sample collection and processing
Samples of 431 marine and freshwater fishes and invertebrates representing 18 families were analysed for δ13C and δ15N with and without chloroform–methanol lipid extraction treatment (Table 1). All samples were stored frozen for a maximum of seven months before analysis. Freezing has not altered δ13C or δ15N values for fish (Sweeting, Polunin & Jennings 2004) or marine invertebrate tissues (Bosley & Wainright 1999), but has affected freshwater zooplankton (Feuchtmayr & Grey 2003). Our comparisons of δ13C and δ15N between lipid extracted and bulk tissue samples were based on aliquots of dried, homogenized samples subjected to identical frozen storage conditions.
Table 1. Freshwater (FW) and marine (M) fish and invertebrate species and tissue types (L = liver, M = muscle, G = gonad, W = whole body) analysed for δ13C and δ15N for bulk tissue and samples lipid extracted using chloroform–methanol. P-values indicate significant increases (+) or decreases (−) in δ13C or δ15N following lipid extraction with an overall α of 0·05. Additional data for European sea bass Dicentrarchus labrax (Linnaeus) liver and muscle (Sweeting et al. 2006) were also used for model fitting
| American eel Anguilla rostrata (Lesueur)||50||FW||M||3·3–9·8||< 0·001+||< 0·001+|
| Atlantic salmon Salmo salar (Linnaeus)||6||FW||M||3·0–3·2||0·109||0·075|
| Brook trout Salvelinus fontinalis (Mitchill)||12||FW||M||3·1–3·6||< 0·001+||0·792|
| Slimy sculpin Cottus cognatus (Richardson)||8||FW||M||3·2–3·2||< 0·001+||0·339|
| Atlantic salmon Salmo salar (Linnaeus)||6||FW||L||4·0–4·3||< 0·001+||0·409|
| Brook trout Salvelinus fontinalis (Mitchill)||6||FW||L||3·9–7·1||0·006+||0·410|
| Slimy sculpin Cottus cognatus (Richardson)||8||FW||L||4·0–8·9||< 0·001+||0·942|
| Bony bream Nematalosa erebi (Günther)||36||FW||W||2·9–8·0||< 0·001+||< 0·001+|
| Golden perch Macquaria ambigua (Richardson)||36||FW||W||3·2–6·4||< 0·001+||0·520|
| Silver tandan Porochilus argenteus (Zeitz)||36||FW||W||3·3–12·1||< 0·001+||< 0·001+|
| Spangled perch Leiopotherapon unicolor (Günther)||29||FW||W||2·7–9·3||< 0·001+||< 0·001+|
| Slimy sculpin Cottus cognatus (Richardson)||8||FW||G||2·5–4·3||< 0·001+||0·107|
| Atlantic bluefin tuna Thunnus thynnus (Linnaeus)||82||M||M||3·1–6·5||< 0·001+||< 0·001+|
| Atlantic bluefin tuna Thunnus thynnus (Linnaeus)||44||M||L||3·9–12·0||< 0·001+||0·048|
| Atlantic herring Clupea harengus (Hildebrand)||11||M||W||6·3–12·5||< 0·001+||0·664|
| Atlantic mackerel Scomber scombrus (Linnaeus)||3||M||W||3·8–9·2||0·083||0·172|
| Silver hake Merluccius bilinearis (Mitchill)||12||M||W||3·5–4·1||< 0·001+||< 0·001−|
| Alderflies (Megaloptera)||7||FW||W||3·7–5·7||0·011+||0·807|
| Caddisflies (Hydropsychidae)||7||FW||W||4·5–5·7||< 0·001+||0·643|
| Dragonflies (Gomphidae)||5||FW||W||4·0–4·4||0·002+||0·651|
| Mayflies (Heptageniidae)||26||FW||W||5·2–10·7||< 0·001+||0·514|
| Stoneflies (Perlidae)||15||FW||W||4·3–5·3||< 0·001+||0·002+|
| Water pennies (Psephenidae)||12||FW||W||4·3–7·5||< 0·001+||0·523|
| Water striders Aquarius remigis (Say)||12||FW||W||4·1–6·0||< 0·001+||0·051|
| Water strider nymphs Aquarius remigis (Say)||6||FW||W||4·0–5·7||< 0·001+||0·501|
| Krill (Euphausiidae)||10||M||W||3·7–4·3||< 0·001+||0·001+|
| Shortfin squid Illex illecebrosus (Leseur)||6||M||W||4·0–4·5||< 0·001+||< 0·001−|
Liver, muscle and gonad samples were removed from selected fish species (Table 1), thawed, lightly rinsed with deionized water, transferred to glass scintillation vials and dried at 60 °C for at least 48 h. The remaining fishes and all invertebrates were dried whole, with pooled samples of two to 20 individuals used for freshwater invertebrates and whole individuals used for marine invertebrates. Larger whole Atlantic herring Clupea harengus (Linnaeus), Atlantic mackerel Scomber scombrus (Linnaeus), silver hake Merluccius bilinearis (Mitchill) and shortfin squid Illex illecebrosus (Leseur) were lightly rinsed with deionized water, then finely minced and dried in aluminium weigh boats at 60 °C for at least 48 h until they reached a constant weight over 3 h. Dried samples were then homogenized with either a Wig-L-Bug® ball and capsule amalgamator (Crescent Industries, Auburn, ME, USA) and stainless steel grinding vials or a mortar and pestle, depending on tissue volume.
Two aliquots were removed from each homogenized sample; one aliquot was immediately prepared for SIA (see below), while the second underwent lipid extraction using a modification of the Bligh & Dyer (1959) method. To extract lipids, dried powdered samples were placed in glass centrifuge tubes and immersed in a 2 : 1 ratio of chloroform : methanol with a solvent volume about three to five times greater than sample volume. Samples were then mixed for 30 s, left undisturbed for greater than 30 min, then centrifuged for 10 min at 1318 g. The supernatant containing solvent and lipids was then discarded. This process was repeated at least three times or more until the supernatant was completely clear and colourless following centrifugation. Samples were dried at 60 °C for 24 h to remove remaining solvent. Euphausiid samples also underwent acid washing after lipid extraction and drying to remove exoskeletal carbonates. Acid washing consisted of addition of 1 N HCl until bubbling ceased (Jacob et al. 2005), and the samples were redried at 60 °C for 24 h.
To determine the relationship between bulk tissue C : N and percentage lipid, a quantitative modification of the Bligh & Dyer (1959) method was performed on four fish species (bony bream Nematalosa erebi (Günther), golden perch Macquaria ambigua (Richardson), silver tandan Porochilus argenteus (Zeitz) and spangled perch Leiopotherapon unicolor (Günther)). Briefly, c. 0·2 g of dry sample was weighed into a centrifuge tube, to which chloroform, methanol and distilled water were added at a ratio of 2 : 2 : 1·8. The mixture was shaken then centrifuged at c. 265 g. for 10 min. The bottom solvent layer was withdrawn and passed through a micropipette column packed with sodium sulphate into an aluminium weigh boat. The fraction remaining in the centrifuge tube was subjected to a second extraction with a 9 : 1 ratio of chloroform : methanol, centrifuged, the bottom layer withdrawn and passed through sodium sulphate as before. The fraction in the weigh boat (the lipid fraction) was evaporated in a drying oven and weighed to determine percentage lipid using the formula percentage lipid = (lipid weight/dry weight) × 100.
stable isotope sample preparation
Aliquots (0·2–1·2 mg) of lipid-extracted and bulk tissue samples were weighed to the nearest 0·001 mg and packed into tin capsules in preparation for SIA. Samples were flash combusted at 1100 °C and resultant gases delivered via continuous-flow for analysis of δ13C, δ15N, percentage carbon and percentage nitrogen using either a DELTAplus or DELTAplus Advantage isotope ratio mass spectrometer at the University of New Brunswick (UNB) and Northern Arizona University (NAU), respectively. C : N ratios were determined from percentage element weight. Measurements of commercially available reference materials across all runs were both accurate and precise with mean ± SD of –33·6 ± 0·16‰ for δ13C and –3·1 ± 0·18‰ for δ15N for acetanilide (n = 214) at UNB and mean ± SD of –25·9 ± 0·04‰ for δ13C and 2·0 ± 0·14‰ for δ15N for NIST 1547 (peach leaves, n = 149) at NAU. Replicate analyses of samples produced SD of 0·16‰ for δ13C and 0·16‰ for δ15N (n = 108) at UNB and SD for δ13C of 0·05‰ and 0·07‰ for δ15N at NAU (n = 61). Samples were also routinely analysed at both labs to ensure data were comparable (e.g. smallmouth bass muscle: UNB δ13C =–23·2 ± 0·11‰, δ15N = 12·5 ± 0·18‰, n = 19; NAU δ13C =–23·3 ± 0·02‰, δ15N = 12·4 ± 0·11‰, n = 3). All C and N isotope data are reported in δ notation according to the following equation: δX = [(Rsample/Rstandard) – 1] × 1000 where X is 13C or 15N and R is the ratio 13C : 12C or 15N : 14N (Peterson & Fry 1987). Standard materials are Vienna Pee Dee belemnite (VPDB) for carbon and atmospheric N2 (AIR) for nitrogen. All δ13C and δ15N values were normalized on the VPDB and AIR scales with IAEA CH6 (–10·4‰), CH7 (–31·8‰), N1 (0·4‰) and N2 (20·3‰).
statistical models and analyses
Differences in mean changes in C and N isotopes (δ13C′ – δ13C and δ15N′ – δ15N) between lipid extracted (δ13C′ and δ15N′) and bulk tissue δ13C and δ15N samples (Table 1) were tested. Differences were determined using paired t-tests (α = 0·05) and a subsequent Holm test to reduce the probability of committing type I errors as a result of multiple comparisons.
To compare lipid corrected δ13C estimates among models, log-likelihood values (assuming normally distributed errors) were calculated for models described by McConnaughey & McRoy (1979) and Fry (2002). The Akaike Information Criterion (AICc) value was also calculated for each model. AICc values are determined according to the equation
where k equals the number of parameters and lower AICc values reflect improved model fits. AICc values are also presented as AIC differences (Δi) according to the equation
Δi = AICci – min AICc,
where AICci corresponds to the AICc value for model i and min AICc is the model with the lowest AICc value among tested models (Burnham & Anderson 1998). AICc differences were calculated between individual models. Models with Δi of about 0–2 have substantial support as best model fits, Δi of 4–7 indicates considerably less support, and Δi > 10 provides essentially no support for a given model (Burnham & Anderson 1998). All tested models used bulk tissue C : N as a predictor of δ13C′ – δ13C.
The first model form that we tested is based on the McConnaughey & McRoy (1979) model (eqn 1),
and L and D represent sample lipid content and protein–lipid discrimination, respectively. The McConnaughey & McRoy (1979) model (eqn 1) was fit with D and θ based on our data set. A new generalized model based on eqn 1 was developed that maintained the nonlinear relationship of the difference in δ13C between bulk tissue and lipid extracted tissue, but aggregated assumed values into three parameters,
The y-asymptote, or D in eqn 1, corresponds to a in eqn 1a. The model estimate C : Nlipid-free is represented by –b/a (x-intercept), whereas b/c (y-intercept) is the δ13C difference corresponding to a C : N value of zero.
The second mathematical lipid correction approach that we tested is based on the Fry (2002) equation (eqn 2),
where P and F represent protein–lipid δ13C discrimination and C : Nlipid-free, respectively. The Fry (2002) equation (eqn 2) was fit with P and F based on our data set.
A new model of the difference in δ13C between bulk and lipid-extracted tissue and log-transformed C : N,
δ13C′ – δ13C = β0 + β1ln(C : N) (eqn 3)
was also explored. The model estimate of C : Nlipid-free is represented by
Normally distributed error terms, ɛ ~ N(0, σ2), were assumed and models were fit based on eqns 1, 1a, 2, and 3 to data from different tissue types and species with sample sizes greater than or equal to 10 (Table 1). For fish species, three nested models of a given type were fitted. The simplest model assumed all model parameters were the same across species and tissue types, the intermediate model assumed that the parameters were tissue-specific and the full model assumed that parameters were both tissue- and species-specific. For invertebrate species, two models of a given type were fitted where the simpler model again assumed that all model parameters were the same across species and the full model assumed parameters were species-specific. Likelihood ratio tests were performed for each model type and species group (fishes or invertebrates) to determine the most parsimonious models. Parameters were estimated for all models using least-squares procedures available in R (R Development Core Team 2006).