Development and Validation of Protocols To ... - ACS Publications

DFO/UBC Centre for Aquaculture and Environmental Research, 4160 Marine Drive, West Vancouver, British Columbia, Canada V7 V 1N6. Environ. Sci. Technol...
3 downloads 4 Views 2MB Size
Download PDF
ARTICLE pubs.acs.org/est

Development and Validation of Protocols To Differentiate PCB Patterns between Farmed and Wild Salmon Mark B. Yunker,*,† Michael G. Ikonomou,*,‡ Paula J. Sather,‡ Erin N. Friesen,§ Dave A. Higgs,§ and Cory Dubetz‡ †

7137 Wallace Drive, Brentwood Bay, British Columbia, Canada V8M 1G9 Fisheries and Oceans Canada, Institute of Ocean Sciences, Sidney, British Columbia, Canada V8L 4B2 § DFO/UBC Centre for Aquaculture and Environmental Research, 4160 Marine Drive, West Vancouver, British Columbia, Canada V7V 1N6 ‡

bS Supporting Information ABSTRACT: Polychlorinated biphenyl (PCB) congener patterns based on full congener PCB analyses of three farmed and five wild species of salmon from coastal British Columbia, Canada are compared using principal components analysis (PCA) and the best fit linear decomposition of the observed PCB composition in terms of Aroclor 1242, 1254, and 1260 end-members. The two complementary analysis methods are used to investigate congener composition pattern differences between species, trophic levels, feeding preferences, and farmed or wild feeding regimes, with the intent of better understanding PCB processes in both salmon and salmon consumers. PCA supports classification of PCB congeners into nine groups based on a) structure activity groups (SAG) related to the bioaccumulation potential in fish-eating mammals, b) Cl number, and c) the numbers of vicinal meta- and para-H. All three factors are needed to interpret congener distributions since SAGs by themselves do not fully explain PCB distributions. Farmed salmon exhibit very similar congener patterns that overlap the PCA and Aroclor composition of their food, while wild salmon separate into two distinct groups, with chinook and “coastal” coho having higher proportions of the higher chlorinated, Aroclor 1260 type, nonmetabolizable congeners, and chum, pink, sockeye, and “remote” coho having higher proportions of the lower chlorinated, more volatile and metabolizable Aroclor 1242 type, congeners. Wild chinook have the highest PCB and toxic equivalent (TEQ) concentrations, and the highest proportions of A1254 A1260, and PCB congeners in the most refractory SAG. Because both “coastal” and “remote” coho groups are likely to be consuming prey of similar size and trophic level, the PCB delivery mechanism (e.g., atmosphere vs runoff) apparently has more influence on the salmon PCB profile than biotransformation, suggesting that the wild chinook PCB profile is determined by feeding preference. Overall, wild salmon distributions primarily relate to trophic level, feeding preferences, and longevity, while metabolism appears at most a minor factor. The new classification protocol takes better advantage of individual congener PCB analyses and provides a better framework for understanding the PCB distributions in salmon and, potentially, the movement of individual PCB congeners through marine food chains than previous classification schemes.

’ INTRODUCTION The vulnerability of fish-eating marine mammals to persistent organic pollutants such as polychlorinated biphenyls (PCBs) is well established, given the biomagnification of these contaminants in aquatic food chains and the mammal’s long lives, high trophic level, and relative inability to metabolize many persistent organic pollutants.1-3 Top predators in British Columbia (B.C.), Canada such as grizzly bear (Ursus arctos horribilis) and the fish eating subspecies of killer whale (Orcinus orca) are exceptionally susceptible to accumulation of trophically amplified PCBs due to the high proportions of fish in their diet and/or high ratios of r 2011 American Chemical Society

blubber to body weight.3-6 As a consequence, killer whales frequenting B.C. are among the most PCB contaminated cetaceans in the world and may be the slowest to recover.3,5 Heavy PCB contamination with associated health effects has also been documented in B.C./Washington state harbor seals (Phoca vitulina), and other smaller and midtrophic level species such as river otter (Lontra canadensis) and the aquatic bird American Received: November 15, 2010 Accepted: January 28, 2011 Revised: January 19, 2011 Published: February 22, 2011 2107

dx.doi.org/10.1021/es1038529 | Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology dipper (Cinclus mexicanus) are of concern due to chronically contaminated food.1,7,8 The stability and nonreactive nature of PCBs is demonstrated by the persistence of compositional profiles of Aroclor mixtures, as established by an Aroclor Calibration Method (ACM) expressing the PCB patterns in terms of the original Aroclors.9,10 While the Aroclor patterns are often still clearly distinguishable in environmental samples, particularly at lower trophic levels and close to population centers, Aroclor PCB mixtures clearly are not impervious to change. The differing molecular weights and structures of PCBs establish the physicochemical properties such as volatility and aqueous solubility (Sw), which in turn determine the key environmental parameters of water/air (H) and the octanol-water (KOW) partitioning and hence the transport, bioavailability, and propensity to bioaccumulate or to be metabolized of each PCB congener.2,11 Given the stability and health implications of PCBs, much research has been devoted to understanding the compositional profiles of PCB congeners in the environment. For salmon, an understanding of the factors affecting PCB compositional profiles can be particularly valuable for providing insight into the likely effects of PCBs on different salmon species as well as the implications for consumers of the salmon. Studies encompassing a large number of PCB congeners have been used to relate PCB congener profiles or “fingerprints” to salmon food sources and feeding preferences and the local sources of contamination in harbor seals using Principal Components Analysis (PCA).1,6,12,13 As well, studies have examined the relationship between the molecular structure of PCB congeners (i.e., their chlorine substitution pattern) and their persistence in fish and marine mammal tissues.11,14-16 Previous studies using PCA to examine PCB distributions in farmed and wild salmon have considered moderate (70 farmed and wild salmon) to large (∼600 farmed and wild salmon and 13 salmon feed) numbers of samples with 45 and 160 individual PCB congeners or congener combinations, respectively, measured.12,13 The earlier study by Carlson and Hites12 used 115 PCB variables with 31 of these variables having coeluting congeners, while the later study by Shaw et al.13 used 45 PCB variables in the PCA model (the full congener list and any coelutions are not detailed). Both studies focus on homologue groups to interpret the PCA results, with little discussion of individual congeners. This level of interpretation is sufficient to compare total PCB concentrations and to examine differences in salmon sample composition due to species and farmed vs wild origins. In contrast, the PCB method used here reports 201 individual PCBs with four coelutions (CB1 to CB3 are not quantified), and the PCA model uses 143 PCB variables with only three coelutions. The much larger number of individual PCB congeners in the present PCA model provide a stronger foundation for answering questions of biomagnification and metabolism of PCB congeners in marine food chains than any used previously by ourselves 1,6,17 or others.12,13 In this paper, PCA and ACM are used to investigate the PCB congener patterns in the five species of wild salmon, namely, chinook (Oncorhynchus tshawytscha), coho (O. kisutch), sockeye (O. nerka), chum (O. keta), and pink (O. gorbuscha) and the three species of farmed salmon, namely, Atlantic (Salmo salar), chinook, and coho present in B.C. The main intent is to compare the PCB congener profiles of each species according to structural type,11 chlorine number, and substitution pattern in order to interpret the composition similarities between wild and farmed salmon in terms of trophic levels and feeding preferences (both

ARTICLE

likely location of feeding and prey species). We further examine the proportions of metabolizable and nonmetabolizable PCB congeners between farmed and wild salmon to assess the relative retention and propensity for bioaccumulation of ingested PCBs by salmon consumers.

’ EXPERIMENTAL SECTION The 123 farmed and 96 wild salmon and eight feed samples (Table S1) were collected in four major regions of coastal B.C. between June 2003 and January 200418 and April-August 2005.19 Salmon were randomly chosen, mature fish of uniform size and weight (except for six juvenile farmed Atlantic salmon from 2005; Table S2). The left fillet (excluding skin, kidney, or residual fins) was homogenized, and subsamples (∼10 g) were analyzed using identical extraction, cleanup, QA/QC, and full congener PCB analysis methods for all samples. Percentage PCB congener compositions for Aroclors 1242, 1254, and 1260 were calculated using an ACM as described previously.9 PCA also followed previous studies,1,6,17 where samples were normalized to the concentration total to remove concentration differences between samples. The centered logratio transformation was then applied to produce a data set that was unaffected by closure.1,6,20-22 Further details are provided in the Supporting Information. ’ RESULTS AND DISCUSSION Salmon PCB and Lipid Content. Trends in the wet weight concentrations of PCBs in the different species of farmed and wild B.C. salmon in the data set (Table S1) have been discussed in detail previously.18,19 The lipid content (Table S2) is weakly, but positively, correlated to the PCB congener total for wild chinook (r2=0.516, p = 0.014) and not significantly (although generally positively) correlated for the other wild salmon (p > 0.2). Because migration-related metabolism normally produces a negative correlation, with a contaminant concentration increase with the lipid decrease,6,23 the wild salmon likely have been collected prior to any significant migration-related changes in PCB concentrations or distributions. Mature farmed Atlantic and chinook salmon samples from 200519 have substantially lower PCB concentrations than in 2003/2004 due to the substitution of less contaminated, chemically cleaned, marine fish oil and the supplementing of the fish oil with vegetable oils or animal lipids.19 Diet preferences and position on the food chain are major determinants of PCB concentration for the different wild B.C. salmon species (Table S1). Chinook and coho eat primarily smaller fish during their marine residency period and generally have the highest PCB concentrations (wet weight mean ( SE: 12600 ( 2200 and 4560 ( 490 pg/g, respectively), while pink, sockeye, and chum salmon eat more copepods, amphipods, euphuasiids, and other invertebrate species and typically have lower concentrations (1890 ( 617, 5730 ( 1020, and 2070 ( 521 pg/g).18,19,24 Other factors, such as lifespan or lipid content,18 influence concentration, and the notably elevated concentration in sockeye likely is most related to the high lipid content (Table S1). The susceptibility to contaminant bioaccumulation for the wild salmon also is complicated by the presence of two wild types of chinook: an ocean-type, which migrates to the ocean in the first months of life and stays near the coast, and a stream-type, which migrates to the ocean in their second year or 2108

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology

ARTICLE

Figure 1. PCA projections for the first two principal components (top plots) and ternary diagrams showing the Aroclor 1242, 1254, and 1260 (A1242, A1254, and A1260, respectively) content from the ACM model (bottom plots). Samples in plots on the left in a) and b) are farmed Atlantic, chinook, and coho salmon and marine fish oil based feed samples, and on the right in c) and d) are wild chinook, coho, chum, sockeye, and pink salmon samples. Ellipses illustrate the 95% confidence interval about the mean PCA projection in a) and c), and the outer four to six points of the six points of variation from ACM in b) and d), with samples classified by origin, species, and year of sampling. Group names follow Table S1 with F- and W- as abbreviations for Farmed and Wild, respectively. Positions of the 224 individual samples in PCA and ACM models are provided in Figure S2 and S3, and samples are listed by group in Table S2.

later and ranges over large portions of the North Pacific.24 Chinook from Alaska are almost entirely stream-type, while chinook from northern British Columbia are mixed stream-type and ocean-type, and those from southern British Columbia and further south are mostly ocean-type. Most other wild Pacific salmon have an offshore marine distribution, with the exception of many coho stocks that have coastal distributions. Stock identifications (Table S3) and the PCA projections (see following) have been used to separate wild coho into “coastal” and “remote” groups in Table S1, with the coastal group having much higher mean concentrations (6240 ( 538 pg/g) than the remote group (1910 ( 213 pg/g; Table S2). PCB Sourcing in Salmon by ACM and PCA. ACM and PCA provide two different, but complementary, approaches to interpreting the PCB composition in salmon, with the two methods providing a valuable check on each other. The ACM9 describes the PCB composition of each sample as a linear combination of

the three main Aroclor end-members (A1242, A1254, and A1260; Figure 1, S1; Table S2). PCA provides the more robust method of sample comparison because it lets the samples describe their own framework without imposing any predetermined structure (Aroclor type or otherwise) to the data.6,25,26 The PCA data preprocessing protocol (normalization followed by the centered logratio transformation) yields a data set where the average concentration and concentration total are identical for every sample and effectively removes interferences from sample concentration differences and spurious variable projections due to closure.20,26 Closure most often manifests in the highest concentration PCA variables, where variables that are strongly correlated using conventional analysis do not covary in the PCA model,21,22 making the model unsuitable for resolving patterns in PCB congeners. The ACM residual values, which provide a measure of how closely the PCB patterns align with Aroclors, are similar for most 2109

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology salmon and feed samples (Table S2) and very low in comparison to many other sample types,9,10 indicating that both sample types have relatively unweathered PCB compositions. Very strong correlations are observed between the t1 sample scores (the samples projections in PC1; Figure 1a, c, S2) for the 224 salmon and feed samples and the calculated Aroclor endmember fractions for each sample from ACM (Table S2), with a very strong negative correlation between t1 and A1260 (r2=0.801, p = 7.2  10-80) and positive correlation with A1242 (r2=0.916, p = 1.1  10-121). The correlation between t1 and A1254 also is significant (r2=0.146, p = 3.6  10-9), but the positive correlation with t2 is much stronger (r2=0.522, p = 2.0  10-37). The correlations between the Aroclor fractions and t3 to t5 are either not significant or marginally significant (the highest correlation is r2=0.062, p = 1.6  10-4 between t3 and A1254). These significant correlations provide a strong foundation to use the PCA and ACM together to interpret congener composition differences between species, trophic levels, feeding preferences, and farmed or wild feeding regimes. The farmed salmon from the 2003/2004 sampling18 form three compact, closely adjoining clusters in both the PCA model and the ACM (Figure 1a, b, S2a) with two clusters overlapping for farmed Atlantic plus coho and a separate cluster for farmed chinook samples. The farmed salmon thus exhibit no difference in PCB composition by sampling date or fish farm sampled, with essentially no composition difference between Atlantic and coho species and only a small difference in PCB composition between these two species and chinook. The 2003/2004 marine fish oil based feed samples project on top of the largest cluster of farmed samples (Figure S2a), demonstrating the similarity in PCB composition between farmed samples and their food and indicating that the feed is the major source of PCBs for farmed salmon (cf., ref 12). The positions of the farmed salmon and feed samples in the ACM ternary diagram indicate a uniform PCB composition with an A1254 and A1260 dominance and only a minor ( 0.2). Harbor seals sampled along a gradient away from the heavily industrialized Puget Sound, through the moderately industrialized Strait of Georgia, and into the more remote Queen Charlotte Strait, also reveal an increasingly “light” PCB pattern, which correlates with log H.1 Accordingly, the significant correlations between the Ucluelet atmospheric samples and A1242 provide strong support for the use of A1242 as a surrogate for the PCB composition end member for the marine aquatic environment (i.e., water, particulate, plankton) in the North Pacific and remote areas of the B.C. coast. Hence, the wild chum, sockeye, pink, and “remote” coho salmon with high proportions of A1242 type congeners, as with harbor seals from remote areas with high proportions of lighter PCB congeners,1 must have gained most of their PCB contaminants in locations in offshore or remote locations where atmospheric transport is the major PCB source. A feeding preference for prey at a lower trophic level by these salmon also is indicated given the lack of any substantive PCB composition change to the less chlorinated congeners during biomagnification. In contrast, chinook and “coastal” coho salmon with high proportions of A1254 and A1260, as with harbor seals from urbanized areas with high proportions of heavier PCB congeners,1 must have gained their PCBs in locations where direct input of low volatility PCBs by 2110

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology rivers or stormwater transport is a major source.28 For these salmon feeding in coastal locations adjacent to urbanized or industrialized areas is suggested, although the shift in PCB profile to more chlorinated congeners also could be partially due to the salmon feeding at higher trophic levels. PCB Congener Groups by Substitution and Bioaccumulation Patterns. Interpretation of PCB congener projections by chlorine (Cl) number, vicinal hydrogen (H) pairs, and structure activity groups (SAGs; see the Supporting Information) based on molecular structure and the bioaccumulation potential of PCBs in fish-eating mammals11 supports classification of PCBs into six well-defined and three diffuse groups (Figure 2a, S3; Table S4). The large number of individual PCB congeners in the present data set greatly facilitates this classification. The six well-defined groups have the smallest 95% confidence ellipses for the PCA variable projections of the PCBs in the group, while the three diffuse groups show more variation in congener projections and/ or have small congener numbers with a larger t-value (shown with dashed lines in Figure 2a). To relate the PCA projections for variables and samples in Figure 1, one looks at the distance and direction from axis center in the samples and variables plots (t1 vs t2 and p1 vs p2, respectively, for PC1 and PC2). The high Cl number PCBs in SAG I (Cl number 5-10; no vicinal H atoms) and II (Cl 5-8; vicinal ortho- and meta-H atoms (o-m-H) with g2 ortho-Cl atoms) project together in compact groups (relative areas of the 95% confidence ellipses are 1.0 and 4.0, respectively) on the left side of the PCA variables plot, with no apparent trends by Cl number or, for SAG II, the numbers of vicinal o-m-H pairs (Figure 2a, S3). On the other side of the PCA variables plot, congeners with 2-4 Cl substituents in SAG III (om-H and 2 ortho-Cl) also show a strong relationship between variable projections and Cl number, but a better separation is obtained by grouping SAG V congeners by vicinal m-p-H, with m-p-H = 1 forming a group (ellipse area 6.4) on the left side, overlapping SAG II. For the diffuse groups, PCBs in SAG V with m-p-H > 1 (ellipse area 17) project on the right with the Cl number 2-4 groups, while PCBs in SAG III with 5-6 Cl substituents (ellipse area 30.3) project in the upper half of the variable plot. The few PCBs in SAG IV (Cl 5-6; m-p-H with e2 ortho-Cl; n = 8 congeners, area 29.5) also are quite variable in their PCA projections but exhibit no relationship to Cl number or the numbers of vicinal m-p-H pairs which define the group. Note that the m-p-H division for SAG V introduces just two exceptions to a separation by Cl number (Table S4), with the m-p-H = 1 group having Cl=6-8 for all except PCB 103 (Cl=5) and m-p-H > 1 having Cl=4-5 for all except PCB 136 (Cl=6). With these classifications, PCB variable loadings in p1 primarily reflect a separation between refractory PCBs in SAG I, II, and V m-p-H = 1 (left side with n = 34, 21, and 11 congeners, respectively; Figure 2a) and metabolizable PCBs in SAG III Cl 2-4, V m-p-H > 1 and VI Cl 2-4 (right side, n = 9, 6, and 37). Increases in p2 primarily reflect increased proportions of PCBs in SAG III Cl 5-6, IV and VI Cl 5-6 (upper center, n = 6, 8, and 12). The most scattered group is SAG III Cl 5-6, which is notable because it constitutes only PCBs with one ortho-Cl substituent that have a TEF (toxic equivalent factor), includes PCBs 105,

ARTICLE

Figure 2. a) PCA variables plot with PCBs classified by SAG, Cl number, and H substitution patterns (Table S4). The 95% confidence interval is indicated by an ellipse for each group, with the three diffuse groups shown with dashed lines. Group abbreviations use Roman numerals I, II, etc. for SAG I, SAG II, etc., an Arabic number range to show Cl number subdivisions, and m-p-H values to indicate vicinal metaand para-H substituents. PCB congeners with a TEF contributing to the TEQ are indicated by the congener number. b) Concentration proportion of PCBs by SAG and PCA group for salmon feed, farmed salmon and wild salmon. The refractory, metabolizable, and variable PCB groups are separated slightly to emphasize the group separations. c) PCA variables plot showing the 16 PCBs identified as refractory (in red) and the 23 PCBs in the PCA model observed to be biotransformed (in green) in rainbow trout.15,16. 2111

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology 114, 118, 123, 156, and 157. The largest outliers in the group are PCB 156, which projects with the SAG I and II congeners in both the salmon PCA model shown here and marine mammals,11 while PCB 157 projects with the SAG III congeners with low Cl number in salmon. PCBs 156 and 157 both have one vicinal o-mH pair with one ortho-Cl (allowing a planar configuration), which fish would be unable to degrade due to the absence of necessary CYP1A enzymes.11,15 The predicted torsional angle between the rings 29 also is equivalent (59.0° and 59.2° for PCBs 156 and 157, respectively), which suggests that the two PCBs should not differ in their conformations or propensity to be degraded by CYP1A enzymes. It may be that the more symmetrical Cl substitution pattern of PCB 157 (3 Cl per ring vs 4 Cl and 2 Cl for PCB 156) facilitates its preferential accumulation at lower trophic levels. Application of PCB Congener Groups to Salmon. With the PCB congener groups established, it becomes apparent that the projection on the left in the PCA samples plot and the A1254 and A1260 dominance for the 2003/2004 farmed salmon samples (Figure 1a, b) corresponds to a predominance in the more highly chlorinated SAG I, II, and V m-p-H = 1 PCBs in the variables plot (Figure 2a). This indicates that the 2003/2004 farmed salmon fed traditional diets based on marine fish oils have high proportions of PCBs in the nonmetabolizable SAG groups I and II (Figure 2b). Furthermore, the small shift to the lower left and the higher content of A1260 for the farmed chinook relative to the Atlantic, coho and most feed samples in both ACM and PCA indicates a higher content of the heavier SAG I and II PCBs for the chinook. The shift of the 2005 farmed Atlantic and chinook to the upper right of the main group of farmed salmon in the PCA model and the higher content of A1242 type congeners both indicate a higher proportion of PCB congeners in SAG III, V, and VI (particularly in the lower chlorinated subgroups of each SAG) for salmon fed a contaminant-reduced diet than salmon fed a traditional marine fish oil diet (Figure 1a, 2a). For the wild chinook, the projection in the upper left in the PCA and the higher content of A1254 than the farmed salmon (Figure 1c, d) indicate a higher proportion of PCBs in SAG III Cl 5-6, IV and VI Cl 5-6 than any of the farmed salmon (Figure 2a, b). The “coastal” wild coho salmon project a little closer to the farmed salmon and have a slightly higher content of A1260 than the chinook, and they have high proportions of PCBs in SAG II as well as SAG III Cl 5-6, IV and VI Cl 5-6. In contrast, the projection of the wild chum, sockeye, pink, and “remote” coho salmon on the right side in the PCA, with their high proportions of A1242 type congeners, correspond to a predominance of PCBs in the lower chlorinated subgroups SAG III Cl 2-4, V m-p-H > 1 and VI Cl 2-4 (Figure 2a, b). Hence, it appears likely that the “remote” coho are feeding primarily in locations distant from direct sources of PCB contamination, while “coastal” coho demonstrate an evident preference for coastal areas receiving direct PCB inputs. Given that both groups of coho are likely to be consuming prey of similar size and trophic level, this suggests that the PCB delivery mechanism (e.g., atmosphere vs runoff) has more influence on the fish PCB profile than composition changes through the food chain and that the wild chinook PCB profile is determined more by feeding preference than metabolism. The close projection of farmed chinook salmon from 2005 and “coastal” coho in the PCA model with wild chinook samples (Figure 1b, c, S2) suggests that the majority of wild chinook samples are likely of the type which stays and feeds near the coast, close to sources of the more chlorinated PCB congeners. All wild chinook with

ARTICLE

available DNA profiles for stock identification have rivers of origin from southern B.C. (Fraser River) and further south (Table S3), and this portion of the Pacific coast generally produces chinook that stay near the coast.24 Salmon Trophic Levels and TEQ. The relative trophic levels of the salmon species in this study are related to the eating habits (both location of feeding and prey species) and the maximum marine residency time of the species.24 Chum salmon have the longest marine residency maximum (4-6 years), followed by chinook (4-5 years), sockeye (3-4 years), coho (3 years), and pink salmon (2 years). If differences in feeding habits and lifespan are considered, the five wild Pacific salmon species can be ranked in order of their susceptibility to contaminant bioaccumulation, with chinook and coho salmon being more susceptible than sockeye, pink, and chum salmon. The likelihood of potential health impacts on consumers of salmon is related to both the PCB concentration and the proportion of PCB congeners with bioaccumulation potential.1,5,11 In the PCA model log KOW30 is significantly correlated with p1 (r2=0.676, p = 8.0  10-37; v = 141) but not p2 (r2=0.010, p = 0.22). Absorption of PCBs into fish tissue from food is an important bioaccumulation mechanism for log KOW > ∼4 (which includes all PCBs in the PCA model), and PCBs with log KOW ∼7 have the longest half-life in fish tissue.15,31 Salmonid species also more easily eliminate the PCBs with lower log KOW values.31 Fish are generally considered to have poor capability to biotransform PCBs,11 particularly at the colder water temperatures typical of the North Pacific Ocean.16 Rainbow trout can biotransform some PCBs in SAG IV and V but not PCBs in SAG I and II and, with the exception of PCB 77, most PCBs in SAG III15,16 and are responsive to CYP2B-like induction by PCBs.16 The 16 PCBs identified as refractory and the 23 PCBs in the PCA model observed to be biotransformed by rainbow trout by Buckman et al.15,16 scatter across the PCA plot (Figure 2c) suggesting that metabolism by fish does not play a major role in PCB separations in the PCA. The rate of PCB biotransformation increases with increasing temperature, with little biotransformation observed at 8 °C.15,16 Since typical temperatures in the North Pacific and around the B.C. salmon pens would be at or below 8 °C for much of the year, the low water temperature is likely to limit biotransformation. Given the likelihood that metabolism is a minor factor and that uptake and depuration relationships with log KOW are similar in all fish, the predominance of refractory higher Cl number congeners in wild chinook and “coastal” coho may be related to both their higher trophic level (with predominately piscivorous diets) and longevity, resulting in the accumulation of longer half-life PCBs relative to more easily depurated congeners. For wild chinook salmon these processes manifest as higher PCB and toxic equivalent (TEQ) concentrations (Table S2; Figure 3), higher proportions of A1254 and A1260 (Figure 1d), and higher proportions of refractory PCB congeners in SAG I and II (Figure 2b) as compared to other wild salmon species. The two groups of coho salmon, which are evident from both the PCA and ACM (Figure 1c, d), clearly illustrate the importance of feeding location, with ∼3 fold lower PCB and TEQ concentrations and much lower proportions of refractory SAG I and II congeners in the “remote” coho. For wild chum, sockeye, and pink salmon, a low trophic level diet and a likely preference for feeding primarily in offshore/remote areas affords the lowest PCB concentrations with the highest proportions of lower chlorinated PCB congeners in SAG III Cl 2-4, V m-p-H > 1 and VI Cl 2-4. 2112

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology

Figure 3. a) Mean TEQ concentration and b) TEF proportion by PCA group as in Figure 2 for salmon feed, farmed salmon and wild salmon showing the contribution of each PCB congener in the PCA model with a defined TEF. PCBs 81, 126, and 169 have not been included in the PCA and this figure due to their low concentrations and high proportions of undetectable values.

Farmed salmon fed a traditional diet of fish meal and oil derived primarily from small pelagic fish can be considered to mimic the higher trophic levels of small fish consumers such as harbor seals.32 The farmed salmon appear different from all groups of wild salmon yet more closely resemble the higher trophic level wild chinook and the “coastal” coho than the lower trophic level chum, sockeye, and pink salmon. The PCB congener composition and SAG differences between the 2003/2004 and 2005 farmed Atlantic and chinook samples (Figure 1a) suggest that attaining both lower PCB concentrations and lower proportions of the SAG I and II PCB congeners are achievable for farmed salmon if the PCB content of the diet is controlled (Table S1; Figure 2b, 3). There is a strong correlation between PCB TEQ and total PCB concentrations both by sample group in Figure 3 for the TEF congeners and the PCBs in the PCA model (r2=0.994, p = 1.72  10-12, v = 10) and by individual sample for all TEF congeners and all measured PCBs (r2=0.776, p = 5.06  10-74, v = 222). This correlation indicates that there is a direct relationship between PCB concentration and toxicity for the wild and farmed salmon. Nevertheless, there are systematic trends between species in the proportion of individual TEF congeners contributing to the TEQ that may affect the potential bioavailability of the toxic congeners to wildlife consumers of salmon. TEF congeners projecting close to the x-axis in Figure 2a show the clearest trends in

ARTICLE

the proportion of individual TEF congeners to the TEQ total in Figure 3, with PCBs 77 and 157 on the far right in Figure 2a contributing less to the TEQ and PCBs 156, 167, and 189 on the far left contributing more to the TEQ as groups progress from wild chum, sockeye, pink, and “remote” coho, to wild “coastal” coho and chinook, to farmed salmon (Figure 3). TEF congeners projecting from the center to upper center along the y-axis in Figure 2a include PCBs 105, 114, and 118, which make the largest contribution to the TEQ and have some of the strongest loadings on p2, and PCB 123. These four PCBs show less difference between species with PCBs 114, 118, and 123 making proportionally the largest contribution to wild chum, sockeye, pink, and “remote” coho and PCB 105 making a marginally greater contribution to farmed salmon than any wild salmon. PCA projections of wild salmon, particularly wild chinook and coho, show much greater variation on t2 than farmed salmon (Figure S2), hence, the projection of PCBs 105, 114, 118, and 123 along the y-axis in Figure 1a appears to have more to do with a larger variation in composition and TEQ for wild chinook and coho than other salmon. Two top predators in B.C. illustrate how the TEQ concentrations and proportions of refractory congeners can affect salmon consumers. The diet of grizzly bears (Ursus arctos horribilis) in remote, coastal regions of B.C. is dominated by pink and sockeye salmon.17 Pink and sockeye have the lowest PCB concentrations (Table S1) with the highest proportions of lower chlorinated PCB congeners in SAG III Cl 2-4, V m-p-H > 1 and VI Cl 2-4 (Figure 2b), reflecting fish that feed primarily in offshore/remote areas at bottom of the food chain. The diet of salmon-eating, northeastern Pacific, resident killer whales (Orcinus orca) is made up of ∼70% chinook salmon with small amounts of coho.3,6 Chinook and coho have the highest PCB concentrations (Table S1) with the highest proportions of highly chlorinated SAG I, II, and V m-p-H = 1 PCBs (Figure 2b), reflecting their longer lifetimes and higher trophic level. The result is the bears are consuming salmon with lower amounts of PCBs and lower proportions of refractory congeners than the killer whales, and the resident killer whales would be most affected by the higher PCB and TEQ concentrations (Table S1; Figure 3) and higher proportions of refractory PCBs (Figure 2b) from chinook.5,6 Many other factors can affect the relative impact of PCBs on a consumer including, for example, that the bears also consume vegetation, which assists with the removal of PCB contaminants.4 Nevertheless, this scenario agrees with other more detailed studies indicating that the resident killer whales are most at risk from chronic PCB exposure.3,5,6 The expanded PCB classification protocol based on PCB congener substitution and fisheating mammal bioaccumulation patterns developed here takes better advantage of individual congener PCB analyses and provides a better framework for understanding the movement of individual PCB congeners through marine food chains than previous classification schemes. The expectation is that a better understanding of the movement of PCBs in food webs will ultimately allow better protection of salmon consuming wildlife. Overall, the congener groups based on SAG, Cl number, and H substitution patterns (Figure 2a) provide an easily visualized framework to interpret the PCA projections for the large number of congeners in the PCA model (Figure S3) in terms of concentrations and proportions of metabolizable and refractory PCB congener groups and the content of refractory and TEQ congeners for each fish species. All three factors are needed to interpret congener distributions since SAGs by themselves do 2113

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology not fully explain PCB distributions. While the classifications have been developed both to help understand PCB distributions in salmon and assist with ongoing work aimed at understanding PCB processes in marine mammal food chains of salmon consumers, the PCB groups also could have value for understanding the health risks associated with the consumption of farmed salmon by humans.12,13,18,33

’ ASSOCIATED CONTENT

bS

Supporting Information. Details are provided regarding the sample collection and full congener PCB analysis along with Aroclor ternary diagrams and PCA samples plots showing all samples by sample type and PCA variable plots showing all PCBs by SAG group. Tables provide salmon size and weight, PCB concentrations, TEQ and Aroclor proportions for each sample and PCA congeners by SAG with chlorine number, ortho-Cl number, number of o-m-H and m-p-H pairs, and congeners with >5% abundance in Aroclor 1242, 1254, and 1260. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected] (M.B.Y.); Michael.Ikonomou@ dfo-mpo.gc.ca (M.G.I.).

’ ACKNOWLEDGMENT The development of the PCB congener groups was inspired by collaborative work with Peter Ross, Jennie Christensen, and Donna Cullon on the application of PCA to understand PCB congener distributions in marine mammals. We thank Janice Oakes from DFO for her help with project planning and sample collection and coordination. We gratefully acknowledge the assistance of all the DFO staff, the numerous personnel at participating B.C. salmon farms as well as the DFO area managers for their assistance with sample collection. The assistance of all the chemists and support staff of the DFO analytical laboratories at IOS who processed and analyzed all of the samples is much appreciated. Financial support was provided by AquaNet, the DFO-ACRDP program, the B.C. Salmon Farmers Association, the B.C. Science Council, the former federal Office of the Commissioner for Aquaculture Development, and Stolt Sea Farm Inc. ’ REFERENCES (1) Ross, P. S.; Jeffries, S. J.; Yunker, M. B.; Addison, R. F.; Ikonomou, M. G.; Calambokidis, J. C. Harbor seals (Phoca vitulina) in British Columbia, Canada, and Washington state, USA, reveal a combination of local and global polychlorinated biphenyl, dioxin and furan signals. Environ. Toxicol. Chem. 2004, 23, 157–165. (2) Macdonald, R. W.; Mackay, D.; Hickie, B. Contaminant amplification in the environment. Environ. Sci. Technol. 2002, 36, 456A–462A. (3) Ross, P. S.; Ellis, G. M.; Ikonomou, M. G.; Barrett-Lennard, L. G.; Addison, R. F. High PCB concentrations in free-ranging Pacific killer whales, Orcinus orca: Effects of age, sex and dietary preference. Mar. Pollut. Bull. 2000, 40, 504–515. (4) Christensen, J. R.; Letcher, R. J.; Ross, P. S. Persistent or not persistent? Polychlorinated biphenyls are readily depurated by grizzly bears (Ursus arctos horribilis). Environ. Toxicol. Chem. 2009, 28, 2206–2215.

ARTICLE

(5) Hickie, B. E.; Ross, P. S.; Macdonald, R. W.; Ford, J. K. Killer Whales (Orcinus orca) face protracted health risks associated with lifetime exposure to PCBs. Environ. Sci. Technol. 2007, 41, 6613–6619. (6) Cullon, D. L.; Yunker, M. B.; Alleyne, C.; Dangerfield, N. J.; O’Neill, S.; Whiticar, M. J.; Ross, P. S. Persistent organic pollutants in chinook salmon (Oncorhynchus tshawytscha): Implications for resident killer whales of British Columbia and adjacent waters. Environ. Toxicol. Chem. 2009, 28, 148–161. (7) Elliott, J. E.; Guertin, D. A.; Balke, J. M. Chlorinated hydrocarbon contaminants in feces of river otters from the southern Pacific coast of Canada, 1998-2004. Sci. Total Environ. 2008, 397, 58–71. (8) Morrissey, C. A.; Elliott, J. E.; Ormerod, S. J. Local to continental influences on nutrient and contaminant sources to river birds. Environ. Sci. Technol. 2010, 44, 1860–1867. (9) Sather, P. J.; Ikonomou, M. G.; Addison, R. F.; He, T. H.; Ross, P. S.; Fowler, B. R. Similarity of an Aroclor-based and a full congenerbased method in determining total PCBs and a modeling approach to estimate Aroclor speciation from congener-specific PCB data. Environ. Sci. Technol. 2001, 35, 4874–4880. (10) Sather, P. J.; Newman, J. W.; Ikonomou, M. G. Congener-based Aroclor quantification and speciation techniques: A comparison of the strengths, weaknesses, and proper use of two alternative approaches. Environ. Sci. Technol. 2003, 37, 5678–5686. (11) Boon, J. P.; van der Meer, J.; Allchin, C. R.; Law, R. J.; Klungsøyr, J.; Leonards, P. E.; Spliid, H.; Storr-Hansen, E.; Mckenzie, C. H.; Wells, D. E. Concentration-dependent changes of PCB patterns in fish-eating mammals: Structural evidence for induction of Cytochrome P450. Arch. Environ. Contam. Toxicol. 1997, 33, 298–311. (12) Carlson, D. L.; Hites, R. A. Polychlorinated biphenyls in salmon and salmon feed: Global differences and bioaccumulation. Environ. Sci. Technol. 2005, 39, 7389–7395. (13) Shaw, S. D.; Brenner, D.; Berger, M. L.; Carpenter, D. O.; Hong, C.-S.; Kannan, K. PCBs, PCDD/Fs, and organochlorine pesticides in farmed Atlantic salmon from Maine, eastern Canada, and Norway, and wild salmon from Alaska. Environ. Sci. Technol. 2006, 40, 5347–5354. (14) Storr-Hansen, E.; Spliid, H.; Boon, J. P. Patterns of chlorinated biphenyl congeners in Harbor seals (Phoca vitulina) and in their food: Statistical analysis. Arch. Environ. Contam. Toxicol. 1995, 28, 48–54. (15) Buckman, A. H.; Wong, C. S.; Chow, E. A.; Brown, S. B.; Solomon, K. R.; Fisk, A. T. Biotransformation of polychlorinated biphenyls (PCBs) and bioformation of hydroxylated PCBs in fish. Aquat. Toxicol. 2006, 78, 176–185. (16) Buckman, A. H.; Brown, S. B.; Small, J.; Muir, D. C.; Parrott, J.; Solomon, K. R.; Fisk, A. T. Role of temperature and enzyme induction in the biotransformation of polychlorinated biphenyls and bioformation of hydroxylated polychlorinated biphenyls by rainbow trout (Oncorhynchus mykiss). Environ. Sci. Technol. 2007, 41, 3856–3863. (17) Christensen, J. R.; MacDuffee, M.; Yunker, M. B.; Ross, P. S. Hibernation-associated changes in persistent organic pollutant (POP) levels and patterns in British Columbia grizzly bears (Ursus arctos horribilis). Environ. Sci. Technol. 2007, 41, 1834–1840. (18) Ikonomou, M. G.; Higgs, D. A.; Gibbs, M.; Oakes, J.; Skura, B.; McKinley, S.; Balfry, S. K.; Jones, S.; Withler, R.; Dubetz, C. Flesh quality of market-size farmed and wild British Columbia salmon. Environ. Sci. Technol. 2007, 41, 437–443. (19) Friesen, E. N.; Ikonomou, M. G.; Higgs, D. A.; Ang, K. P.; Dubetz, C. Use of terrestrial based lipids in aquaculture feeds and the effects on flesh organohalogen and fatty acid concentrations in farmed Atlantic salmon. Environ. Sci. Technol. 2008, 42, 3519–3523. (20) Bonn, B. A. Polychlorinated dibenzo-p-dioxin and dibenzofuran concentration profiles in sediment and fish tissue of the Willamette basin, Oregon. Environ. Sci. Technol. 1998, 32, 729–735. (21) Johansson, E.; Wold, S.; Sj€odin, K. Minimizing effects of closure on analytical data. Anal. Chem. 1984, 56, 1685–1688. (22) Yunker, M. B.; Macdonald, R. W.; Veltkamp, D. J.; Cretney, W. J. Terrestrial and marine biomarkers in a seasonally ice-covered 2114

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115

Environmental Science & Technology

ARTICLE

Arctic estuary — Integration of multivariate and biomarker approaches. Mar. Chem. 1995, 49, 1–50. (23) Kelly, B. C.; Gray, S. L.; Ikonomou, M. G.; Macdonald, J. S.; Bandiera, S. M.; Hrycay, E. G. Lipid reserve dynamics and magnification of persistent organic pollutants in spawning sockeye salmon (Oncorhynchus nerka) from the Fraser River, British Columbia. Environ. Sci. Technol. 2007, 41, 3083–3089. (24) Higgs, D. A.; Macdonald, J. S.; Levings, D. D.; Dosanjh, B. C. Nutrition and feeding habits in relation to life history stage. In Physiological Ecology of Pacific Salmon; Groot, C., Margolis, L., Clarke, W. C., Eds.; UBC Press: Vancouver, B.C., 1995; pp 159-315. (25) Meglen, R. R. Examining large databases: a chemometric approach using Principal Components Analysis. Mar. Chem. 1992, 39, 217–237. (26) Yunker, M. B.; Belicka, L. L.; Harvey, H. R.; Macdonald, R. W. Tracing the inputs and fate of marine and terrigenous organic matter in Arctic Ocean sediments: A multivariate analysis of lipid biomarkers. Deep-Sea Res., Part II 2005, 52, 3478–3508. (27) No€el, M.; Dangerfield, N.; Hourston, R. A.; Belzer, W.; Shaw, P.; Yunker, M. B.; Ross, P. S. Do trans-Pacific air masses deliver PBDEs to coastal British Columbia, Canada? Environ. Pollut. 2009, 157, 3404–3412. (28) Johannessen, S. C.; Macdonald, R. W.; Wright, C. A.; Burd, B.; Shaw, D. P.; van Roodselaar, A. Joined by geochemistry, divided by history: PCBs and PBDEs in Strait of Georgia sediments. Mar. Environ. Res. 2008, 66, S112–S120. (29) Dorofeeva, O. V.; Novikov, V. P.; Moiseeva, N. F.; Yungman, V. S. Density functional theory study of conformations, barriers to internal rotations and torsional potentials of polychlorinated biphenyls. J. Mol. Struct. 2003, 637, 137–153. (30) Hawker, D. W.; Connell, D. W. Octanol-water partition coefficients of polychlorinated biphenyl congeners. Environ. Sci. Technol. 1988, 22, 382–387. (31) Fisk, A. T.; Norstrom, R. J.; Cymbalisty, C. D.; Muir, D. G. Dietary accumulation and depuration of hydrophobic organochlorines: Bioaccumulation parameters and their relationship with the octanol/ water partition coefficient. Environ. Toxicol. Chem. 1998, 17, 951–961. (32) Cullon, D. L.; Jeffries, S. J.; Ross, P. S. Persistent organic pollutants in the diet of harbor seals (Phoca vitulina) inhabiting Puget Sound, Washington (USA), and the strait of Georgia, British Columbia (Canada): A food basket approach. Environ. Toxicol. Chem. 2005, 24, 2562–2572. (33) Huang, X.; Hites, R. A.; Foran, J. A.; Hamilton, C.; Knuth, B. A.; Schwager, S. J.; Carpenter, D. O. Consumption advisories for salmon based on risk of cancer and noncancer health effects. Environ. Res. 2006, 101, 263–274.

2115

dx.doi.org/10.1021/es1038529 |Environ. Sci. Technol. 2011, 45, 2107–2115