Hydrogen isotope assimilation and discrimination in green turtles

The use of stable isotope analysis (SIA) to characterize habitat and resource use by organisms has grown exponentially (Peterson and Fry, 1987; Post, 2002; Solomon et al., 2011) and is now routinely used to assess diet composition (Burgett et al., 2018; Díaz-Gamboa et al., 2017), the flow of energy within and among ecosystems (Doucett et al., 1996; Finlay and Kendall, 2007), trophic level and food chain length (Boecklen et al., 2011; Post, 2002), and movement/migration patterns (Hobson and Wassenaar, 2018; Zbinden et al., 2011). In particular, studies of individual- and population-level foraging strategies have greatly benefited from SIA, which provides information on resource assimilation over a variety of timescales depending on the type of tissue analyzed (Martínez del Río et al., 2009; Vander Zanden et al., 2013). This is one of the primary advantages of SIA in comparison to conventional methods that directly assess diet composition, such as observation of gut/scat content analysis, which only provides dietary ‘snapshots’ and may not accurately reflect variation in how different resources are assimilated by a consumer (Hobson, 1999; Newsome et al., 2010).

Carbon isotope (δ¹³C) and nitrogen isotope (δ¹⁵N) analyses have been most frequently used by animal ecologists to characterize resource and habitat use (DeNiro and Epstein, 1978; Hobson, 1999; Phillips, 2012), while hydrogen isotope (δ²H) analysis has been primarily used to track the geographic origin and seasonal movement of birds (Chamberlain et al., 1996; Hobson, 2005a), insects (Flockhart et al., 2017; Hobson et al., 2012) and even humans (Ehleringer et al., 2008) across regional to continental scales. Only a few studies have used δ²H as a tracer of food resources, specifically to quantify autochthonous (instream algae) and allochthonous (riparian plants) resources in freshwater ecosystems (Berggren et al., 2014; Doucett et al., 2007; Finlay et al., 2010), or to coarsely assess trophic level (Birchall et al., 2005; Solomon et al., 2009).

Although the use of δ²H in ecological studies has increased rapidly over the past decade, a persistent challenge is to understand the ecological and physiological factors that contribute to variation in consumer tissue δ²H values. Generally, as carbon and nitrogen isotopes move up a food chain, they are sorted by organisms in predictable ways, which leads to offsets in isotope values between the tissues of consumers and their diet, commonly referred to as trophic discrimination and denoted by Δ¹³C or Δ¹⁵N (Martínez del Río et al., 2009; Newsome et al., 2010). To expand the use of δ²H analysis to study animal ecology, we must better understand how animals assimilate and sort hydrogen isotopes from the two distinct sources available for metabolism and tissue synthesis: food and body water (Ehleringer et al., 2008; Wolf et al., 2011, 2013; Fig. 1). This requires knowledge of (1) the proportion of the hydrogen in tissues from food and body water, and (2) the offset in δ²H values of consumers and these two sources (i.e. trophic discrimination factors, Δ²H_Net). In regard to the first challenge, controlled feeding experiments show that the majority (∼70–80%) of hydrogen in consumer tissues is derived from diet, whereas only ∼20–30% is sourced from environmental or drinking water (Rodriguez Curras et al., 2018; Newsome et al., 2017; Solomon et al., 2009; Wolf et al., 2011). Fewer studies have focused on quantifying Δ²H_Net, which is mediated by physiological processes during resource assimilation, tissue synthesis and excretion. Similar to patterns observed in Δ¹³C and Δ¹⁵N, variation in Δ²H is likely influenced by tissue amino acid composition and the potential for disproportionate routing of dietary protein to tissue synthesis (Ambrose and Norr, 1993; Schwarcz and Schoeninger, 1991). A suite of studies have revealed that consumer tissues generally have δ²H values greater than food sources but lower than environmental water (Peters et al., 2012; Solomon et al., 2009), a pattern that results from the mixing of these two sources and causes apparent increases in δ²H values with increasing trophic level (Birchall et al., 2005; Hobson, 2005b; Magozzi et al., 2019).

Hydrogen isotopes have the potential to provide numerous ecological inferences. For example, significant variation in δ²H values among marine primary producers (Estep and Dabrowski, 1980; Smith and Epstein, 1970) suggests that hydrogen isotopes may be useful for quantifying resource and habitat use in marine ecosystems (Ostrom et al., 2014), especially in nearshore ecosystems fueled by a combination of phytoplankton, macroalgae, seagrass and even terrestrial plants (e.g. mangroves). In addition, more recent work has identified predictable δ²H discrimination among primary producer sources in marine ecosystems (Hondula and Pace, 2014), similar to that observed in freshwater aquatic and terrestrial food webs (Birchall et al., 2005; Peters et al., 2012; Soto et al., 2013). Thus, δ²H may complement δ¹³C and δ¹⁵N data to better quantify the relative contributions of different sources of primary production and characterize food web dynamics in coastal marine ecosystems. However, before ecologists can confidently apply this tool to assess resource and habitat use by wild populations, additional controlled feeding experiments are needed to better understand the factors that influence hydrogen isotope assimilation and discrimination in a variety of marine consumers and their tissues.

Here, we investigate hydrogen isotope assimilation and discrimination in an omnivorous marine ectotherm by sampling tissues from captive green turtles, Chelonia mydas (Linnaeus 1758), fed controlled diets in two separate feeding experiments. Our objective was to estimate trophic discrimination factors (Δ²H_net) between the sources of hydrogen available for tissue synthesis and green turtle skin (epidermis), blood serum/plasma and red blood cells. We analyzed the isotopic composition of dietary macromolecules and used a mass balance mixing model framework to examine the factors that cause variation in δ²H values among tissue types and life stages. We anticipate that our results will help refine the use of δ²H values as a tool to study resource and habitat use by marine consumers that inhabit nearshore ecosystems characterized by large variation in δ²H values among primary producers.

We analyzed green turtle tissues and diet samples collected during two previous controlled feeding experiments described in Seminoff et al. (2006) and Vander Zanden et al. (2012). For convenience, we refer to Seminoff et al. (2006) as Feeding Experiment 1 (FE1) and Vander Zanden et al. (2012) as Feeding Experiment 2 (FE2). In FE1, turtles were kept in a large oval fiberglass tank (10×3×1.5 m) filled with seawater. The water temperature was maintained at 24±1°C. Fluorescent light fixtures suspended above each tank provided full-spectrum radiation for 12 h each day; the tanks were also exposed to ambient light. Water quality was maintained between the following levels: pH=8.0–8.3, salinity=33–35 and ammonia <0.1 mg⁻¹. Monthly water changes pumped directly from the ocean prevented accumulation of high levels of ammonia, bacteria and fungi. Salinity was maintained by adding locally sourced fresh water. In FE2, turtles were kept in unshaded 40,000-gallon concrete tanks that received 400–1100 l min^–1 of unfiltered saltwater pumped directly from the ocean. Both feeding experiments used green turtles hatched at Cayman Turtle Center (Grand Cayman, Cayman Islands). However, turtles from FE1 were transferred to the University of British Columbia (Vancouver, Canada) as yearlings. In FE1, eight juvenile green turtles with a mean (±s.d.) body mass of 11.7±0.7 kg and mean straight carapace length (SCL) of 45.2±1.2 cm were fed a pellet diet composed of 41% protein, 12% lipids and 47% carbohydrates by weight (Aquamax Grower 500; PMI Nutrition International, Brentwood, MO, USA) for 619 days (Table 1). Samples of the pelleted diet, epidermis and blood components – including whole blood, blood plasma and red blood cells – were collected at the end of FE1. In FE2, 70 green turtles (30 adults and 40 juveniles) were fed a constant pellet diet containing 35.5% protein, 4.2% lipids and 60.3% carbohydrates by weight (Southfresh Feeds, Demopolis, AL, USA) for 1460 days (Table 1); all adults in this experiment were sexually mature females. Adults had a mean (±s.d.) body mass of 125±29.2 kg and a mean curved carapace length (CCL) of 101±5.2 cm; juveniles had a mean body mass of 42±8.1 kg and a mean CCL of 73±5.5 cm. Samples of pelleted diet, epidermis, serum and red blood cells were collected at the end of FE2.

Dietary lipid and protein δ²H values were obtained via lipid extraction followed by hydrolysis and cation exchange separation. Lipids were extracted from homogenized diet pellets via three ∼24-h soaks in petroleum ether. After each soak, the removed petroleum ether was saved in a separate glass scintillation vial and transferred to pre-weighed silver capsules for δ²H lipid analysis. Lipid-extracted diet samples were then rinsed with deionized (DI) water to remove any residual solvent and freeze dried. Approximately 6–7 mg of the lipid-extracted diet samples were hydrolyzed in 6 N HCl for 20 h at 110°C, breaking proteins into constituent amino acids, then dried under a flow of N₂ gas at 110°C; the HCl used for hydrolysis had a δ²H value of −95‰. A cation exchange resin (Dowex 50WX8, 100–200 mesh) was then used to separate the amino acids from carbohydrates (Amelung and Zhang, 2001). The hydrolyzed sample was added to a column containing Dowex resin and then washed with 3–4 ml of 0.01 N HCl solution. Subsequently, 4 ml of 2 N NH₄OH was added to the column, and the effluent containing amino acids representing dietary proteins was collected in another pre-combusted glass vial. The amino acid fraction was dried down at 80°C under N₂ and temporarily re-suspended in DI water for transfer to silver capsules, which were then thoroughly dried at ∼45°C prior to δ²H analysis (Amelung et al., 1996; Andrews, 1989). We measured δ²H values of four to six replicates of the lipid and protein fractions isolated from each diet. δ²H values for the carbohydrate portion of each diet were estimated with a mixing model using measured δ²H values for dietary lipids and protein in combination with data on the relative weight percent proportion of these macromolecules in each experimental diet (Table 1). Note that dietary macromolecular proportions in Table 1 sum to 100%, but they need to be multiplied by 0.8 to calculate Δ²H_Net using Eqns 3–5 (see below) to account for the ∼20% contribution of hydrogen from environmental water to tissue synthesis (Fig. 1). We acknowledge that this model does not account for potential sources of hydrogen from vitamins and minerals in each diet that could not be isolated and analyzed for their δ²H composition, but they represent minor sources in comparison to the major macromolecular components (protein, carbohydrates, lipids) that we directly measured or modeled.

Samples of blood serum/plasma and red blood cells were dried at 60°C for 24 h and homogenized with a mortar and pestle. Epidermis samples were rinsed with DI water, dried at 50°C for 24–48 h, and then lipid extracted in three sequential 24-h soaks with petroleum ether, followed by thorough rinsing in DI water before being dried at 50°C for ∼48 h (Seminoff et al., 2006). To evaluate the influence of lipids on δ²H values, we compared δ²H values of epidermis (N=20; 10 juveniles and 10 adults) from FE2 before and after lipid removal.

Approximately 0.2–0.3 mg of sea turtle tissue or pelleted diet samples were weighed into 3.5×5 mm silver capsules sealed for δ²H analysis. Proteins can exchange ∼10–20% of their hydrogen with ambient water vapor (Bowen et al., 2005; Coplen and Qi, 2012), so we used a bench top equilibration method in which samples and associated in-house reference materials used to correct for exchangeable hydrogen sat for at least 3 weeks before analysis to ensure equilibration with local water vapor (Bowen et al., 2005; Wassenaar and Hobson, 2003).

To correct for exchangeable hydrogen, we used a series of keratin in-house reference materials for which the hydrogen isotope composition of the non-exchangeable portion of the tissue (δ²H_{non-exchangeable}) had previously been estimated via room-temperature-exchange experiments identical to those used by Bowen et al. (2005). δ²H_{non-exchangeable} values for the three keratin internal reference materials ranged from −55‰ to −175‰. We used a series of keratin external standards (KHS and CBS) to check the values obtained by these internal reference materials. Muscle and whole blood from cows raised in Wyoming and Florida, USA, were also used to compare δ²H in similar tissues, which ranged from −70‰ and −150‰ for muscle and −82‰ and −160‰ for blood, respectively. We used the oil standard NBS-22 (δ²H: –120‰) to correct lipid samples extracted from pelleted diet because lipids do not contain exchangeable hydrogen. All standards had a within-run hydrogen isotope variation (s.d.) of ≤3‰. Weight percent [H] values were measured via analysis of powdered benzoic acid analyzed in each run of unknown sea turtle tissue or diet samples.

We used one- and two-way analysis of variance (ANOVA), followed by post hoc Tukey honest significant difference (HSD) pairwise comparisons with Bonferroni correction to evaluate differences between tissues for Δ²H_Net in experiment FE1. For experiment FE2, we used a two-way ANOVA with two factors: age (juvenile or adult) and tissue (epidermis, EPI; serum/plasma, SER/PLA; red blood cells, RBC) corrected for unbalanced data (lsmeans), followed by pairwise comparison using the Bonferroni correction. When comparing between experiments, we used a two-way ANOVA with two factors: experiment (FE1 or FE2) and tissue (EPI, SER/PLA, RBC) corrected for unbalanced data (lsmeans), followed by pairwise comparison using Bonferroni correction. We used a Max-t-test to compare the mean δ²H values of bulk and lipid-extracted diet and diet macromolecules (proteins and lipids) between FE1 and FE2 (Herberich et al., 2010). This procedure is useful to assess datasets that include unbalanced group sizes, non-normal distributions and non-heteroscedasticity (Herberich et al., 2010). All models were run using R software (https://www.r-project.org/).

Table 1 presents mean (±s.d.) δ²H values, [H] and relative proportions (weight %) for bulk diet, lipid-extracted bulk diet, and dietary protein and lipids in the two feeding experiments. Mean δ²H values of bulk diets (t=5.3, P=0.001) and dietary lipids (t=3.4, P=0.010) were higher in FE2 than in FE1. Mean δ²H values of lipid-extracted bulk diets were lower in FE2 than in FE1 (t=−3.5, P=0.007). Mean δ²H values of dietary protein were not different between diets (t=−1.4, P=0.19).

Results of the sensitivity analysis revealed that changing p_Water from 15% to 30% or varying the δ²H value of this source of hydrogen by 10‰ resulted in an overall difference in estimated Δ²H_Net within tissues or among feeding experiments of 5–6‰. These values are slightly higher than analytical precision (3–4‰) of δ²H measurements of organic materials via TCEA-IRMS and are similar to the observed variation in δ²H within tissues, but much smaller than differences in δ²H among tissues (Table 2).

We combined isotope data for blood plasma and serum because these blood components had similar δ²H values. Epidermis δ²H values were higher than values for blood components in FE1 (F=47.2; d.f.=31, P<0.001; Fig. 2, Table 2). By contrast, δ²H values of all tissues differed within and between life stages (F=560.9, d.f.=204, P<0.001; Fig. 2, Table 2) in FE2, with the exception of red blood cells and serum/plasma for the juvenile group. Adults had lower serum/plasma δ²H values and higher weight percent [H] values than juveniles in FE2 (Fig. 3, Table 2), which resulted in a significant negative relationship between serum/plasma δ²H values and weight percent [H] (R²=0.88; P<0.001; Fig. 3). The only tissue that differed between feeding experiments was epidermis, which was slightly but significantly higher in juveniles in FE2 (F=313, d.f.=143, P<0.001; Fig. 2, Table 2). Lastly, there were no significant differences in mean (±s.d.) δ²H or weight percent [H] values between bulk {δ²H: −74±8; [H]: 5.6±0.1} and lipid-extracted epidermis {δ²H: −73±6; [H]: 6.0±0.1} (t=−0.27, d.f.=38, P=0.79).

Epidermis had significantly higher mean Δ²H_Net values than blood components in FE1 (F=53.7, d.f.=23, P<0.001; Fig. 4, Table 2). In FE2, multiple pairwise comparisons between life stage and between tissues show that all tissues except serum/plasma and red blood cells in juveniles had significantly different Δ²H_Net values (F=572.9, d.f.=205, P<0.001). Mean Δ²H_Net values in epidermis and serum/plasma were higher in juveniles than in adults, whereas in red blood cells the opposite pattern was observed (Fig. 4, Table 2). Estimates of Δ²H_Net for juveniles were slightly but significantly different between FE1 and FE2 experiments for serum/plasma and red blood cells (F=315.5, d.f.=143, P<0.001), but similar for epidermis (F=315.5, d.f.=143, P=0.14).

We investigated the potential for using δ²H values as an ecological tracer in marine consumers by providing the first quantification of hydrogen isotope discrimination factors (Δ²H_Net) in four tissues from juvenile and adult green turtles using samples collected during two previous controlled feeding experiments (Seminoff et al., 2006; Vander Zanden et al., 2012). In the following sections, we assess the factors that influence Δ²H_Net and discuss possible physiological mechanisms responsible for the observed differences among diet treatments and age classes. Overall, our results corroborate patterns observed in previous controlled feeding experiments on other taxa (Rodriguez Curras et al., 2018; Fogel et al., 2016; Newsome et al., 2017; Wolf et al., 2013), and we anticipate that these data will enable ecologists to accurately apply δ²H isotopes to study diet composition, habitat use and movement in marine consumers.

When comparing results from different feeding experiments, it is important to consider differences in dietary macromolecule composition and the proportional contribution of food and body water to tissue synthesis, both of which have been shown to influence tissue δ²H and associated Δ²H_Net values. Our models assumed that ∼80% and ∼20% of the hydrogen in green turtle tissues was synthesized from diet versus water (Fig. 1), respectively, which is similar to contributions found in a wide variety of vertebrate taxa (Rodriguez Curras et al., 2018; Hobson et al., 1999; Newsome et al., 2017; Soto et al., 2013). Given the relatively small contribution (20%) of hydrogen from water that is used to synthesize tissues, calculations from our sensitivity analysis showed that a small variation (5–10‰) in our estimated seawater δ²H values did not heavily impact our estimates of Δ²H_Net. With respect to diet composition, the diets used in the two feeding experiments were similar in terms of macromolecular composition and bulk diet δ²H values (Table 1), and green turtle tissues had remarkably similar δ²H values for nearly all tissue types (Fig. 2); the only exception to this was blood serum/plasma for adult turtles in FE2 (Fig. 3). Finally, previous δ²H-based work on tropical seabirds (Ostrom et al., 2014) has hypothesized that variation in feather hydrogen isotope values may be caused by isotopic fractionation associated with water loss during salt excretion via the nasal salt glands. We assume that this mechanism has a negligible isotopic effect on body water of sea turtles because they have (1) a near-fully aquatic lifestyle, (2) a large body size that dampens evaporative effects, and (3) high water flux rates relative to their body size with complete turnover of body water in ∼10 days (Jones et al., 2009). In addition, baseline ²H concentrations of turtles in FE1 reported in Jones et al. (2009) show that body water δ²H values are similar to those of local seawater. Unlike seabirds, sea turtles are considered seawater drinkers with water turnover rates as high as ∼6% of total body water per day when fasting, suggesting that water turnover is primarily driven by imbibing seawater (Jones et al., 2009).

Green turtle epidermis had higher mean δ²H values relative to blood components by ∼20‰ in both feeding experiments (Fig. 2), which is similar to patterns in δ²H values of keratinaceous tissues (feathers and claws) and blood components observed in feeding experiments on birds (Hobson et al., 1999; Wolf et al., 2011, 2012, 2013). There are two possible explanations for these patterns. First, previous feeding experiments reporting δ¹³C and δ²H data have suggested that such tissue-specific isotopic patterns are likely to be driven by differences in the amino acid composition among tissues (Rodriguez Curras et al., 2018; Newsome et al., 2017; Wolf et al., 2012). Hydrogen isotope values of individual amino acids within a single organism can vary by 200–300‰ (Fogel et al., 2016); thus, subtle changes in a tissue amino acid composition could easily drive variation in bulk tissue δ²H values (Rodriguez Curras et al., 2018; Newsome et al., 2017). A second potential explanation for the observed tissue-specific δ²H patterns between epidermis and blood components is that the hydrogen in blood is directly routed from dietary protein, which had lower δ²H values than that of bulk diet or dietary carbohydrates in both experiments (Table 1 and Fig. 2). Here, we define protein routing as the preferred use of dietary protein to synthesize proteinaceous tissues, which decreases the costs of de novo synthesis of non-essential amino acids from dietary carbohydrates and/or lipids. Protein routing has been observed in other δ²H-based controlled feeding experiments (Rodriguez Curras et al., 2018; Newsome et al., 2017); however, the animals in those experiments were fed diets that had much lower weight percent protein contents than the diets fed to green turtles in our experiments, which was exceptionally high (35.5–41.0%) for an omnivore. Because dietary protein was readily available, the potential for protein routing to particular tissues (e.g. epidermis) was likely minimal in our experiments. Furthermore, tissue-specific patterns in δ²H values are similar among a wide range of unrelated taxa (birds, mammals and reptiles), and thus the first potential explanation above involving variation in amino acid composition among tissues is the most parsimonious reason for the observed differences in δ²H values among green turtle tissues.

The significantly lower serum/plasma δ²H values observed in adult female versus juvenile turtles from the FE2 experiment (Vander Zanden et al., 2012) are likely to be related to elevated lipid content in this tissue type. Lipids have relatively high weight percent [H] contents (11–12%) and lower δ²H values in comparison to associated proteins owing to a large isotopic discrimination during the formation of acetyl CoA from pyruvate (Hayes, 2001; Schmidt et al., 2003; Sessions and Hayes, 2005). Thus, the significant negative relationship between δ²H values and weight percent [H] observed in serum/plasma (Fig. 3) supports the idea that these samples had higher lipid content than serum/plasma collected from juveniles. Because all adults in our experiment were sexually mature females, higher concentration of free lipids in blood serum/plasma could be related to reproduction (i.e. follicle development, egg production). Females mobilize lipids prior to and throughout the nesting season to help with vitellogenesis (yolk deposition). In this process, lipids and proteins are stored in the oocytes during egg formation (Milton and Lutz, 2003) and other related reproductive processes (Hamann et al., 2002). Thus, the relationship between serum/plasma δ²H and [H] could potentially be used as a proxy to identify sexually mature females in the field. Finally, comparison of δ²H for lipid and non-lipid extracted epidermis suggests that this tissue has sufficiently low lipid content such that lipid extraction prior to hydrogen isotope analysis is unnecessary. These results mirror those found in previous studies that showed that δ¹³C values of sea turtle epidermis were unaffected by lipid removal (Turner Tomaszewicz et al., 2017; Vander Zanden et al., 2012, 2014).

By definition (Eqn 5), patterns in tissue-specific trophic discrimination factors are influenced by tissue hydrogen isotope values; therefore, patterns in Δ²H_Net typically mirror those in tissue δ²H values. Thus, the observed differences in Δ²H_Net values between epidermis and blood components could simply reflect differences in tissue-specific discrimination that we suggest are likely driven by variation in amino acid composition among these tissues. Δ²H_Net estimates are also sensitive to the proportional contribution of dietary macromolecules (Eqn 4), especially protein because it can be directly routed from diet and thus could contribute more to proteinaceous tissue synthesis relative to its dietary content. Previous feeding experiments on tilapia (Oreochromis niloticus; Newsome et al., 2017) and house mice (Mus musculus; Rodriguez Curras et al., 2018) showed that when animals were fed a low-protein diet (5–10%), 34–48% of the hydrogen in proteinaceous tissues (muscle and liver) was derived from dietary protein. In this study, turtles in both experiments were fed diets with a high protein content (35.5–41.0%), and thus protein routing was likely minimal in comparison to the previous work mentioned above. Overall, our estimates of Δ²H_Net should be considered preliminary because the protein content of the diets used in FE1 and FE2 are ∼3–4 times higher than those consumed by wild green turtles in the eastern Pacific Ocean or Caribbean (Bjorndal, 1980; López-Mendilaharsu et al., 2005; Seminoff et al., 2002). Our previous work shows that the relative importance of protein routing versus de novo synthesis of non-essential amino acids from dietary carbohydrates and lipids has an important influence on Δ²H_Net (Newsome et al., 2017; Rodriguez Curras et al., 2018). As such, our study provides a valuable first approximation of the range in Δ²H_Net values for multiple tissues of an omnivorous marine ectotherm and provides a framework for investigating this important parameter in other marine species.

Our study is an initial step for expanding the use of δ²H analysis to evaluate resource and habitat use of marine consumers that inhabit coastal habitats, especially those that are fueled by a combination of marine (micro/macroalgae and seagrass) and terrestrial producers (e.g. mangroves). The hydrogen isotope composition of primary producers in coastal marine ecosystems can vary by more than 100‰ (Estep and Dabrowski, 1980), providing a sufficient degree of natural variation to trace resource use by consumers with δ²H analysis. We suggest that additional feeding experiments are needed to better understand variation in Δ²H_Net among species so that mixing models can be accurately applied to quantify diet composition and energy flow in marine food webs. When possible, feeding experiments should vary diet quality (e.g. protein content) but attempt to include diets that mimic the macromolecular composition of prey consumed by wild populations. In addition to designing more realistic feeding experiments, direct δ²H measurement of body water distilled from blood would help refine our understanding of the processes that influence discrimination of hydrogen isotopes in animals and enable us to broaden our use of this tool to study the ecology of wild populations of marine consumers.

This research was supported by staff from NOAA Fisheries, particularly Joel Schumacher and Garrett Lemons, who provided assistance with database and sample logistics. We thank John P. Whiteman, Emma A. Elliot Smith and Mauriel Rodriguez Curras for lab assistance, and Viorel Atudorei and Laura Burkemper of the UNM-CSI for analytical support. We also thank Bryan Wallace for his input on the development of conceptual models.