C. Barnes, School of Marine Science and Technology, Ridley Building, University of Newcastle upon Tyne, NE1 7RU, UK. E-mail: Carolyn.Barnes@hotmail.co.uk
1Stable isotope data are widely used to track the origins and transformations of materials in food webs. Reliable interpretation of these data requires knowledge of the factors influencing isotopic fractionation between diet and consumer. For practical reasons, isotopic fractionation is often assumed to be constant but, in reality, a range of factors may affect fractionation.
2To investigate effects of temperature and feeding rate on fractionation of carbon and nitrogen stable isotopes in a marine predator, we reared European sea bass Dicentrarchus labrax on identical diets at 11 and 16 °C on three ration levels for 600 days.
3Nitrogen trophic fractionation (Δδ15N) was affected by temperature. Bass Δδ15N was 4·41‰ at 11 °C and 3·78‰ at 16 °C.
4Carbon fractionation (Δδ13C) was also affected by temperature. Bass Δδ13C was 1·18‰ at 11 °C and 1·64‰ at 16 °C. The higher lipid content in the tissues of bass reared at cooler temperatures accounted for the temperature effect on Δδ13C. When Δδ13C was determined using mathematically defatted values, there was a direct effect of ration size and Δδ13C was 2·51, 2·39 and 2·31‰ for high, medium and low rations, respectively.
5Reported Δδ15N for all treatments exceeded the mean of 3·4‰ widely used in ecological studies of fish populations and communities. This would confound the interpretation of δ15N as an indicator of trophic level when comparing populations that are exposed to different temperatures.
6The Δδ13C of 0–1‰ commonly applied in food web studies did not hold under any of the temperature or feeding regimes considered and a value of 2‰ would be more appropriate.
Nitrogen and carbon stable isotope data are widely used to describe the trophic levels of individuals, populations and communities, to identify the source materials that support them and to understand their trophic dynamics. Reliable interpretation of stable isotope data requires knowledge of the isotopic fractionation between diet and consumer, estimates of which will not apply in all circumstances. Understanding variation in trophic fractionation is important because small changes in fractionation affect estimates of trophic level and the contribution of different source materials to production (McCutchan et al. 2003). Much variation in fractionation has been observed within trophic groups, although the reasons for this are often poorly understood (Vander Zanden, Cabana & Rasmussen 1997; Sponheimer et al. 2003) and it remains unclear whether this variation is random or due to specific, predictable influences (Gannes, O’Brien & del Rio 1997). If some of this variation can be explained, then it will improve the interpretation of stable isotope data and ensure that stable isotope techniques can be applied with greater confidence.
Nitrogen trophic fractionation (Δδ15N) is usually assumed to be 3·4 ± 1·1‰ (1 SD) (Minagawa & Wada 1984). Although this estimate was based on a small number of individual estimates (16 new values plus seven from the literature), and a mixture of whole-body homogenates and specific tissues, the reported mean fractionation has proven remarkably resilient as further results have been compiled (e.g. Post 2002). However, there have been ongoing calls for more comparative laboratory experiments to aid interpretation of the natural variations in Δδ15N, as cases of enrichment, depletion and no change in 15N between diet and consumer have been reported (Ruess et al. 2004; MacNeil, Skomal & Fisk 2005). Carbon trophic fractionation (Δδ13C) has usually been assumed to vary from 0 to 1‰ based on whole-body experimentally derived values (DeNiro & Epstein 1978) and literature reviews of mixtures of species and tissues (Fry & Sherr 1984), and the most commonly applied value for Δδ13C is 0‰ or a small increase.
Trophic fractionation may be affected by temperature and/or feeding rate (e.g. Bosley et al. 2002; Olive et al. 2003). Only one study has examined temperature effects using a controlled feeding experiment (Power, Guiguer & Barton 2003). They reported a significant negative relationship between the δ13C and δ15N of Daphnia magna and Hyalella sp. across a temperature gradient, implying an increase in Δδ13C and decrease in Δδ15N as temperature rose (Power et al. 2003). Fractionation estimates applied to studies of wild animals are usually derived from captive feeding experiments where individuals are typically well fed and in good condition. Wild animals may, however, be under nutritional stress and the fractionation estimates applied may not be appropriate (Phillips & Koch 2002). For example, in Nile tilapia Oreochromis niloticus, Δδ13C and Δδ15N were significantly higher at higher feeding rates (Focken 2001) but when the study was repeated using larger sample sizes and a longer period, δ13C and δ15N decreased significantly with increasing feeding rate (Gaye-Siessegger et al. 2003). Hobson, Alisauskas & Clark (1993) fed juvenile Japanese quail Coturnix japonica a rationed diet designed to maintain but not increase, body mass, which showed significantly enriched tissue δ15N values over a control group fed the same diet ad libitum.
The European sea bass Dicentrarchus labrax is of importance to aquaculture and fisheries. The bass is both euryhaline and eurythermic, plays an important role in the ecology of many European estuaries and seas and can be maintained on constant diet under experimental conditions (Pickett & Pawson 1994). In the natural environment, bass populations live at a range of temperatures and food availability will not be constant, so the effect of temperature and feeding rate on fractionation needs to be known if wild populations are to be compared. The temperatures chosen for this experiment (11 and 16 °C) are well within the known range encountered by bass in the wild (Pickett & Pawson 1994). The aim of this study was to simultaneously assess effects of temperature and feeding rate on bass Δδ13C and Δδ15N. If temperature and feeding rate contribute to the variation in Δδ13C and Δδ15N, then interpretation of stable isotope data will be improved allowing more rigorous application of stable isotope techniques in the field.
Materials and methods
Approximately 1000 European sea bass fry were supplied by ‘Ecloserie Marine de Gravelines’ (Gravelines, France) and transported to the Centre for Environment, Fisheries and Aquaculture Science Laboratory, Suffolk in July 2003. As larvae they had been fed zooplankton (rotifers then Artemia) and as fry on commercially produced trash fish meal. Individuals were randomly assigned in approximately equal numbers to six 1·15 m3 (1150 litre) tanks – three of which were at 11 °C and three at 16 °C (Fig. 1). Both systems were in the same room with a 12L : 12D photoperiod. The seawater inlet was fed from the North Sea via an underground seawater storage tank in which the salinity was measured once a week.
The bass were fed on a diet of finely diced sandeel Ammodytes marinus, an important component of their wild diet (Pickett & Pawson 1994), prepared from a single thoroughly mixed batch of whole frozen sandeels supplied by Ammodytes Co. Ltd (Cornwall, UK). Owing to the large quantity of diet required, the batch of sandeels was processed in 15 lots. Prepared diet was frozen until use and the days on which bass were fed with each lot were recorded. A sample of approximately 10 g of diet was put aside from each lot for stable isotope analysis, from which at least two 1-mg samples were analysed.
The bass were fed once daily on all days excluding weekends and UK public holidays (maximum 3 days unfed) at three different ration levels: a satiation ration (hereafter called high ration, H), 66% of satiation (medium, M) and 33% (low, L) in one tank at each temperature (warm, W and cool, C). Bass eat more at a higher temperature (Hidalgo, Alliot & Thebault 1987; Russell, Fish & Wootton 1996) so the satiation level was determined in the warmer tank. The same satiation-feeding rate was used in the high ration cool tank with any uneaten excess being removed after the fish were satiated. After 6 months the medium and low rations were increased to 75% and 50% on welfare grounds as the average weight of the medium and low ration bass was less than 85% of those fed to satiation.
Three hundred and sixty bass were sampled during the experiment (three bass from six tanks monthly for 20 months). Each month three individuals from each tank were netted and humanely killed by a single blow to the head then immediately frozen to −16 °C. Freezing does not affect δ15N or δ13C (Sweeting, Polunin & Jennings 2004). Subsequently, the frozen individuals were weighed and measured to the nearest 0·1 g and 0·1 cm (fork length) and a dorsal sample of approximately 2·0 g of white muscle tissue was removed.
Bass muscle and diet samples were freeze-dried for grinding and homogenization using a pestle and mortar. One-milligram samples were weighed into tin capsules for stable isotope analysis. Fifty-four samples were analysed (Newcastle University) using a PDZ Europa 20–20 elemental analyser combustion continuous flow isotope ratio mass spectrometer (CF-IRMS) (PDZ Europa Ltd, Northwich, UK) linked to an Automated Nitrogen Carbon Analysis (ANCA) Solid/Liquid Preparation Module (PDZ Europa Ltd). Three hundred and thirty-two samples were analysed (Scottish Crop Research Institute, Dundee) using a Europa Scientific ANCA-NT 20–20 Stable Isotope Analyser with ANCA-NT Solid-Liquid Preparation Module (Europa Scientific Ltd, Crewe, UK). Replicates of cod Gadus morhua muscle tissue from a single individual were included in every batch so that adjustment could be made for interlab machine differences.
C : N ratios are reported as atomic values, calculated as (%C/12)/(%N/14). Ratios of 14N : 15N and 12C : 13C are expressed in conventional delta notation (δ) relative to the international standards, atmospheric air (N) and Pee Dee Belemnite (C). Two capsules of an internal reference were analysed every 10 samples to compensate for machine drift and as a quality control measure. Within-batch precision (standard deviation of the internal standard) ranged from 0·13 to 0·44‰ for δ15N and 0·04–0·25‰ for δ13C. Overall analytical precision was calculated using the standard deviation of δ15N and δ13C of the replicated cod muscle. The values were 0·15‰ for δ15N and 0·10‰ for δ13C after a correction factor was applied to each SCRI batch to compensate for differences due to time and/or laboratory procedures.
Lipids are depleted in 13C relative to the other major components (protein and carbohydrate) of tissue (DeNiro & Epstein 1977). Thus, variation in lipid content influences δ13C. Accounting for the differential effects of lipid synthesis and storage allows for more precise and interpretable δ13C comparison of tissue (DeNiro & Epstein 1978). Comparisons between δ13C values for bass tissue following chemical lipid removal and mathematical lipid correction have shown that the two methods provide consistent results (Sweeting, Polunin & Jennings 2006). The C : N ratios were used to mathematically correct the bass muscle δ13C values to lipid-free equivalents using the mass balance equation:
( eqn 1)
where δsample and C : Nsample are the experimentally derived values and C : Nprotein is 3·663, the atomic C : N of pure bass protein (Sweeting et al. 2006).
For the statistical analysis, the replication of bass sampled at each of the different sampling dates allows the interactions between ration, temperature and day to be estimated, as well as their main effects. The data from the first six sampling dates were excluded to avoid any equilibration effect during the first 200 days, as bass are known to take this long to equilibrate to their diet at this age (Sweeting, Jennings & Polunin 2005). The fact that different bass were sampled on each occasion means that we can treat each observation as independent. Thus, we are left with a 3 × 2 factorial design (three rations, two temperatures) at each of 14 sampling days. In our analysis we treat sampling day as a covariate. Formally we can write the full linear model in the form:
where ΔδnnXijk is the trophic fractionation (nnX = 13C or 15N), Ti is a parameter denoting the effect of the ith level of temperature, Rj is a parameter denoting the jth level of ration, (TR)ij denotes the interaction between temperature and ration, f(Dayk) is either a linear or quadratic function of the 14 values of sample day, (Tf(Day))ik denotes the interaction between temperature and day (Rf(Day))jk denotes the interaction between ration and day and ɛijk is an error term that is assumed to be normally distributed with mean 0 and variance σ2. Minitab 13·30 (Minitab Inc.) and R (R Development Core Team 2005) were used to perform the statistical analyses. In reporting the results of the analyses, mean values ±1 confidence interval (α = 0·05) are reported unless stated.
The temperature was 16·0 ± 0·03 °C and 11·1 ± 0·05 °C in the warm and cool systems, respectively. The salinity was 28·6 ± 0·25. The mean quantity of sandeel fed per feed over the course of the experiment was 59·5 ± 0·36 g, 43·6 ± 0·34 g and 27·8 ± 0·44 g in the high, medium and low ration tanks, respectively, equating to 100%, 73% and 47% of satiation ration.
At the start of the experiment mean length of the sampled fish was 7·3 ± 0·3 cm and weight 4·9 ± 0·7 g (n = 18). At the end of the experiment (day 597) mean length was 13·4 ± 0·9 cm and weight 31·0 ± 5·6 g (n = 18). As expected, the growth rates differed between tanks as the bass grew more in the warmer conditions and within temperature growth was higher on larger rations, but growth was achieved in all treatments throughout the experiment. Recorded natural mortalities accounted for approximately 10% of total mortality. Four deaths were due to two failed attempts at cannibalism. Six exceptionally large individuals were removed on welfare grounds to reduce competition. At 21 months the remaining bass (n = 454) were sampled for other investigations.
Overall elemental and stable isotope results for the sampled bass and sandeel diet are summarized (Table 1).
Table 1. Elemental and stable isotope results for European sea bass fed a sandeel diet over 597 days in a controlled environment
C : N
The C : N ranged from 3·61 to 8·09 (mean 4·19 ± 0·04). The C : N data did not conform to the assumptions underlying anova so the nonparametric Kruskal–Wallis test was performed. The bass from the cool tanks had a higher C : N than those from the warm at all ration levels (H > 42, P < 0·001 in all three cases) and those fed higher rations had higher C : N at both temperatures (H > 19, P < 0·001 in both cases) (Fig. 2).
The mean δ15N and δ13C of the diet was 12·39 ± 0·41‰ and −18·38 ± 0·22‰, respectively. Despite the sandeel diet being prepared from a single mixed batch the stable isotope signatures of the processed lots varied. To calculate the trophic fractionation, weighted mean values were calculated to account for the feeding duration for each lot. These calculated values were 12·43‰ and −18·39‰ for δ15N and δ13C, respectively.
nitrogen trophic shift
Δδ15N ranged from 3·13 to 5·42‰ and was affected by temperature with the cooler values consistently greater than the warmer. Temperature by day and temperature by ration interactions were statistically significant (F2,251 = 8·81, P < 0·001 and F2,251 = 40·60, P < 0·001). The magnitude of the temperature effect became greater with increasing experimental duration (Fig. 3a). The effect of ration level depended on temperature (Fig. 3b). At all rations Δδ15N was higher in the cooler temperature but the difference in Δδ15N between cool and warm varied depending on ration, with greatest difference (−0·2‰ per 1 °C) at medium ration and least difference at high ration.
carbon trophic shift
Bulk (nondefatted) Δδ13C ranged from −1·07 to 2·39‰ and was affected by temperature with the warmer values consistently greater than the cooler. As for δ15N the magnitude of the temperature effect changed with experimental duration (F2,251 = 3·74, P < 0·05) (Fig. 4a). The temperature effect also depended on ration (F2,251 = 5·00, P < 0·01) (Fig. 4b). At all rations Δδ13C was lower in the cooler temperature but the difference in Δδ13C between cool and warm varied depending on ration, with greatest difference at high ration and least difference at low ration.
‘Arithmetically defatted’Δδ13C ranged from 1·04 to 3·13‰ and was affected by temperature; however, the effect was not consistent through the duration of the experiment (F2,251 = 5·34, P < 0·01) (Fig. 5a). Ration level also had a direct significant effect on defatted Δδ13C (F2,251 = 11·88, P < 0·001) (Fig. 5b) with means of 2·51‰ ± 0·06, 2·39‰ ± 0·06 and 2·31‰ ± 0·06 for the high, medium and low rations, respectively. Tukey's post-hoc pairwise comparisons showed that high ration was significantly different to low ration (P < 0·001).
Both nitrogen and carbon trophic fractionation were affected by temperature and feeding rate. The results demonstrate that temperature and feeding rate effects would confound the interpretation of δ15N as an indicator of trophic level, or δ13C as an indicator of source and trophic level, when comparing populations exposed to different temperatures and/or feeding opportunities. Δδ15N exceeded 3·4‰ in 94% of sampled fish and lipid-free Δδ13C always exceeded 1‰. As the Δδ15N commonly applied in stable isotope studies (3·4‰) was lower than the fractionation recorded in our experiments, it is important to quantify the biases introduced by assumed fractionation in ecological studies. This conclusion also applies to δ13C, given that commonly applied values of Δδ13C (0–1‰) are much lower than those recorded in this and other studies on fish (e.g. Pinnegar & Polunin 1999). Indeed, the results suggest that a value of 2‰ would be more appropriate for lipid-free fish muscle.
Experimental estimates of trophic step fractionation may not reflect those occurring in the natural environment for a number of reasons. Experiments usually require controlled environments and focus on small species or the young of large species at high stocking densities. In this study, the Δδ15N for high ration at the cool temperature (3·66‰) was consistent with that reported by Sweeting et al. (2005) for bass fed to satiation on sandeel and dab diets in a much larger experimental system and exposed to natural temperature (mean 12 °C) and light regimes (3·83‰ on dab and 3·98‰ on sandeel). This observation, and the fact that sandeel are an important component of wild bass diet (Pickett & Pawson 1994), suggests that the temperature and diet effects on fractionation that we recorded will be indicative of patterns observed in wild bass. Moreover, wild young-of-the-year bass show a preference for temperatures above 10 °C but are known to commonly encounter temperatures as low as 6 °C and as high as 18 °C (Russell et al. 1996), and so the experimental temperatures were well within their usual range.
The C : N ratios reflect the quantity of fat reserve and show that the cooler, well-fed bass were in the best condition. Russell et al. (1996) reported that an increase in temperature between 10 and 18 °C generally resulted in a decline in the condition of bass that were fed a fixed ration. We conclude that the higher lipid content at the cooler temperature was responsible for the change in δ13C with temperature. This is because there was no consistent effect of temperature on Δδ13C following mathematical lipid correction and because the higher lipid corrected Δδ13C at higher rations was consistent with the results of theoretical modelling (Olive et al. 2003).
Temperature affected Δδ15N and Δδ13C and thus single fractionation values are not necessarily applicable in all environments. The δ15N signatures of the bass reared at 16 °C were, on average, 0·63‰ less enriched than those at 11 °C. Our finding that Δδ15N was lower at the warmer temperature is in agreement with the work of Power et al. (2003) on the primary consumers Daphnia magna and Hyalella sp. For carbon, there was a consistent temperature effect on bass bulk Δδ13C only, but further investigation is needed to see whether there is a physiological effect on lipid-free muscle. This study adds weight to the argument that temperature affects fractionation in all consumers regardless of their trophic level.
Notably, although the Δδ15N for all treatments fell within 1 SD of the mean of 3·4‰ widely used in ecological studies and reported by Minagawa & Wada (1984), Δδ15N was consistently higher than 3·4‰. An assumed fractionation of 3·4‰ would lead to overestimates of trophic level for bass. In consumers, nitrogen fractionation occurs during assimilation and excretion. Observed trophic step fractionation therefore results from the balance between these processes (Ponsard & Averbuch 1999). There are two components to nitrogenous excretion in fish: the endogenous (for maintenance) and exogenous fractions; the former being affected by fish size and temperature (Brett & Groves 1979). Within limits, temperature accelerates most physiological processes (Schmidt-Nielsen 1997) thus higher growth rates are achieved. Higher growth rates lead to reduced excretion and, therefore, increased retention of the lighter isotope is expected. This would result in an inverse relationship between δ15N and growth rate, which is also in keeping with the reported δ15N enrichment in fasting animals (e.g. Hobson et al. 1993; Adams & Sterner 2000; Vanderklift & Ponsard 2003) due to low excretion and nitrogen recycling. However, a direct link between intake rates and Δδ15N can be ruled out as this study found the highest Δδ15N on the medium ration at the cooler temperature.
To conclude, our results show that assumptions of constant Δδ15N or Δδ13C will not apply when populations are living at different temperatures and different amounts of food are available to them. Given that many ecological (field) studies use isotopes to look at populations of many species in many environments, and given the considerable cost and logistical challenge of conducting comparatively simple experiments, it is unlikely that trophic fractionation will be reliably predicted for all populations in all environments. However, as stable isotope analyses often provide the only available tool for describing the trophic levels of individuals, populations and communities, identifying source materials and for understanding trophic dynamics (Post 2002), it is important to work towards applying isotope analyses more rigorously. A first step towards more rigorous application is to treat assumed fractionation as uncertain and to quantify and assess the effects of errors in assumed fractionation on the results and conclusions of analyses. This can be achieved by conducting sensitivity analyses and will help to identify the circumstances when inaccurate estimates of fractionation, and/or the confounding effects of environment, can alter the conclusions of studies. These circumstances will vary. For example, if estimated Δδ15N were used to determine the relative trophic levels of individuals of different body sizes in a single population living at a given temperature, the results would be much less sensitive to assumed fractionation and the effects of environment, than if estimated Δδ15 were used to compare the absolute trophic levels of individuals, from a range of species and populations, in different environments.
This work was funded by a NERC CASE studentship (NER/S/A/2003/11886) and CEFAS (DEFRA project M0731). We thank Stuart Hetherington at CEFAS for the daily care of the bass, Charlie Scrimgeour and all his team at SCRI and Gillian Taylor at Newcastle for the stable isotope analysis. Two anonymous reviewers provided useful feedback, which helped us improve the manuscript.