(Uria aalge) from the Baltic Sea - ACS Publications - American

Multivariate Data Analyses of Chlorinated and Brominated Contaminants and Biological Characteristics in Adult Guillemot (Uria aalge) from the Baltic S...
0 downloads 0 Views 205KB Size
Environ. Sci. Technol. 2005, 39, 8630-8637

Multivariate Data Analyses of Chlorinated and Brominated Contaminants and Biological Characteristics in Adult Guillemot (Uria aalge) from the Baltic Sea K A T R I N L U N D S T E D T - E N K E L , * ,†,‡ ANNA-KARIN JOHANSSON,‡ MATS TYSKLIND,§ LILLEMOR ASPLUND,| KERSTIN NYLUND,| M A T S O L S S O N , ‡,| A N D J A N O ¨ RBERG† Environmental Toxicology, Department of Physiology and Developmental Biology, Evolutionary Biology Centre, Uppsala University, Norbyva¨gen 18A, SE-752 36, Sweden, Contaminant Research Group, Swedish Museum of Natural History, P.O. Box 50007, SE-104 05 Stockholm, Sweden, Environmental Chemistry, Department of Chemistry, Umea˚ University, SE-901 87 Umea˚, Sweden, and Institute of Applied Environmental Research, Stockholm University, SE-106 91 Stockholm, Sweden

Adult guillemot (Uria aalge) birds, 10 females and 10 males, drowned in trawl nets near Stora Karlso¨ in the Baltic Sea, were collected in 2000. Several of the animals’ biological characteristics were recorded. The birds’ pectoral muscles were individually analyzed for their concentrations of organochlorines (OCs) and brominated flame retardants (BFRs), dichlorodiphenyltrichloroethanes (DDTs), polychlorinated biphenyls (PCBs), hexachlorocyclohexanes, trans-nonachlor, hexachlorobenzene, hexabromocyclododecane (HBCD), and polybrominated diphenyl ethers (PBDEs). The dominating contaminant was p,p′dichlorodiphenyldichloroethylene (p,p′-DDE) with a geometric mean concentration of 12 900 ng/g lipid weight (lw). The concentration of ΣPBDE (80 ng/g lw) was similar to that of HBCD (65 ng/g lw). The total concentration of all OCs was approximately 150 times higher than that of all BFRs. For the statistical evaluation of the data, we used multivariate analysis techniques such as principal components analysis, partial least-squares (PLS) regression, and PLS discriminant analyses. No differences between the two sexes were found, either in contaminant concentrations or in biological characteristics. We found that some biological characteristics covaried with the concentrations of several OCs and BFRs, e.g., a negative correlation between liver weight and concentration of contaminants. The concentrations of most OCs but not of BFRs showed a decrease with increasing lipid content. Further, a PLS model with OCs as X and BFRs as Y showed that the contaminants formed two groups, each with distinctive correlation patterns. The PLS model could be used to predict with varying accuracy * Corresponding author phone: +46-18-4716498; fax: +46-18518843; e-mail: [email protected]. † Uppsala University ‡ Swedish Museum of Natural History § Umeå University. | Stockholm University. 8630

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005

the concentration of BFRs in the individual muscles from their concentration of OCs.

Introduction Guillemots (Uria aalge) are year-round residents in the Baltic Sea, which feed almost exclusively on pelagic clupeids, sprat (Sprattus sprattus) and herring (Clupea harengus) (1-3), two fatty fish species known to be contaminated with organochlorines (OCs) and brominated flame retardants (BFRs) (4-6). From around 4-5 years of age, guillemot females lay one large (∼110 g) egg in the spring each year (4, 7), and guillemot eggs have been analyzed for their contaminant contents since 1969 within the Swedish monitoring program (8). In one study, biomagnification factors (BMFs) between herring (muscle) and guillemot (egg) were calculated, and there it was found that some OCs biomagnify to a high degree, e.g., p,p′-dichlorodiphenyldichloroethylene (p,p′-DDE) and 2,3′,4,4′,5-pentachlorinated biphenyl (CB118), while some OCs do not biomagnify, e.g., 2,2′,4,5,5′-pentachlorinated biphenyl (CB101) (9). The broad span in BMFs reflects variations both in chemical characteristics of the contaminants and in biological factors. Chemical characteristics of the contaminants essential for biomagnification, e.g., bioavailability and persistence, are determined by a number of physicochemical properties such as molecular size, structure, presence of reactive groups, lipophilicity (Kow), and water solubility (10-14). Biological factors important for biomagnification are related to the animals, the predator and its prey, e.g., feeding habits and habitats, lipid characteristics and content, sex and age, and the animals’ capacity to biotransform/excrete contaminants (15-18). Differences between male and female contaminant concentrations have been shown for a number of species where the female allocates a large amount of lipids for the production of eggs or progeny or for transfer to the offspring during lactation (15, 19, 20). The first aim of this work was to investigate if there were any differences between guillemot males and females in their contaminant patterns and/or in their biological characteristics. Exposure to environmental stressors, i.e., contaminants that are potentially detrimental to animal well-being, might lead to changes in biological characteristics, e.g., body weight and plumage condition. Such effects caused by pollutants in birds have recently been reviewed in a number of papers (21-24). The second aim of this work was to investigate relationships between guillemot biological characteristics (e.g., liver weight and lipid content), hereafter called biological variables, and muscle residue levels of contaminants. Earlier work has shown that concentrations of some OCs covary. For instance, CB153 can be used to calculate/predict (henceforth called predict), for example, CB180 concentrations in many matrixes (4). The third aim of this work was to investigate relationships between concentrations of OCs and BFRs and if OC concentrations can be used to predict the animals’ BFR concentrations. Investigations measuring large numbers of contaminants in many individuals generate large data sets. In general, these investigations use classical statistics to present levels as means together with some measure of variation and/or regression/ correlation analysis to study the relationship between a few selected variables. As many variables in nature are dependent on each other, we address our three main questions by applying multivariate data analysis techniques, i.e., principal 10.1021/es051118o CCC: $30.25

 2005 American Chemical Society Published on Web 10/12/2005

components analysis (PCA), partial least-squares (PLS) regression, and PLS discriminant analyses (PLS-DA), to extract additional and essential information from our data that is not possible to obtain using classical statistics.

Materials and Methods Guillemot. Adult guillemot (Uria aalge) birds (n ) 20) belonging to the colony at the Karlso¨ islands (coordinates 57°17′ N, 17°58′ E), situated ∼6 km to the west of the Swedish island of Gotland in the Baltic Sea, were used. All of the birds collected died accidentally in trawl nets south of the islands of Karlso¨ in the middle of April 2000. Three of the collected birds were ringed, one female, 7 years of age and two males, 27 and 29 years of age. All nonringed birds were of unknown age, though all of the birds had adult plumage. The birds died ∼1 month before egg-laying, and all females had small ∼4.8 mm (3.6-6.4 mm) (geometric mean (GM) with 95% confidence interval (CI)) oocytes in their ovaries. Further, from the anatomy of the oviduct, a majority of the females (8 of 10) had lain at least one egg and were thus a minimum of 5 years old (7). Directly after landing, the bird specimens were individually frozen (-20° C) in low-density polyethylene (LDPE) plastic bags and transported to the sample preparation laboratory at the Swedish Museum of Natural History in Stockholm, Sweden, where the birds were kept frozen until time of sampling. The birds were then allowed to thaw for ∼12 h, and the birds’ biological variables were recorded according to previously described procedures (25). These biological variables were sex (determined by the gonads), body length (BL), i.e., from the tip of the tail to the tip of the bill ((1 mm), tail-feather length (TL), i.e., from the base of the tail to the tip of the longest feather ((1 mm), wing length (WL), i.e., the length from the carpal joint to the tip of the longest primary feather ((1 mm), wing span (WS), i.e., tip to tip measured on outstretched wings ((1 mm), body weight (BW) ((1 g), muscle weight (MW), i.e., the sum of the left and right pectoral muscles ((1 g), liver weight (LW) ((1 g), and kidney weight (KW) ((1 g). The muscle somatic index (MSI) was calculated (MSI ) MW (g)/BW (g) × 100) as well as the liver somatic index (LSI) (LSI ) LW (g)/BW (g) × 100) and kidney somatic index (KSI ) KW (g)/BW (g) × 100). For the chemical analysis, 10 g per individual of the right and left pectoral muscles were sampled, placed individually in ethanol/acetone rinsed glass jars, and stored at -20 °C pending the chemical analysis. A total of 23 different OCs and BFRs were analyzed from 10 female and 10 male guillemots. The lipid content (F%) of the muscles were determined gravimetrically during the chemical analysis (see below). Analyzed OCs and BFRs. The following OCs and BFRs were analyzed: 2,2-bis(4-chlorophenyl)-1,1,1-trichloroethane (p,p′-DDT) and its’ metabolites p,p′-dichlorodiphenyldichloroethylene (p,p′-DDE) and p,p′-dichlorodiphenyldichlorethane (p,p′-DDD), polychlorinated biphenyls (PCBs) with the congeners’ respective International Union of Pure and Applied Chemistry (IUPAC) number within parentheses, 2,4,4′-tri-CB (CB28), 2,2′,5,5′-tetra-CB (CB52), 2,2′,4,5,5′penta-CB (CB101), 2,3,3′,4,4′-penta-CB (CB105), 2,3′,4,4′,5penta-CB (CB118), 2,2′,3,4,4′,5′-hexa-CB (CB138), 2,2′,4,4′,5,5′hexa-CB (CB153), 2,3,3′,4,4′,5-hexa-CB (CB156), 2,2′,3,4,4′,5,5′hepta-CB (CB180), hexachlorocyclohexane (isomers R-, β-, and γ-HCH), trans-nonachlor (t-nonaCl), and hexachlorobenzene (HCB). The BFRs were hexabromocyclododecane (HBCD) and the polybrominated diphenyl ethers (PBDEs) 2,2′,4,4′-tetra-BDE (BDE47), 2,2′,4,4′,5-penta-BDE (BDE99), 2,2′,4,4′,6-penta-BDE (BDE100), 2,2′,4,4′,5,5′-hexa-BDE (BDE153), and 2,2′,4,4′,5,6′-hexa-BDE (BDE154). Homogenization and lipid extraction of all the samples were carried out at the Institute for Applied Environmental

Research (ITM) at Stockholm University. The chemical analyses of OCs were performed at the Special Analytical Laboratory (RSL) of the Swedish Museum of Natural History, and the analyses of BFRs were performed by ITM. Sample treatment, analytical methods for quantification of individual OC and BFR isomers/congeners, and laboratory quality assurance/quality control procedures have been previously described for OCs (9, 26, 27) and BFRs (28). In short, the samples were extracted with a mixture of acetone/n-hexane and n-hexane/diethyl ether and subjected to further cleanup. For analysis of OCs, the sample extracts were injected into gas chromatographs equipped with electron capture detectors. Through the use of a combination of different capillary columns, all compounds could be separated except for CB138, where the interfering peak caused by CB163 meant that CB138 was overestimated (∼20-30%). To facilitate reading, CB138 is henceforth written alone. For analysis of BFRs, the sample extracts were injected into gas chromatographs/mass spectrometers (GC/MS), a VG Trio 1000 (Manchester, U. K.) with a Carlo-Erba gas chromatograph (MEGA MFC 500, Italy) equipped with a split-splitless injector (270 °C) and a 30 m DB-5MS (J&W Scientific) fused-silica column (0.25 mm i.d., 0.25 µm film thickness). The standards used were individual BDE congeners: BDE47, BDE99, BDE100, BDE153, and BDE154 (Cambridge Isotope Laboratories) and HBCD (Michigan Chemical, St. Louis, MI). As an internal standard, Dechlorane 603 (Hooker Chemical Corp.) was used. Helium was used as the carrier gas. The temperature program was 80 °C (2 min), 25 °C/min to 200 °C, 4 °C/min to 315 °C (15 min). Electroncapture negative ionization (ECNI) was used for the MS analysis. For the MS conditions, ammonia was used as the reaction gas as well as an electron energy of 70 eV and transfer line and ion source temperatures of 300 and 220 °C, respectively. The mass fragments monitored for quantitative analysis were m/z 79 and 81 for PBDEs and HBCD and 237 and 239 for the internal standard Dechlorane. The concentrations of OCs are expressed as ng/g on a lipid weight (lw) basis unless otherwise specified. Concentrations below the Level of Quantification. A few contaminants had nonquantifiable concentrations; when possible, the level of quantification (LOQ) was determined from each individual chromatogram. For t-nonaCl, this was not possible; therefore, a general LOQ was determined. The LOQ was defined as 10 times the standard deviation of the measurements, and in practice this was estimated by repetitive measurements of a sample containing analytes at a concentration close to the expected LOQ (29). When the multivariate analysis was performed, only the 16 contaminants with quantifiable concentrations were used (Table 2). Univariate Statistics. For all univariate statistics, the software GraphPad Prism 4.03 (30) was used. We performed column statistics, Shapiro-Wilk normality tests, unpaired two-tailed t-tests (if an F-test showed unequal variances between groups, Welch’s correction was used), as well as correlation analysis (Pearson). The significance level was set to 0.05 for all tests. Multivariate Data Analysis. All multivariate data analyses (MVAs), principal component analyses (PCAs), partial leastsquares projections to latent structures (PLS), and PLS discriminant analyses (PLS-DAs) were performed using the software SIMCA-P 10.0.4 (31). For all MVAs, a significance level of 0.05 was used, and centered and scaled data (to variance 1) were fitted to a principal component (PC) model (32). Determinations of the significant number of components were made by cross validation. Values of the explained variation, R2, and predicted variation, Q2, were calculated, and R2 values > 0.7 and Q2 values > 0.4 denote a good model when analyzing biological data (33). PLS is a regression VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8631

TABLE 1. Body Length (BL), Tail-Feather Length (TL), Wing Length (WL), Wing Span (WS), Body Weight (BW), Muscle Weight (MW), Muscle Somatic Index (MSI), Liver Weight (LW), Liver Somatic Index (LSI), Kidney Weight (KW), Kidney Somatic Index (KSI), and Lipid Content (F%)a female (n ) 10) GM

-CI

TABLE 2. Concentrations (Geometric Mean (GM) ( 95% Confidence Interval (-CI and +CI)) of Organochlorines and Brominated Flame Retardants in Pectoral Muscles of Guillemots (Uria aalge) (n ) 20) Collected in 2000 South of Stora Karlso1 in the Baltic Seaa

male (n ) 10) +CI

GM

-CI

female (n ) 10) (ng/g lw)

+CI GM

BL (mm) 439 428 451 446 439 453 TL (mm) 50.7 49.1 52.2 51.2 50.4 52.0 WL (mm) 207 204 209 205 204 206 WS (mm) 718 707 730 712 707 718 BW (g) 1050 1010 1100 1080 1040 1110 MW (g) 162 153 171 160 152 168 MSIb (%) 15.4 14.8 16.0 14.9 14.0 15.8 LW (g) 38.5 35.8 41.4 40.0 36.0 44.7 LSIc (%) 3.7 3.4 3.9 3.7 3.4 4.1 KW (g) 7.1 6.4 8.0 7.2 6.3 8.2 d KSI (%) 0.68 0.61 0.75 0.67 0.59 0.75 F% 2.9 2.6 3.2 2.9 2.6 3.1 a Values (geometric mean (GM) ( 95% confidence interval (-CI and +CI)) are based on 20 individual guillemots (Uria aalge) collected from south of Stora Karlso¨ in the Baltic Sea in 2000. b MSI ) MW (g)/BW (g) × 100. c LSI ) LW (g)/BW (g) × 100. d KSI ) KW (g)/BW (g) × 100.

extension of PCA to model the relationship between two blocks of variables, X and Y, that can both be multidimensional. This was performed by modeling X and Y separately and then relating their respective scores to each other to find if there were any systematic changes in the X matrix that correlated to any systematic changes in the Y matrix (34-36). Finally, PLS discriminant analyses (PLS-DAs) were used to determine whether an observation belonged to a preestablished class by modeling and quantifying the eventual discriminating variables that contributed to the class (e.g., X/Y matrices) separation (37). In the present work, this was performed after introducing so-called “dummy variables”, i.e., 0 for “male” and 1 for “female”, and then quantifying which variable or combination of variables that discriminates between the sexes.

Results and Discussion Biological Variables. The biological variables for the guillemots are presented (GM ( 95% CI) in Table 1. Univariate statistics, i.e., testing female vs male, each biological variable separately, showed no significant differences between the sexes. Concentrations of OCs and BFRs. Concentrations (GM ( 95% CI) in pectoral muscles are presented in Table 2. The total concentration of OCs was approximately 150 times higher than that of BFRs. The contaminant with the highest concentration in both females and males was p,p′-DDE with a concentration close to twice (1.95 times) as high as the concentration of ΣPCB. The same relationship exists also in guillemot eggs from Stora Karlso¨ with a p,p′-DDE/ΣPCB concentration ratio close to 2 (9). In contrast, the opposite relationship has been found in tissues from other circumpolar seabirds with a p,p′-DDE/ΣPCB concentration ratio below 1 (38-41). The reasons for this relatively high p,p′-DDE/ΣPCB concentration ratio for guillemots from Stora Karlso¨ are not known but differences between species regarding, for example, exposure and capacity to biotransform and excrete contaminants are likely to be involved. The concentration in guillemot muscles of β-HCH was ∼20 times higher than that of both R-HCH and γ-HCH (Table 2). The insecticide Lindane is a technical mixture of HCH isomers with 10-15% of the active isomer γ-HCH and the rest 60-70% R-HCH, 5-12% β-HCH, and 6-10% δ-HCH (42, 43). The relatively high concentration of β-HCH in guillemot muscle is in accordance with an earlier finding that β-HCH 8632

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 22, 2005

-CI

+CI

male (n ) 10) (ng/g lw) GM

-CI

+CI

p,p′-DDE 13 400 11 500 15 500 12 400 11 100 13 800 p,p-DDD NDb ND p,p-DDT ND ND R-HCHc 12.0 9.7 14.8 10.1 9.9 10.3 β-HCH 279 238 327 250 203 307 c γ-HCH 14.4 11.6 17.9 11.5 10.1 13.2 HCB 605 537 681 571 509 641 t-nonaCl 10d 10 CB28 51.6 40.8 65.2 43.6 38.8 49.0 CB52 ND ND CB101c 20.9 17.6 24.9 17.7 15.5 20.2 CB105 672 553 817 620 553 695 CB118 1700 1350 2140 1520 1380 1670 CB138 1940 1460 2580 1700 1500 1920 CB153 2690 2020 3600 2230 1980 2520 CB156 544 444 665 560 495 635 CB180 920 672 1260 760 645 895 ΣPCBe 7370 5630 9640 6290 5670 6980 BDE47 71.5 43.2 118 46.7 36.9 59.0 BDE99 13.4 8.9 20.1 9.6 8.4 11.1 BDE100 4.9 3.4 7.1 3.6 3.2 4.1 BDE153 1.9 1.1 3.2 1.3 1.0 1.7 BDE154 4.1 3.4 4.8 4.6 3.8 5.6 ΣPBDEf 96.9 61.0 154 66.5 55.1 80.3 HBCD 66.7 55.0 80.9 62.7 42.6 92.4 a For abbreviations, see the Materials and Methods section. b ND ) not detected. c R-HCH, γ-HCH, and CB101 have individually determined LOQs. d General level of quantification for trans-nonachlor ) 10 ng/g lw. e ΣPCB ) sum of ICES seven marker PCBs: IUPAC nos. CB28, CB52, CB101, CB118, CB138, CB153, and CB180. For CB52, ND ) 0. f ΣPBDE ) sum of BDE congener nos. BDE47, BDE99, BDE100, BDE153, and BDE154.

biomagnifies ∼20 times between herring (Clupea harengus) muscle and guillemot egg while the R-HCH and γ-HCH isomers do not biomagnify (9). The concentration of ΣPBDE was not significantly different from that of HBCD (Table 2). The concentrations of different BFRs were in decreasing order: BDE47 ) HBCD > BDE99 > BDE100 ) BDE154 > BDE153. BDE47 and BDE99 are the major congeners in the PentaBDE formulation, and the concentrations of these congeners in guillemot egg from the Baltic Sea peaked in ∼1985-1990, followed by a decline over the last 15 years. This decline was rapid only during the first 10 years (28). In herring gull eggs from the Laurentian Great Lakes (U. S.), the levels of PBDE are still increasing (44). Levels and time trends of BFRs in the environment have recently been reviewed in a number of papers (45-49). The time trend of HBCD is different from that of PBDEs in guillemot egg from Stora Karlso¨, Baltic Sea. During the 10 years 19701980, a peak HBCD concentration was found in 1975 followed by a rapid decline to fairly low levels (∼30 ng/g lw) in 1980. During the subsequent two decades, the HBCD levels in guillemot eggs have increased steadily to around 160 ng/g lw (28), which is in fact higher than the HBCD concentration found in muscle sampled at the time of egg formation (Table 2). This might indicate a specific accumulation of HBCD in eggs. The use of the technical products PentaBDE, OctaBDE, and DecaBDE are the primary sources for PBDEs in the environment (46, 50). The dominating congener in guillemot muscle was BDE47, a major constituent in the PentaBDE mixture. Apart from PentaBDE, some BDE47 might originate from reductive debromination of other PBDE congeners.

FIGURE 1. Principal component analysis (PCA) (R 2X ) 0.62 and Q 2 ) 0.25, two components) based on biological characteristics from guillemot (Uria aalge) from Stora Karlso1 , Baltic Sea, in 2000 and concentrations of contaminants in pectoral muscles. (a) Score plot with 20 individually numbered females (F) and males (M). The ellipse shows Hotellings T 2 (0.05). (b) Loading plot. (c) Model overview with R 2X (lightbars) and Q 2 (dark bars). For abbreviations, see the Materials and Methods section. For instance, both OctaBDE and DecaBDE have been shown to undergo debromination to less brominated congeners (51-54). In a study regarding atmospheric long-range transport and environmental fate of PBDE, air samples were collected over the island Gotska Sando¨n in the Baltic Sea, ∼130 km north of the guillemot colony at Stora Karlso¨ (55). In that study, BDE209 was found to be the most abundant BDE congener, and the level of PBDE was 40 times higher than that of ΣPCB. The authors attributed this to the currently high consumption volume of DecaBDE. Accordingly, in the eggs of peregrine falcons (Falco peregrinus) from Sweden, levels of BDE209 up to 430 ng/g lw have been found (56). In a screening study, where the same guillemot muscles as in this paper were sampled, divided in a female and a male pool, and analyzed with the expressed purpose to identify especially “new” BFRs not yet analyzed in biological samples, no BDE209 was detected. However, in that study three different methoxylated PBDEs (MeO-BDEs) were identified, 6-MeO-BDE47, 6′-MeO-BDE49, and 6′-MeO-BDE68 at levels ∼1-5 ng/g lw (57).

Since 2004 the use of PentaBDE and OctaBDE has been banned in the European Union (58), and the major manufacturer in the U. S. has agreed to phase out production during 2005. DecaBDE, HBCD, and other BFRs are still in use, and this will change the BFR composition in biota in the future. However, the PentaBDE and OctaBDE already present in enormous quantities in products still in use will continue to contribute to the contaminant burden in biota for a long time. Multivariate Data Analysis. Today, there are numerous articles regarding the levels of contaminants in nature. However, few of these address the issue of how different contaminants are interconnected. When working with multivariate data, i.e., three or more variables determined per observation (in this study, an observation is an individual bird), initially it is valuable to obtain a general overview of how the different observations and variables are related (32). PCA: Biological Variables and Contaminants. An initial PCA with 9 biological variables (Table 1, MSI, LSI, and KSI excluded) and 16 contaminants (Table 2, concentrations VOL. 39, NO. 22, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8633

FIGURE 2. Coefficient plot with 95% CI for the respective variables for PLS model (R 2X ) 0.50, R 2Y ) 0.34, Q 2 ) 0.09) between liver weight (LW) (g) and concentrations of organochlorines and brominated flame retardants in individually analyzed muscles (n ) 20) from guillemot (Uria aalge) from Stora Karlso1 , Baltic Sea, 2000. For abbreviations, see the Materials and Methods section.