present on parent congener number of para chlorines present on parent congener 19 variables for chlorophenyl group presence on the parent congener number of flanked para chlorines on parent congener number of doubly flanked para chlorines on parent congener number of flanked meta chlorines on parent congener number of doubly flanked meta chlorines on parent congener loss of an ortho chlorine from the parent congener loss of a meta chlorine from the parent congener loss of a para chlorine from the parent congener

compiled through analysis of structures

planarity of the parent congener vapor pressure of the parent congener* water solubility of the parent congener* Henry’s Law constant of the parent congener* mean atomic Sanderson electronegativity of the parent congener mean atomic polarizability of the parent congener*

(46)

a Attributes appearing with an asterisk were considered to be homologue-like and were therefore omitted from classification tree analyses used to elucidate true homologue range.

number or position of ortho chlorines and the number of flanked meta chlorines. DP data present an unusual challenge to predictive modeling. Although pathways belonging in a given DP are explicitly reported, reports of excluded pathways are likely sediment-specific. The 840 dechlorination pathways were divided into three groups for each DP: explicitly reported pathways; explicitly reported pathways in other DPs; and pathways that have not been explicitly reported to be in any DP. The first two sets of pathways, totaling 108, were used to create CTs. Explicitly reported pathways in other DPs are a first approximation of pathways that were considered for inclusion in the given DP but were determined not to belong in that process. This inexact classification stems from a lack of information regarding which congeners were measured in cases where Aroclors were investigated as well as congener quantification affected by analytical limits. Then CTs were used to categorize the remaining 732 pathways as belonging or not belonging in each DP. Sets of pathways predicted to be in a DP by a CT plus those explicitly reported to be in the process are termed CTDPGs. The base CT analysis presented here does not incorporate negative findings. Consequently, evaluations of pathway inclusion in CTDPGs are not limited by the subset of biogeochemical conditions that were present in the studies 2844

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 8, 2009

used to identify DPGs. Thus, we hypothesize that CTDPGs represent the suite of pathways that the microorganism(s) responsible for a DP is capable of degrading across all possible biogeochemical sediment conditions. We expect that some of the predicted pathways will be inconsistent with the DPGs and observed pathways in the field and laboratory studies. It may be that these unexpected pathways are less likely to be active due to unfavorable geochemical sediment conditions or the presence of more energetically favorable dechlorination pathways. The identification of predicted yet unobserved pathways is of particular interest for future exploration of the role of sediment conditions controlling PCB dechlorination.

Results and Discussion The CTDPG analyses added 486 pathways to the 108 explicitly reported pathways in DPs (a breakdown by DP appears in Table 2). Performance measures of CTs convey their ability to sort the 108 pathways used in their creation into their preanalysis categories. Performance measurements for the eight CTs appear in Table 3. Overall performance, measured in terms of the percent of correctly classified pathways (defined as predicting a pathway’s prior classification), ranged from 90% to 99%. These measurements are comparable to

TABLE 2. Comparison of the Number of Pathways that Were Explicitly Reported, and Appear in Dechlorination Process Generalizations, Classification Tree Dechlorination Process Generalizations without and with Ortho Dechlorination Exclusiona dechlorination process

H

H′

LP

M

N

P

Q

T

all

explicitly reported DPG CTDPG CTDPG with ortho exclusion

22 69 91 91

22 55 76 76

33 54 199 73

17 53 108 88

29 124 204 204

28 92 117 116

22 57 76 34

6 10 25 39

108 330 594 448

a Note that dechlorination process LP is most recently and also most thoroughly studied DP. While its dechlorination process generalization in 2003 is straightforward (15), in 2005 and 2008 terms like “sometimes” are used to describe targeted chlorophenyl groups (16, 17). The set of pathways included in dechlorination process generalization LP is conservatively interpreted here due to uncertainty regarding which congeners within some chlorophenyl groups are targeted.

TABLE 3. Classification Tree Performance Criteria for All 8 Dechlorination Processes

c

dechlorination process

H

H′

LP

Ma

Mb,c

N

P

Qa,c

Qb

T

tree size correct classification rate true positive rate true negative rate

25 0.90 1.00 0.87

23 0.96 0.86 0.99

25 0.96 0.94 0.97

11 0.99 1.00 0.99

17 0.96 1.00 0.96

11 0.94 0.97 0.94

13 0.96 1.00 0.95

15 0.94 0.91 0.95

23 0.96 0.86 0.99

9 0.99 1.00 0.99

a Classification tree created using all 46 attributes. b Classification tree created using 41 nonhomologue like attributes. Classification tree selected to represent the dechlorination process.

or significantly exceed percentages reported in other CT applications (39–41). Perfect true positive rates, which describe the rate at which a CT includes a pathway that was explicitly reported to be in a DP, were achieved by half of the CTs. The CTs’ exhibit slightly lower true negative rates (0.87 to 0.99), consistent with the greater uncertainty associated with pathways determined a priori to be excluded from a DP. Application of the CT method to Process M is presented here in detail while discussions of the remaining DPs appear in the Supporting Information. DPG M states that flanked and unflanked meta chlorines on chlorophenyl groups 3-, 23-, 25-, 34-, 234-, and 236- in homologues 2 through 4 are targeted for removal (15). However, the true homologue range of DPG M is unknown due in large part to a limited number of studies (12, 44). The creation of two CTs provides insight. The first CT (Figure S4 in the Supporting Information) was created using all 46 attributes. Pathways predicted to be in CTDPG M by this first CT directly reflect the extent of the known homologue range (2-4) through inclusion of the homologue attribute. However, it is possible that exclusion of pathways with parent congeners outside the known 2 to 4 homologue range is an artifact of the small number of available studies, as opposed to a characteristic of the DP. Therefore, a second CT (Figure 1) that omits attributes that split the data into homologue-like groups (noted in Table 1) was created and then applied to all 840 pathways as follows: a pathway starts at the top of the tree and moves down through it by comparison of its attribute values to those appearing below each attribute’s node in the CT. If these values are in agreement, then the pathway follows the line leading away from that attribute and to the right (and to the left if the values are not in agreement). It eventually reaches the end of a branch, where it is classified as being included or excluded from the DP. For example, pathway 25-35 to 35-2 describes the loss of a meta chlorine. Thus, the root node’s criterion is met and the pathway is next tested for its ability to meet the criterion of the 234-chlorophenyl group attribute. This pathway continues moving through the tree until it reaches the end of a branch, which specifies inclusion in CTDPG M. The second tree shares four attributes with the first tree and in addition to predicting the inclusion of the same pathways, it includes 53 pathways having greater than 4 chlorines on the parent congener. While both trees have

perfect true positive rates, the second tree has a slightly smaller false positive rate because four pathways assumed to be out of the process a priori are predicted to belong in DP M. Because these pathways describe the meta dechlorination of flanked or doubly flanked chlorines they are not considered a significant hindrance to the adoption of this CT. Thus, we cannot exclude the possibility that congeners with greater than 4 chlorines can be degraded as part of DP M. For DP M, the CT method yields a CTDPG suggesting a dechlorination capability that was not identified by the more traditional approach. In the latter CTDPG M, 104 of the 108 pathways were correctly classified. Nodes describing the removal of a meta chlorine and the exclusion of pathways with parent congeners having chlorophenyl groups 235-, 245-, and 2345- are in agreement with the DPG M. Further, the CT indicates that the governing microorganism(s) is unable to degrade congeners having greater than two ortho chlorines. This indication is in agreement with recently reported behavior by the microorganism(s) believed to govern DP LP (16). Negative findings associated with DP M are the absence of para dechlorination of congeners in Aroclor 1242 and no accumulation of congeners 2-3, 24-3, 25-3, and 26-3 (12). Pathways in CTDPG M (Figure 2) do not describe the para dechlorination of a congener in any of the Aroclor 1242 lots (38). However, pathways leading to the accumulation of congeners 2-3, 24-3, 25-3, and 26-3 are predicted three, one, one, and two times, respectively, indicating some false positive predictions. All 17 explicitly reported pathways are included among the 53 pathways in DPG M, which shares 43 pathways with the CTDPG. All but four of the 65 pathways that appear in the CTDPG M but not in the DPG have parent congeners with greater than four chlorines. A significant difference between CTDPGs and the previously developed DPGs is the presence of ortho dechlorination pathways. CTDPGs LP, Q, and T predict, based solely on structural similarities to explicitly reported pathways, that the responsible microorganisms are capable of carrying out 137 (of 320) ortho dechlorination pathways. While ortho dechlorination was not explicitly reported in any DP, three such pathways were observed in laboratory studies of Woods Pond sediment, in which DP LP was observed (17). These three pathways appear in CTDPG LP, suggesting that the microorganism(s) responsible for DP LP is also responsible VOL. 44, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2845

FIGURE 2. Explicitly reported pathways (dashed lines) and pathways added to dechlorination process M through the classification tree analysis (solid lines). Note that the numbers are arranged by homologue and correspond to congener structures assigned in the original work of Ballschmiter and Zell (47) with corrections to congener numbers 199-201 by Schulte and Malisch (48) and corrections to numbers 107-109 by Guitart et al. (49) (0 represents biphenyl).

FIGURE 3. Results of 100 stratified 10-fold cross-validations for each dechlorination process. for at least these three ortho dechlorination pathways. Additionally, ortho chlorine removal from congener 2356-, which is reported for bacterium o-17 appears in CTDPG LP (45). To partially evaluate the implications of including negative findings in CT creation, a second analysis was performed. The results of prohibiting ortho dechlorination from all DPs a priori are reported in the Supporting Information. As expected, all of the CTs exclude ortho dechlorination when this restriction is forced. Tree sizes and performance criteria of CTDPGs that did not include ortho dechlorination in the base analysis did not change significantly. However, significant decreases in true positive rates (-0.12 on average) were observed for DPs LP, Q and T, whose base CTDPGs included ortho dechlorination. Because the a priori exclusion of ortho dechlorination hindered the CTs’ ability to correctly classify the same set of explicitly reported pathways, it is 2846

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 8, 2009

likely that such a broad exclusion may be counterproductive, especially in a first screening for pathway inclusion in DPs. A focus on the geochemical conditions that might limit ortho dechlorination in sediments where DPs LP, Q and/or T have been detected is strongly indicated. Application of the CT method was validated by 100 stratified 10-fold cross validations (43). In this standard statistical method, 90% of the data with known classification are used to construct a CT and the remaining 10% of the data set are used to test its performance. The withheld data are randomly chosen and this process is repeated 1000 times (43). Results are summarized in Figure 3. The relatively small error bars associated with true positive rates for most DPs and the small variations in the standard deviations of the CT structures (approximated by CT size) suggest that the estimated CTs are robust.

Implications The alternative approach to DPGs presented here quantitatively and systematically generates a set of dechlorination pathways that, based solely on their structure, may be degraded by the microorganism(s) responsible for DPs. Notably, this method may also be applied to the results of experimental studies in order to identify unobserved pathways that could be active under optimal biogeochemical conditions by the microorganism(s) active during the time of the study. It is possible that the set of candidate pathways identified by such a CT analysis agree with those appearing in DPs. For example, Dehalococcoides sp. Strain CBDB1, was reported to produce a pattern that “matches PCB dechlorination Process H” (19). A CT analysis for pathways reported in (19) generated a CT with true and false positive rates of 0.94 and 1.00, respectively (details appear in the Supporting Information). However, the CT for microorganism CBDB1 shares only four predictive attributes with the CT for DP H (eight are not shared). This difference suggests that CBDB1 is closely related to or coincidentally shares some dechlorination pathways with the microorganism(s) responsible for DP H but is not the microorganism responsible for it. In the course of the CT analyses, the unknown homologue ranges of two DPs and the value of negative findings were elucidated. Complete validation of the CTDPGs would require isolation of the microorganisms responsible for each DP and extensive experimental analysis that is likely time and cost prohibitive. The CTDPG can reduce the extent of the required analysis by providing a set of dechlorination pathways to target in the planning phase of experimental studies, thus providing a complementary approach to understand microbially mediated dechlorination of PCBs.

Acknowledgments This work has been funded by the Strategic Environmental Research and Development Program (Project ER 1495: “Modeling and Decision Support Tools Based on the Effects of Sediment Geochemistry and Microbial Populations on Contaminant Reactions in Sediments”). The authors thank members of Carnegie Mellon University’s Center for Water Quality in Urban Environmental Systems for fruitful discussions of laboratory methods. The valuable comments in e-mail exchanges with Dr. Donna Bedard and the comments of the anonymous reviewers are also gratefully acknowledged.

Supporting Information Available Discussions and figures of the seven remaining base CTs as well as secondary analyses in which negative findings are included in CT creation appear. Also included are tables presenting the distribution of explicitly reported pathways across homologue and pathway inclusion in DPs and a CT analysis of bacterium CBDB1. This information is available free of charge via the Internet at http://pubs.acs.org/.

Literature Cited (1) Breivik, K.; Sweetman, A.; Pacyna, J. M.; Jones, K. C. Towards a global historical emission inventory for selected PCB congenerssA mass balance approach 1. Global production and consumption. Sci. Total Environ. 2002, 290 (1-3), 181–198. (2) Hutzinger, O.; Safe, S.; Zitko, V., The Chemistry of PCB’s; CRC Press: Cleveland, 1974; p 269. (3) Abramowitz, R. Aerobic and anaerobic biodegradation of PCBssA review. Crit. Rev. Biotechnol. 1996, 10, 241–249. (4) Field, J. A.; Sierra-Alvarez, R. Microbial transformation and degradation of polychlorinated biphenyls. Environ. Pollut. 2008, 155 (1), 1–12. (5) Borja, J.; Taleon, D. M.; Auresenia, J.; Gallardo, S. Polychlorinated biphenyls and their biodegradation. Process Biochem. 2005, 40 (6), 1999–2013.

(6) Bedard, D. L.; May, R. J. Characterization of the polychlorinated biphenyls in the sediments of Woods Pond: Evidence for microbial dechlorination of Aroclor 1260 in situ. Environ. Sci. Technol. 1996, 30 (1), 237–245. (7) McFarland, V.; Clarke, J. Environmental occurrence, abundance, and potential toxicity of polychlorinated biphenyl congeners: Considerations for a congener-specific analysis. Environ. Health Perspect. 1989, 81, 225–239. (8) Brown, J. F.; Bedard, D. L.; Brennan, M. J.; Carnahan, J. C.; Feng, H.; Wagner, R. E. Polychlorinated biphenyl dechlorination in aquatic sediments. Science. 1987, 236 (4802), 709–712. (9) Brown, J. F.; Wagner, R. E.; Feng, H.; Bedard, D. L.; Brennan, M. J.; Carnahan, J. C.; May, R. J. Environmental dechlorination of PCBs. Environ. Toxicol. Chem. 1987, 6 (8), 579–593. (10) Brown, J. F.; Wagner, R. E. PCB movement, dechlorination, and detoxification in the Acushnet Estuary. Environ. Toxicol. Chem. 1990, 9 (10), 1215–1233. (11) Bedard, D. L.; Bunnell, S. C.; Smullen, L. A. Stimulation of microbial para-dechlorination of polychlorinated biphenyls that have persisted in Housatonic River sediment for decades. Environ. Sci. Technol. 1996, 30 (2), 687–694. (12) Bedard, D. L.; Quensen, J. F. “Microbial Reductive Dechlorination of Polychlorinated Biphenyls.” In Microbial Transformation and Degradation of Toxic Organic Chemicals; Young, L., Cerniglia, C., Eds.; John Wiley & Sons: New York, 1995. (13) Bedard, D. L.; VanDort, H. M.; May, R. J.; Smullen, L. A. Enrichment of microorganisms that sequentially meta, paradechlorinate the residue of Aroclor 1260 in Housatonic River sediment. Environ. Sci. Technol. 1997, 31 (11), 3308–3313. (14) Wu, Q. Z.; Bedard, D. L.; Wiegel, J. Temperature determines the pattern of anaerobic microbial dechlorination of Aroclor 1260 primed by 2,3,4,6-tetrachlorobiphenyl in Woods Pond sediment. Appl. Environ. Microbiol. 1997, 63 (12), 4818–4825. (15) Bedard, D. L. “Polychlorinated Biphenyls in Aquatic Sediments: Environmental Fate and Outlook for Biological Treatment.” In Dehalogenation: Microbial Processes and Environmental Applications; Haggblom, M. M., Bossert, I. D., Eds.; Kluwer Academic Publishers: Boston, 2003; pp 443-465. (16) Bedard, D. L.; Pohl, E. A.; Bailey, J. J.; Murphy, A. Characterization of the PCB substrate range of microbial dechlorination process LP. Environ. Sci. Technol. 2005, 39 (17), 6831–6838. (17) Bedard, D. L. A case study for microbial biodegradation: Anaerobic bacterial reductive dechlorination of polychlorinated biphenyls-from sediment to defined medium. Annu. Rev. Microbiol. 2008, 62 (1), 253–270. (18) Bedard, D. L.; Jerzak, G. V.; Bailey, J. J. Strategies for the selective enrichment of microorganisms carrying out reductive dechlorination of polychlorinated biphenyls in freshwater sediments. Fresenius Environ. Bull. 2003, 12 (3), 276–285. (19) Adrian, L.; Dudkova´, V.; Demnerova´, K.; Bedard, D. L. Dehalococcoides sp. strain CBDB1 extensively dechlorinates the commercial polychlorinated biphenyl (PCB) mixture Aroclor 1260. Appl. Environ. Microbiol. 2009, 75 (13), 4516–4524. (20) Hiraishi, A. Biodiversity of dehalorespiring bacteria with special emphasis on polychlorinated biphenyl/dioxin dechlorinators. Microbes Environ. 2008, 23 (1), 1–12. (21) Cutter, L. A.; Watts, J. E. M.; Sowers, K. R.; May, H. D. Identification of a microorganism that links its growth to the reductive dechlorination of 2,3,5,6-chlorobiphenyl. Environ. Microbiol. 2001, 3 (11), 699–709. (22) Wu, Q. Z.; Watts, J. E. M.; Sowers, K. R.; May, H. D. Identification of a bacterium that specifically catalyzes the reductive dechlorination of polychlorinated biphenyls with doubly flanked chlorines. Appl. Environ. Microbiol. 2002, 68 (2), 807–812. (23) Fagervold, S. K.; Watts, J. E. M.; May, H. D.; Sowers, K. R. Sequential reductive dechlorination of meta-chlorinated polychlorinated biphenyl congeners in sediment microcosms by two different Chloroflexi phylotypes. Appl. Environ. Microbiol. 2005, 71 (12), 8085–8090. (24) May, H. D.; Miller, G. S.; Kjellerup, B. V.; Sowers, K. R. Dehalorespiration with polychlorinated biphenyls by an anaerobic ultramicrobacterium. Appl. Environ. Microbiol. 2008, 74 (7), 2089–2094. (25) Fennell, D. E.; Nijenhuis, I.; Wilson, S. F.; Zinder, S. H.; Haggblom, M. M. Dehalococcoides ethenogenes strain 195 reductively dechlorinates diverse chlorinated aromatic pollutants. Environ. Sci. Technol. 2004, 38 (7), 2075–2081. (26) Bedard, D. L.; Bailey, J. J.; Reiss, B. L.; Jerzak, G. V. Development and characterization of stable sediment-free anaerobic bacterial enrichment cultures that dechlorinate Aroclor 1260. Appl. Environ. Microbiol. 2006, 72 (4), 2460–2470. VOL. 44, NO. 8, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2847

(27) Wu, Q. Z.; Bedard, D. L.; Wiegel, J. Effect of incubation temperature on the route of microbial reductive dechlorination of 2,3,4,6-tetrachlorobiphenyl in polychlorinated biphenyl (PCB)-contaminated and PCB-free freshwater sediments. Appl. Environ. Microbiol. 1997, 63 (7), 2836–2843. (28) Cho, Y. C.; Sokol, R. C.; Frohnhoefer, R. C.; Rhee, G. Y. Reductive dechlorination of polychlorinated biphenyls: Threshold concentration and dechlorination kinetics of individual congeners in Aroclor 1248. Environ. Sci. Technol. 2003, 37 (24), 5651–5656. (29) Rhee, G. Y.; Bush, B.; Bethoney, C. M.; Denucci, A.; Oh, H. M.; Sokol, R. C. Reductive dechlorination of Aroclor 1242 in anaerobic sedimentssPattern, rate and concentration-dependence. Environ. Toxicol. Chem. 1993, 12 (6), 1025–1032. (30) Fagervold, S. K.; May, H. D.; Sowers, K. R. Microbial reductive dechlorination of Aroclor 1260 in Baltimore Harbor sediment microcosms is catalyzed by three phylotypes within the Phylum Chloroflexi. Appl. Environ. Microbiol. 2007, 73 (9), 3009–3018. (31) Barreira, L. A.; Bebianno, M. J.; Mudge, S. M.; Ferreira, A. M.; Albino, C. I.; Veriato, L. M. Relationship between PCBs in suspended and settled sediments from a coastal lagoon. Cienc. Mar. 2005, 31 (1B), 179–195. (32) Battley, E. H. The development of direct and indirect methods for the study of the thermodynamics of microbial growth. Thermochim. Acta 1998, 309 (1-2), 17–37. (33) Baughman, G. L.; Paris, D. F.; Steen, W. C., “Quantitative Expression of Biotransformation Rate.” In Biotransformation and Fate of Chemicals in the Environment, Agency, U. S. E. P., Ed. American Society for Microbiology: Washington, D.C., 1980; pp 105-111. (34) Magar, V. S.; Johnson, G. W.; Brenner, R. C.; Quensen, J. F.; Foote, E. A.; Durell, G.; Ickes, J. A.; Peven-McCarthy, C. Longterm recovery of PCB-contaminated sediments at the Lake Hartwell superfund site: PCB dechlorination. 1. End-member characterization. Environ. Sci. Technol. 2005, 39 (10), 3538– 3547. (35) Williams, W. A. Stimulation and enrichment of two microbial polychlorinated biphenyl reductive dechlorination activities. Chemosphere. 1997, 34 (3), 655–669. (36) Bzdusek, P. A.; Lu, J. H.; Christensen, E. R. PCB congeners and dechlorination in sediments of Sheboygan River, Wisconsin, determined by matrix factorization. Environ. Sci. Technol. 2006, 40 (1), 120–129. (37) Imamoglu, I.; Li, K.; Christensen, E. R.; McMullin, J. K. Sources and dechlorination of polychlorinated biphenyl congeners in the sediments of Fox River, Wisconsin. Environ. Sci. Technol. 2004, 38 (9), 2574–2583.

2848

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 8, 2009

(38) Frame, G. M.; Cochran, J. W.; Bowadt, S. S. Complete PCB congener distributions for 17 Aroclor mixtures determined by 3 HRGC systems optimized for comprehensive, quantitative, congener-specific analysis. HRC-J. High Res. Chromatogr. 1996, 19 (12), 657–668. (39) Bennett, D. H.; McKone, T. E.; Kastenberg, W. E. Evaluating multimedia chemical persistence: Classification and regression tree analysis. Environ. Toxicol. Chem. 2000, 19 (4), 810–819. (40) De’ath, G.; Fabricius, K. E. Classification and regression trees: A powerful yet simple technique for ecological data analysis. Ecology 2000, 81 (11), 3178–3192. (41) Steen, P. J.; Zorn, T. G.; Seelbach, P. W.; Schaeffer, J. S. Classification tree models for predicting distributions of Michigan stream fish from landscape variables. Trans. Am. Fish. Soc. 2008, 137 (4), 976–996. (42) Quinlan, J. R. C4.5 Programs for Machine Learning; Morgan Kauffman: San Mateo, 1993. (43) Witten, I.; Frank, E., Data Mining: Practical Machine Learning Tools and Techniques, 2nd ed.; Morgan Kaufmann: San Francisco, 2005. (44) Bedard, D. L. Rensselaer Polytechnic Institute: Troy, NY., personal communication. 2009. (45) May, H. D.; Cutter, L. A.; Miller, G. S.; Milliken, C. E.; Watts, J. E. M.; Sowers, K. R. Stimulatory and inhibitory effects of organohalides on the dehalogenating activities of PCB-dechlorinating bacterium o-17. Environ. Sci. Technol. 2006, 40 (18), 5704–5709. (46) Oberg, T. Prediction of physical properties for PCB congeners from molecular descriptors. Internet J. Chem. 2001, 4 (11), . (47) Ballschmiter, K.; Zell, M. Analysis of polychlorinated-biphenyls (PCB) by glass-capillary gas-chromatographyscomposition of technical Aroclor-PCB and Clophen-PCB mixtures. Fresenius Z. Anal. Chem. 1980, 302 (1), 20–31. (48) Schulte, E.; Malisch, R. Calculation of the real PCB content in environmental-samples: 1. Investigation of the composition of 2 technical PCB mixtures. Fresenius Z. Anal. Chem. 1983, 314 (6), 545–551. (49) Guitart, R.; Puig, P.; Gomezcatalan, J. Requirement for a standardized nomenclature criterium for PCBssComputerassisted assignment of correct congener denomination and numbering. Chemosphere 1993, 27 (8), 1451–1459.

ES902109W