Hydrogen isotope ratios in animal body protein reflect trophic level



    1. Research Laboratory for Archaeology, University of Oxford, 6 Keble Road, Oxford OX1 3QJ, UK;
    Search for more papers by this author

    1. Research Laboratory for Archaeology, University of Oxford, 6 Keble Road, Oxford OX1 3QJ, UK;
    2. McDonald Institute for Archaeological Research, University of Cambridge, Downing Street, Cambridge CB2 3ER, UK; and
    Search for more papers by this author

    1. NERC Isotope Geosciences Laboratory, Keyworth, Nottingham NG12 5GG, UK
    Search for more papers by this author

    Corresponding author
    1. Research Laboratory for Archaeology, University of Oxford, 6 Keble Road, Oxford OX1 3QJ, UK;
      Robert E. M. Hedges, Research Laboratory for Archaeology, University of Oxford, 6 Keble Road, Oxford OX1 3QJ, UK. E-mail: robert.hedges@rlaha.ox.ac.uk
    Search for more papers by this author

Robert E. M. Hedges, Research Laboratory for Archaeology, University of Oxford, 6 Keble Road, Oxford OX1 3QJ, UK. E-mail: robert.hedges@rlaha.ox.ac.uk


  • 1Carbon and nitrogen isotope ratios in consumer tissues are known to correlate with diet isotope composition, and nitrogen isotope ratios are observed to increase with increasing trophic level.
  • 2We analysed nitrogen and hydrogen isotope ratios of collagen from 19 species of British fish, birds and mammals to investigate how δD also correlated with trophic level and with feeding environment (terrestrial or aquatic).
  • 3A strong relationship between trophic level and δD was discovered for both terrestrial and aquatic consumers.
  • 4The correlation between trophic level and δ15N was apparent for terrestrial consumers, but less so for aquatic consumers.
  • 5No differentiation was found between δD of aquatic and terrestrial consumers at the same trophic level.
  • 6This observation should provide an additional tool in the study of current and ancient animal and human food web ecology.


It is well established that the stable isotope composition of animal body tissues corresponds closely to that of their diets, with modifications according to the way food components are partitioned and by isotope fractionation during metabolism (Gannes, del Rio & Koch 1998). In the last few years, ecological applications of this observation include studies of animal migrations (Hesslein et al. 1991; Best & Schell 1996; Hobson 1999), dietary composition of carnivores (Angerbjörn et al. 1994; Ben-David, Flynn & Schell 1997), and investigations of habitat usage (Koch et al. 1995; Schoeninger, Iwaniec & Nash 1998). The technique has also been used widely in the study of prehistoric human diet (Van der Merwe & Vogel 1978; Ubelaker & Katzenberg 1995; Ambrose et al. 1997; Richards et al. 2000; Privat, O’Connell & Richards 2002).

The case of 15N/14N ratios in body protein shows a particularly clear dependence on diet and trophic level. Nitrogen isotope ratios (expressed as δ15N) in consumer body proteins are typically enriched over those in dietary protein by three or four parts per thousand (‰) (the ‘trophic level effect’: Minagawa & Wada 1984), due supposedly to catabolic processes involving nitrogen excretion (Ambrose 1991). This has led to the use of δ15N in trophic level and food web studies (Hobson 1990; Gu, Schelske & Hoyer 1996; O’Connell & Hedges 1999; Kelly 2000).

We report here on an investigation of D/H (2H/1H) and trophic level in mammals, birds and fish from Britain. We aimed to test whether an individual's hydrogen isotope ratios depended on diet type (aquatic/terrestrial) or trophic level.

Measurement of D/H ratios in animal tissues have been used more frequently as an indicator of climate and environment, or animal origin and migration (Cormie, Schwarcz & Gray 1994a; Chamberlain et al. 1997; Hobson & Wassenaar 1997; Hobson 1999; Kelly et al. 2002). Such studies make use of the fact that isotope fractionation associated with the atmospheric transport and evaporation of meteoric waters leads to geographical variations in the D/H ratios of surface water (Rozanski, Araguas-Araguas & Gonfiantini 1993), and hence of drinking water. In regions where these variations are large, they are the dominant cause of variations in the D/H ratio of animal tissue protein (Cormie et al. 1994a,b; Hobson, Atwell & Wassenaar 1999). However, while hydrogen food-chains will have local water values at their base, the relationship between body tissue D/H and diet is likely to be more complicated than simply reflecting local water values. First, only tissue H that is not exchangeable in the post mortem environment can be measured meaningfully. For collagen, the non-exchangeable hydrogen (that bound to carbon) constitutes 75% of the total hydrogen. Of this non-exchangeable hydrogen, a significant portion is located on indispensable (essential) amino acids (25% for collagen, 51% for keratin) which come directly from ingested protein, while the remainder may come from food or from water via amino acid synthesis. By comparing bone collagen D/H from species living solely in Britain, we sought to eliminate climatic effects on animal δD. With a temperate maritime climate, the geographical variation in surface water δD in Britain is small − a range of c. 15‰ over 90% of the mainland (Darling, Bath & Talbot 2003). Any differences in δD between animal species will therefore be related to diet rather than to local water.

Materials and methods

We measured the 15N/14N and non-exchangeable D/H of bone collagen extracted from 97 individuals of 19 species of mammals, birds and fish in Great Britain (Table 1). We chose non-migratory species with only terrestrial or freshwater diets. All individuals died between 1998 and 2001, with samples being obtained from various sources, including academic collaborators, ecological societies, veterinarians and museums. Location of recovery was recorded for all samples.

Table 1.  Species mean bone collagen δD (non-exchangeable hydrogen) and δ15N values
Species*Common nameFigure codeNAquatic/ TerrestrialTrophic statusδD (‰)SD of δDδ15N (‰)SD of δ15N
  • *

    Full details of information sources for species identification given in Birchall (2002).

  • P, piscivore; C, carnivore; O, omnivore; I, insectivore; H, herbivore. First letter indicates predominant feeding habit where two are present. Full details of information sources for diet identification given in Birchall (2002).

  • δDVSMOW = [(D/Hsample/D/HVSMOW) – 1] × 1000; δ15NAIR = [(15/14Nsample/15/14NAIR) – 1] × 1000 (Gonfiantini 1978; Mariotti 1983).

Lutra lutraOtter 16AP/C 103·212·0 15·51·5
Mustela visonAmerican mink 25A/TC/P 65·210·5 12·70·6
Mustela nivalis vulgarisCommon weasel 36TC 69·217·7 7·90·8
Mustela putoriusWestern polecat 45TC 48·6 9·8 7·91·6
Vulpes vulpesRed fox 53TC/O 31·615·2 7·72·5
Sciurus carolinensisGrey squirrel 65TO −5·610·7 2·81·2
Apodemus sylvaticusWood mouse 75TO −8·410·8 4·70·4
Capreolus capreolusRoe deer 85TH −16·1 3·2 3·11·6
Cervus elaphusRed deer 95TH −24·0 6·8 3·00·4
Oryctolagus cuniculusRabbit105TH −34·3 8·2 0·71·5
Clethrionomys glareolusBank vole115TH −15·314·3 −0·51·8
Esox luciusPike125AP/I 43·411·1 11·60·8
Rutilis rutilisRoach137AO −14·510·7 10·42·9
Carassius carassiusCrucian carp145AO −26·4 7·1 13·20·5
Accipiter nisusSparrow hawk155TC 106·825·7 8·01·2
Tyto albaBarn owl165TC 87·015·5 9·70·5
Mergus merganserGoosander175AP 40·424·2 13·70·6
Ardea cinereaGrey heron185A/TP/C 49·225·9 11·32·0
Cygnus olorMute swan195A/TH −49·223·7 13·15·0

Collagen was extracted from each sample using a modified Longin methodology (Longin 1971; Birchall 2002). Mammal and bird bones were cleaned physically by scraping, and fish scales were ultrasonicated repeatedly in deionized and distilled water until cleaned of skin and guanine crystals. Samples were then defatted by two rinses (12 h each) in a 2 : 1 (v/v) mixture of methanol : chloroform at room temperature, then three rinses in water, ultrasonicating for 30 min each time, to remove all traces of solvent. Samples were then left in 0·5 m aqueous hydrochloric acid at room temperature until demineralized (2–14 days), rinsed three times in water, gelatinized in water of pH 3 at 80 ± 5 °C for 24 h, and the soluble collagen lyophilized.

15N/14N analyses were performed by combustion of samples in an automated carbon and nitrogen analyser (Carlo Erba, Turin, Italy) coupled to a Geo 20/20 isotope ratio mass spectrometer (PDZ-Europa, Crewe, UK).

D/H analyses of the non-exchangeable hydrogen in collagen were performed using a method adapted from Cormie et al. (1994a,b) (see Birchall (2002) for full details). Collagen samples were loaded into Pyrex tubes with CuO, and allowed to equilibrate with the vapour of water of known D/H composition, before evacuating and sealing the tubes, and combusting the contents at 550 °C. The resultant water was separated cryogenically, transferred to vessels containing zinc turnings, and the vessels heated to 500 °C to produce H2 gas for D/H analysis in a VG SIRA dual-inlet mass spectrometer. As this H2 analysed represents the total hydrogen in the sample, a correction to yield the δD value of only the non-exchangeable hydrogen was performed by analysing each sample twice, after equilibrating with the vapour of two waters with markedly different D/H ratios.

The 15N/14N and D/H ratios are reported in the conventional manner as δ15N and δD, in units per thousand (‰) against the international standards of AIR for nitrogen and V-SMOW for hydrogen (see footnote to Table 1). Replicate measurement errors on collagen standards were less than ± 0·2‰ for δ15N, and less than ± 4‰ for δD.

Results and discussion

All collagen samples analysed had atomic C : N ratios between 2·9 and 3·6, within the accepted range for uncontaminated/unaltered collagen (DeNiro 1985). The nitrogen and hydrogen isotope ratios for all 19 species of mammals, birds and fish are shown in Table 1. When comparing herbivores and carnivores (Fig. 1), it is apparent that both δ15N and δD register the difference in trophic level, but δD is more discriminating. The species include herbivorous and carnivorous/piscivorous birds, fish and mammals (ruminant and non-ruminant), and the δD of all species conform to the same pattern. For δ15N, there is a 4–5‰ difference between terrestrial carnivores and herbivores compatible with the trophic level enrichment in 15N. Although similar trophic level enrichments have been reported in aquatic systems (Fry et al. 1999), we found little difference between the δ15N of aquatic carnivores/piscivores and aquatic herbivores/omnivores (Fig. 1). In contrast, δD values for carnivores/piscivores differ by about 90‰ from herbivores/omnivores in both terrestrial and aquatic systems. Notably, while aquatic herbivores/omnivores in this study had substantially higher δ15N values than terrestrial herbivores/omnivores, both groups had similar δD values. It appears that δD does not discriminate between the aquatic and terrestrial environments.

Figure 1.

Species mean bone collagen non-exchangeable δD and δ15N values. Species number labels as detailed in Table 1. Error bars are standard errors.

At present we lack a convincing and testable explanation for the observed enrichment (of c. 90‰) of carnivores over herbivores in protein δD. For nitrogen, the widely observed isotopic enrichment with trophic level has been demonstrated to be a direct result of enrichment between dietary intake and body proteins (DeNiro & Epstein 1981; Sponheimer et al. 2003). The full mechanism underlying this has not yet been elucidated, but transamination reactions of amino acids are known to cause isotopic fractionation (Macko et al. 1986). As the carnivores and herbivores analysed in this study are from the same ecosystems, we postulate that the difference between carnivore and herbivore δD supports the hypothesis of a diet-to-body enrichment in carnivores. We have no firm data as to the dietary δD of the herbivores analysed, but because plant leaf tissues are often depleted in D relative to local groundwater (Estep 1980; Hayes 2001), while here herbivore protein δD is similar to groundwater δD (UK average of about −55 to −40‰: Darling et al. 2003), this suggests that herbivores, like carnivores, may also be enriched in D with respect to their diet. Published δD measurements on zooplankton and insect chitin also appear to exhibit isotopic enrichment with trophic level (Stiller & Nissenbaum 1980; Schimmelmann & DeNiro 1986).

When considering possible mechanisms for the observed differences in δD, differences in both diet and metabolism between herbivores and carnivores must be recognized. Our data show most clearly that carnivore body protein is enriched with respect to its dietary protein (i.e. herbivore flesh). An explanation for this may also serve to explain general trophic level enrichment in protein H; alternatively, other processes may be involved in determining the isotopic composition of herbivore protein. For carnivores, non-exchangeable protein hydrogen has two sources: from dietary amino acids that are not catabolized, and from (dispensable) amino acids which are synthesized in vivo from glycolytic and TCA (tricarboxylic acid) cycle intermediates. In the first case, enrichment could occur through selective catabolism of isotopically lighter amino acids. However, any isotopic discrimination would result from a second-order kinetic isotope effect (as deamination or transamination is the first reaction) and we do not know if this would be sufficient. In the second case, the synthesis of amino acids during metabolism enables body water hydrogen to be incorporated, bringing in a second source of hydrogen, and also the possibility of fractionation during incorporation (e.g. from fumarate to malate). The issue here is (a) whether the proportion of carnivore protein which has been synthesized is significant, (b) whether body water provides a source of ‘enriched’ H and (c) whether H incorporation reactions are isotopically discriminating. For (b), the body water pool is a combination of drinking water and that derived from diet during catabolism of dietary macronutrients, and its isotopic signal should be a mixture of the two, with allowance made for fractionation during body water loss (i.e. respiration and transcutaneous losses (Schoeller, Leitch & Brown 1986)). While body water δD and its dependence on diet and drinking water is well studied, particularly with regard to the doubled-labelled-water method of assessing energy expenditure (Schoeller et al. 1986; Nagy 1988), experimental data concerning body protein δD and the relative importance of food and drinking water as hydrogen sources are scarce: the few studies that have attempted to quantify the contributions of diet and drinking water to body protein δD have either not controlled for the contribution of food-derived water to the body water pool, or have been performed over such short time-scales so as not to allow sequestration within tissues (Hobson et al. 1999; Sharp et al. 2003). As to (c), whether H incorporation reactions are isotopically discriminating, no specific data exist, although one would expect depleted water to be preferentially incorporated. We are not aware of any studies that address (a), the extent to which carnivore protein amino acids are synthesized in situ, and thus we cannot speculate on whether this can be a significant mechanism.

The same arguments can be used to consider possible enrichment of D in herbivore body protein. However, in this case it is clear that much of the non-essential amino acid content in protein will have been synthesized in vivo (including during bacterial fermentation in the gut), and that the role of body water will be correspondingly more important.

The general finding of δD distinctions between herbivores and carnivores over a wide range of habitats has several implications. Currently it challenges metabolic explanations, suggesting that further research should lead to better understanding and quantification of whole body hydrogen isotope dynamics (e.g. in labelled water studies in bioenergetics, or in dietary validation and metabolic studies). Whatever the metabolic explanation, this pioneer study suggests that while 15N/14N ratios can be affected by both trophic level and environment type, within a localized ecosystem the D/H ratio of animal collagen is a more discriminating indicator of trophic level, being independent of habitat type. δD analyses can therefore provide an additional tool in disentangling food web patterns in both ecology and palaeodietary studies − for example, where δ15N measurements give ambiguous interpretations about an individual's trophic level.


J. B. thanks NERC and English Heritage for a Case studentship. Thanks to Martin Humm at the Radiocarbon Accelerator Unit, University of Oxford, and Carol Arrowsmith at NIGL at Keyworth for technical assistance, and to all who supplied samples for analysis.