Environ. Sci. Technol. 2007, 41, 2723-2729
Organochlorine Compounds in Trout from Lakes over a 1600 Meter Elevation Gradient in the Canadian Rocky Mountains M A R C J . D E M E R S , * ,†,§ E R I N N . K E L L Y , ‡ JULES M. BLAIS,† FRANCES R. PICK,† VINCENT L. ST. LOUIS,‡ AND DAVID W. SCHINDLER‡ Centre for Advanced Research in Environmental Genomics, Department of Biology, University of Ottawa, 30 Marie Curie, Ottawa, Ontario, K1N 6N5, Canada, National Guidelines and Standards Office, Environment Canada, 351 St. Joseph Boulevard, Gatineau, Que´bec, K1A 0H3, Canada, and Department of Biological Sciences, University of Alberta, Edmonton, Alberta T6G 2E9, Canada.
The effect of altitude on the concentration and composition of organochlorine compounds (OC) in trout was investigated along an elevation gradient of 1600 m in the Canadian Rocky Mountains. The eight lakes sampled were within or adjacent to national parks in sparsely settled parts of Alberta and British Columbia, thus contaminants were assumed to have derived from long-range atmospheric transport. Concentrations of several OCs in trout increased significantly with lake elevation. In general, these increases were most pronounced for the higher Kow pesticides (i.e., dieldrin and DDTs), and less pronounced for lower Kow pesticides (e.g., HCHs and HCB) and PCBs. Most OC concentrations in trout were inversely correlated with fish growth rate. Growth rate explained more of the variation for some OCs (particularly PCBs) than lake elevation. Differences in trophic position (indicated by δ15N) explained little of the variation in OC concentration in comparison to other factors such as lake elevation and the growth rate and age of trout. Using principal component analysis (PCA), we identified the importance of lake elevation and octanol/water partition coefficient (Kow) to the OC composition of trout.
Introduction Organochlorine compounds (OCs) are omnipresent in the environment (1) as a consequence of their high stability, volatility, and capacity to biomagnify in foodwebs. One explanation for their ubiquitous distribution is global distillation (2, 3): volatile OCs follow successive evaporation and concentration steps from warmer temperate areas toward cooler regions where low temperatures suppress their subsequent evaporation (2, 3). Efficient scavenging of OCs by snow may augment loadings to regions receiving high precipitation (4, 5). As a result, according to the global distillation theory, one would expect the concentration of more volatile OCs to increase with elevation. * Corresponding author phone: (819) 994-8749; fax: 819-9565602; e-mail:
[email protected]. † University of Ottawa. § Environment Canada. ‡ University of Alberta. 10.1021/es062428p CCC: $37.00 Published on Web 03/08/2007
2007 American Chemical Society
OC concentrations in air (6, 7), soil (8), snow (4), plants (9, 10), and amphipods (11) from mountainous regions appear to increase with elevation and show patterns consistent with an altitudinal chemical fractionation. This is in contrast to studies on European mountain fishes (12-14) which reported concentration increases with elevation only for less volatile OCs such as DDT and the more chlorinated PCBs, but no relationship with elevation for more volatile chemicals. Daly and Wania (5) questioned the conclusions of these studies because the elevation gradient examined was confounded by proximity to different contaminant source regions. In particular, low altitude lakes (>1100 m.a.s.l.) were typically in northern Europe and farther from point sources, while high (>2400 m.a.s.l.) and intermediate (1600-2300 m.a.s.l.) altitude lakes were all in the more populated Alps and Pyrenees, respectively (12-14). In this study, we tested the hypothesis that OCs increase in trout with increasing lake elevation. In contrast to previous studies, all lakes are confined to a relatively small area (∼108 000 km2) in the Canadian Rocky Mountains distant from point sources of OCs. By selecting lakes within a confined and remote geographic region, differences in contaminant concentration and composition with respect to elevation should not be confounded by the proximity to point sources of contamination. The sampling location and study design allowed us to test whether patterns of altitudinal fractionation observed in soil, snow, plants, and amphipods apply also to fishes, in this case trout, which are typically the top predators in Rocky Mountain aquatic food webs. In addition to relating OC concentration in trout to lake elevation, we examined the variation in OC composition in relation to fish growth characteristics, lipid content and trophic position (indicated by δ15N).
Materials and Methods Site Description. Eight study lakes were chosen in Alberta and British Columbia (Figure 1, Supporting Information) to encompass a range in elevation from 760 to 2360 m.a.s.l. All lakes were located within national parks (Banff, Jasper, or Yoho National Parks) with the exception of Shere Lake which was located ∼20 km west of Mt. Robson Provincial Park. Lake elevation, latitude, longitude, lake surface area, and catchment area (Table 1, Supporting Information) were determined using digital elevation models derived from 1:50 000 topographic maps from the National Topographic Database analyzed in ArcGIS 9 (ESRI). Lakes were sampled in 2001 after ice-out in late May for low-elevation montane lakes and in early July for the higher elevation alpine lakes, with the exception of Bighorn Lake, which was sampled in July 1997. Surface water temperature (°C) and Secchi depth (m) were recorded. Surface waters for analyses of dissolved organic carbon (DOC), color (absorbance at 350 nm), total dissolved nitrogen (TDN), total dissolved phosphorus (TDP), and chlorophyll a (from Whatman GF/F filters) were filtered in the field within 1 h of collection (Table 1, Supporting Information). Unfiltered water samples were analyzed for alkalinity, pH, conductivity, ammonium (NH4+), nitrite + nitrate (NO2-2 + NO3-), total phosphorus (TP), and turbidity. All water chemistry analyses were conducted at the University of Alberta Limnological Services Unit (Edmonton, Alberta, Canada) as per standard methods. Sampling of Fishes. Brook trout (Salvelinus fontinalis), rainbow trout (Oncorhynchus mykiss), and bull trout (Salvelinus confluentus) were collected using gillnets in August 2001 from six of the eight study lakes. Brook trout were collected VOL. 41, NO. 8, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2723
FIGURE 1. Concentrations of selected organochlorine pesticides and PCBs in trout versus elevation. Concentrations are expressed on per lipid basis. Each symbol represents an individual fish, with the black squares representing brook trout, and open circles representing other trout species. The solid and dashed lines represent the linear regression of all trout species and only brook trout respectively. r2T corresponds to coefficient of variation for all trout species, whereas r2BK indicates the coefficient of variation for only brook trout. All regressions were significant (p < 0.05). from Bighorn Lake in 1997 and lake trout (Salvelinus namaycush) were captured from Pyramid Lake in August 2003. Additional rainbow, lake, and bull trout were collected at Moab Lake in 2003 (Table 2, Supporting Information). Each fish was identified (15, 16), measured for fork and total length (mm) and weight (g), and frozen on dry ice in the field. Fishes were aged at the University of Alberta (Edmonton, Alberta, Canada) using otoliths. Growth rate was estimated for each fish as weight/(age-1) (17). A fillet with skin attached was taken from each fish using dissection equipment rinsed with distilled water and pesticide-grade acetone and hexane. Each fillet was individually wrapped in prewashed aluminum foil and sealed within two Whirlpak bags. Fillets were shipped to the University of Ottawa (Ottawa, Ontario, Canada) on dry ice and then stored at -20 °C until extraction. Extraction of OCs and Analytical Methods. Approximately 8-10 g of fish dorsal tissue (including skin) were homogenized with Hydromatrix using a mortar and pestle. Prior to extraction, standards (National Laboratory for Environmental Testing) were added to determine extraction efficiencies. The extraction was performed using an accelerated solvent extractor (ASE 200, Dionex) first using dichloromethane (DCM), and then hexane. The two extracts were combined and dried by passing through sodium sulfate. Lipids were removed from the extract by automated gel permeation chromatography (GPC) using a precalibrated GPC Autoprep 1002A (Analytical Bio Chemistry Laboratories) 2724
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 8, 2007
with a flow rate set to 5 mL/min. The glass column (internal diameter 25 mm, length 600 mm with Teflon fittings) contained 65 g (dw) of Bio-beads preswelled in 1:1 hexane: DCM. The extract was fractionated on an 8 g silica gel column. The first fraction (50 mL hexane) contained the nonpolar organochlorine pesticides (OCPs) and PCBs, while the second fraction (80 mL of 1:1 hexane:DCM) contained the more polar OCPs. Iso-octane was added to each fraction, which was then evaporated down to 0.5 mL. Mirex (Ultra Scientific) was added as an internal standard. Extracts were analyzed by a Hewlett-Packard 6890 series II gas chromatograph (GC) equipped with a 63Ni microelectron capture detector. One µL of extract was injected in the splitless mode with an initial injector temperature of 250 °C. Helium was used as the carrier gas (flow rate 3.1 mL/min). Mid-level standards for the OCPs and PCBs were analyzed at intervals throughout sample analyses and used to recalibrate the instrument after every 20 injections. Sample extracts were screened for 27 OCPs and 127 PCB congeners, some of which coeluted to give 94 peaks. Analytes were considered present if the sample peak and its corresponding reference peak eluted within a retention time window of 0.02 min. Peak areas were quantified using HP Chemstation software. Method blanks and a certified reference material NIST 2978 mussel tissue were analyzed with every sample batch. Percent recovery of standards ranged from 52.7 ( 25.1% for
endrin ketone, 63.3 ( 10.6% for PCB 30, and 66.3 ( 10.1 for PCB 204, 79.9 ( 20.6% for 1,2,4,5-tetrabromobenzene, and 114.1 ( 27.5% for δ-HCH (hexachlorocyclohexane). All samples were blank and recovery corrected. In fraction A, PCB 30 was used to correct for the recovery of the lower Kow compounds (Kow of 3.5-5.5) while PCB 204 was used to correct for the higher Kow compounds (Kow of 5.6-7.5). In fraction B, δ-HCH was used to correct for the lower Kow compounds (Kow of 3.5-5.5) while endrin ketone was used to correct for the higher Kow compounds (Kow of 5.6-7.5). The lipid content of each sample was determined by extracting 2-4 g of fish dorsal muscle (including skin) using the ASE 200. The extract was evaporated to complete dryness, and the lipids were calculated gravimetrically. Stable Isotope Analysis. Since fish accumulate OCs mainly through the ingestion of food, stable isotope analysis was conducted to measure the variability in fish diet. Primary consumer δ15N varies among lakes; therefore, basal differences are also reflected in the δ15N of upper trophic level consumers (18, 19). To correct from basal δ15N differences, filter feeding Pelecypoda (Mollusca) were sampled for analysis of their δ 15N content because they were found in all lakes and are relatively long-lived. Pelecypoda were collected using an Ekman grab, and were sorted from the sediment, rinsed, put in Whirlpak bags and frozen. Skinless dorsal muscle samples from individual fishes and Pelecypoda were freeze-dried. Pelecypoda samples were acid treated to dissolve any remaining shells and surficial carbonates (20, 21). All samples were then homogenized to a fine powder using a nitric acid washed mortar and pestle. Samples were weighed into tin capsules (0.3-0.5 mg for fish, 1-2 mg for Pelecypoda), and analyzed at the G.G. Hatch Isotope Laboratories (University of Ottawa, Ottawa, Ontario, Canada) for carbon and nitrogen isotopes. Samples were combusted in an automated Carlo Erba (CE-1110) coupled to a Finnighan Mat Delta PLUS isotope ratio mass spectrometer (IRMS) with a Conflow III interface. Helium served as the carrier gas. Water was removed using a magnesium perchlorate trap. Stable isotopes were expressed in “delta” notation (δ) as parts per thousand (per mil) difference from the standard as follows:
δR‰ ) [(Rsample/Rstandard) - 1] × 1000 where R ) 15N/14N or 13C/12C. δ15N and δ13C values were calculated using a normalized calibration curve based on National Institute of Standards and Technology (NIST) NBS22 and IAEACH-6 standards for carbon, and IAEA-N-1 and IAEA-N-2 standards for nitrogen. An internal laboratory standard (caffeine) was run every 10th sample to correct for instrument drift. Precision was found to be 0.2‰ for nitrogen and 0.3‰ for carbon (based on analyses of 20 replicates). All δ15N trout values for a given lake were corrected for baseline δ15N by subtracting the δ15N of Pelecypoda from the δ15N of each trout from the same lake. Statistical Analysis. All concentrations were log transformed to normalize the data and equalize the variance. Weight, length, and growth rate were also log transformed. Data were analyzed using SYSTAT 10. Correlation analyses were used to relate lake and fish characteristics and fish isotopic signatures to lake elevation and OC concentrations in fishes, whereas simple linear regression analyses were used to distinguish relationships between OC concentrations in fishes and lake elevation and fish growth rate. A backward stepwise multiple regression (with alpha-to-enter and alphato-remove set to 0.05) was conducted to establish the relative predictive importance of elevation, growth rate, and trophic position. In the above analyses, fishes were considered as independent samples to illustrate the within lake variation.
We conducted a principal component analysis (PCA) to determine the variation in OC composition in trout using Canoco 4.5. We used the relative abundance of OC data as opposed to the absolute concentrations. Data were screened to only include analytes which accounted for >0.5% of the total OC concentration and were present in >75% of trout. This screening step retained 33 out of a possible 154 analytes (for further details see the Supporting Information). The PCA scores for the first two axes were then used as independent variables in subsequent correlation and regression analyses with respect to lake elevation and Kow (for those OCs with a known Kow).
Results and Discussion Lake Characteristics. Lake temperature, average air temperature, color, DOC, TDP, TP, NH4, and alkalinity were negatively related to lake altitude, while catchment area, lake volume, and NO2 + NO3 were positively correlated with elevation (Table 1, Supporting Information). As anticipated most of the lakes were oligotrophic (defined as TP 4-10 µg‚L-1 (22)). Fish Characteristics. Brook trout was the dominant fish species collected, appearing in six of the eight study lakes, and constituted 49 of the 92 fishes analyzed (Table 2, Supporting Information). Individual fishes (all species of trout) ranged from 161 to 589 mm in length, and weighed between 46 to slightly over 2000 g. Both fork length and weight were negatively correlated with elevation (r2 ) 0.108, p ) 0.0008, r2 ) 0.147, p < 0.0001 respectively). This is in agreement with previous work by Wilhelm et al. (23)., which showed a significant negative correlation between the weight of fish and lake elevation in bull trout from Rocky Mountain lakes. In the present study, trout ranged from 2 to 15 years of age, while lipid content varied from 0.39 to 23%. There was no correlation between age or lipid content and elevation. Growth rate was negatively correlated with elevation (r2 ) 0.211, p < 0.0001) and varied between 24 and 322 g‚yr-1. Isotope Composition and Inferred Feeding Behavior of Trout. The δ15N values of Pelecypoda used for baseline correction ranged from 0.15 to 4.39‰ among study lakes. The baseline corrected δ15N of fishes ranged from 2.29 to 13.12‰ (Table 2, Supporting Information) and was not correlated with elevation, implying that the length of the food chain did not change consistently with elevation. The baseline-corrected δ15N values from trout in the mid-altitude lakes (Pyramid, Patricia, and Moab in 2003) were higher than in trout from other lakes. Trout from the other lakes had a range of δ15N baseline corrected values between 1.90 and 9.06‰. The range in mean δ15N indicated that the feeding habits of trout were quite variable among the lakes. Trout with the lowest δ15N were feeding as primary or secondary consumers (24), whereas trout with the highest δ15N values were piscivorous. Trout with intermediate δ15N values were omnivorous, feeding on a range of prey from different trophic levels. The feeding behavior inferred from isotopes was supported by stomach content analyses. Fish stomach contents consisted primarily of invertebrates (ranging from grazers like gastropods to carnivores like Mysis) and small fish (25). δ13C signatures in trout varied from -31.96 to -17.52‰ (Table 2, Supporting Information) and was positively correlated with lake elevation (r2 ) 0.263, p < 0.0001). Generally, a δ13C signature greater than -20‰ implies that the carbon source of a fish’s diet comes from benthic algae and benthic invertebrates (26, 27) and/or insects of terrestrial origin, whereas a δ13C signature below -25‰ suggests carbon source of mainly pelagic planktonic zooplankton (26, 27). Stomach content analyses revealed that terrestrial flying insects comprised a large proportion of the brook trout diet in Bighorn Lake (28). The stomach analyses, in addition to δ13C VOL. 41, NO. 8, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2725
FIGURE 2. Concentrations of selected organochlorine pesticides and PCBs in trout versus growth rate. Growth rate was estimated for each individual fish as weight/(age-1). Concentrations are expressed per lipid gram weight. Each symbol represents an individual fish, with the black squares representing brook trout, and open circles representing all other trout species. The solid and dashed lines represent the linear regression of all trout and only brook trout respectively. r2T corresponds to the coefficient of variation for all trout species, whereas r2BK indicates the coefficient of variation for only brook trout. All regressions were significant (p < 0.05). and δ15N values indicated that the trout from our study lakes had pelagic, benthic and terrestrial components to their diet, as has been observed in other trout populations from mountain lakes (16, 23, 25, 29). Organochlorine Concentrations in Trout. Mean concentrations of individual OCs in trout varied from 61 ( 52 pg‚g-1 for γ-chlordane to 4050 ( 5400 pg‚g-1 for p,p′-DDE. The sum of PCB concentration in trout averaged 7700 ( 10 400 pg‚g-1, and tetra, penta, and hexa-chlorinated congeners were the dominant homologues. These concentrations are comparable to those previously recorded in fish from alpine lakes in the Rockies (30) and Europe (12-14, 29, 31). Concentrations of all organochlorine pesticides (OCPs) and most PCBs analyzed in trout were significantly (p < 0.05) correlated with lake elevation (Figure 1). Elevation explained between 6.5 and 61% of the variation when OCs were normalized by lipid weight, whereas only 1.2-47% of the total variation was explained when OCs were normalized by wet weight. These relationships remained significant, and the variation explained did not change appreciably when only one species, brook trout, was analyzed (present as black squares in Figure 1). When the fish muscle OC data were expressed as an average for each lake, elevation explained between 3.6 and 77% of the variation. These correlations with elevation remained significant (p < 0.05, n ) 8) for all OCPs except for R and γ-HCH (the pesticides with the lowest Kow). The correlations between lake-average PCB concentration in muscle and elevation were no longer significant, due 2726
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 8, 2007
to the reduction in degrees of freedom and statistical power. The slope of the regression lines of lake elevation vs trout OC concentrations were steeper for the more hydrophobic (log Kow 5.2-6.5), less volatile OCPs (e.g., dieldrin and DDTs) than for the less hydrophobic, more volatile OCPs (e.g., HCH and HCB) (Figure 2, Supporting Information). Concentrations of PCBs in trout were only weakly correlated with lake elevation, and no relationship was observed with Kow and the slopes of the regression for PCBs. Effect of Growth Rate and Other Variables. Most OCs were inversely correlated with fish growth rate, suggesting that growth dilution has a significant impact on OC concentrations (Figure 2). In many cases, the r2 was somewhat improved when only the brook trout were analyzed (Figure 2). Growth tends to result in the dilution of a chemical as the contaminant is divided among more biomass as a fish grows causing a decline in contaminant concentration (11, 32, 33). Increases in growth rate cause an appreciable dilution of a chemical that has a slow clearance rate (i.e., small respiratory clearance, metabolic clearance and/or egestion rate), especially for the most persistent OCs (i.e., PCB 180) (33, 34). Less persistent OCs tend to be largely influenced by changes in metabolic clearance. This is consistent with our results; in general, growth rate correlated best with the higher Kow compounds, and sometimes superseded the influence of lake elevation in our multiple regressions (Figure 2, Table 1).
TABLE 1. Multiple Regressions for Selected Organochlorine Compound Concentrations in Trout Based on Lipid Weight (pg/g Lipid Weight) of Individual Trout independent variable
coefficient ( SE
p (partial)
r2
SEest
log R-HCH
constant growth rate
3.489 ( 0.602 -0.682 ( 0.309
