The influence of drinking water on the δD and δ¹⁸O values of house sparrow plasma, blood and feathers

It is often stated by isotope ecologists that animals are what they eat plus a discrimination factor. This observation recognizes that animals incorporate the isotopic composition of their resources, but that the match between the isotopic makeup of animal's tissues and its diet is rarely perfect (Bearhop et al., 2002; Pearson et al., 2003). Knowing the tissue-to-diet discrimination factors (denoted by ΔXδX_tissues–δX_diet, where X is an element) between an animal's tissues and its food sources is an essential aspect of most stable isotope applications in ecology (reviewed by Martinez del Rio et al., 2009b). For hydrogen and oxygen, discrimination factors are not measured or expressed in the traditional sense as the difference in delta values between consumer and diet (McCutchan et al., 2003). Instead they are reported as the difference in isotopic composition of an animal's tissue (typically bird feathers) and the delta value of local precipitation [δdeuterium (D)_{feather–precipitation} or Δ¹⁸O_{feather–precipitation} (Hobson, 2005)]. This usage is unorthodox but reflects researchers' interests in using hydrogen and oxygen isotopes to measure animal movements. If the discrimination factors for hydrogen and/or oxygen are known, we can calibrate the relationship between the isotopic composition of a tissue and the isotopic composition of precipitation in an ‘isoscape’ (sensuWest et al., 2006) to determine where the tissue was grown (Hobson, 2008; Hobson et al., 2009). Wunder describes in detail the state-of-the-art methods currently used to assign a likely site of origin for a sample given its isotopic composition (Wunder, 2010).

For deuterium, discrimination values are often estimated from the linear relationship between the delta values of feathers of wild-caught birds and the average value of growing season precipitation at the site of feather growth estimated from modeled values derived from long-term (40+ years) delta values of precipitation available in the International Atomic Energy Agency – Global Network of Isotopes in Precipitation (IAEA-GNIP) database (reviewed by Hobson, 2008). Wunder highlighted the importance of calibrating the isotopic value of feathers for correct geographical assignment and identified the complexities involved in these calibrations. He also emphasized how mechanistic understanding can clarify some of the issues involved in the estimation of tissue to precipitation discrimination factors (Wunder, 2010).

One of the important issues that clouds the estimation of tissue δD_{feather–precipitation} or Δ¹⁸O_{feather–precipitation} values is the existence of multiple sources for hydrogen and oxygen in organic matter. Hobson et al. observed that the source of hydrogen in the organic molecules that make up animal tissues can be ultimately derived from two potential sources: the hydrogen in the organic compounds in food and pre-formed water (including free drinking water and water in food) (Hobson et al., 1999). Similarly, Hobson et al. noted that the possible sources for oxygen in the organic molecules that make up animal tissues are food, pre-formed water (including free drinking water and water in food), and air (Hobson et al., 2009). Hobson et al. in their previous paper also noted that sometimes there are large differences in the isotopic signatures between drinking water and food at a single locality and speculated that this disparity could lead to variation in the relationship between the isotopic content of an animal's tissues and that of the precipitation of the area in which this animal lives (Hobson et al., 1999).

Several experiments have attempted to measure the relative contribution of preformed water and diet to the hydrogen and oxygen in animal tissues. Hobson et al. reported that only 18–32% of the hydrogen in bird tissues (blood, muscle, liver, lipids, feathers and claws) was derived from preformed water (Hobson et al., 1999). Wang et al., Solomon et al. and Nielson and Bowen reported similar ranges (20–39%) for aquatic invertebrate tissues (including chitin produced by chironomids, mosquitoes and brine shrimp) and muscle from salmonid fishes (12%) (Wang et al., 2009; Solomon et al., 2009; Nielson and Bowen, 2010). Wang et al. and Nielson and Bowen also found that 70% of the oxygen in chitin produced by larval chironomids and brine shrimp was derived from water (Wang et al., 2009; Nielson and Bowen, 2010). This value is approximately twice as high as the values found in two previous studies investigating human hair keratin [35% (Ehleringer et al., 2008) and 27% (O'Brien and Wooller, 2007)].

The methods used by Hobson et al., Solomon et al. and Wang et al. rely on a basic mixing model, which does not account for several potential process that can affect the hydrogen and oxygen signatures of tissues (Hobson et al., 1999; Solomon et al., 2009; Wang et al., 2009). For example, a simple two-end mixing model does not account for the potential contribution of atmospheric oxygen to the isotopic composition of body water (Gretebeck et al., 1997) and, in terrestrial animals, for the potential effect of an isotopically fractionated loss of body water (McKenchnie et al., 2004). Ehleringer et al. and Bowen et al. developed a more complete, but also more complex, model to explain the relationship between δD and δ¹⁸O in human hair and local tap water (Ehleringer et al., 2008; Bowen et al., 2009). Unfortunately, despite recent advances (Nielson and Bowen, 2010), some of the many parameters of this model are unknown for most species, and can be difficult to measure.

Rather than attempting to estimate the fractional contribution of drinking water, we adopted a phenomenological strategy and tried to determine the relationship between the isotopic composition of drinking water and the δD and δ¹⁸O of bird tissues. We offered drinking water with contrasting δD and δ¹⁸O values to house sparrows (Passer domesticus) fed on the same diet. Based on the results of the studies discussed above we expected a linear relationship between the δD and δ¹⁸O of drinking water and bird tissue.

Fifteen house sparrows, Passer domesticus L. [initial body mass 23.6±2.0 g, final body mass 24.5±2.4 g; means ± s.d.) were captured with mist nets in Albany County, WY, USA in October and November 2007. The birds were marked with unique combinations of colored leg bands and housed in groups of three per cage (0.6 m×0.6 m×0.6 m) at the University of Wyoming's Animal Care Facility at the Department of Zoology and Physiology. Throughout the experiment, the birds were housed at 26.0°C. This temperature is within the thermal neutral zone (25.0–28.0°C) reported for house sparrows (Nzama et al., 2010). For the first 180 days, all five groups received the same millet (Pennisetum glaucum) diet (δD=–65.3±6.1‰, δ¹⁸O=32.5±0.4‰) and drinking water (δD=–133.5‰, δ¹⁸O=–17.4‰) ad libitum. Following this 180-day equilibration period, the birds were randomly assigned to one of five treatment groups of three birds each. The five treatment groups were fed the same millet diet, but each group received a drinking water treatment with different δD and δ¹⁸O values [δD(δ¹⁸O) equaled –132.0(–17.4)‰, –96.7(–15.1)‰, –59.0(–12.3)‰, –23.3(–10.2)‰, 8.3 (–8.1)‰]. The coupled variation in δD and δ¹⁸O values of these waters approximates the natural variation in environmental waters lying along the Global Meteoric Water Line (Craig, 1961). The δD and δ¹⁸O of the drinking water in each treatment were achieved by adding 99 atom percent deuterated water (D₂O) and 98 atom percent ¹⁸O water (H₂O¹⁸) to distilled local tap water.

Following the water switch, we took weekly blood samples from each bird until the tissues reached constant asymptotic values (16 weeks). Blood samples (≈50 μl) were taken from the brachial vein using a 0.5-ml syringe with a 30-gauge needle, and centrifuged (2 min in a microhematocrit centrifuge) to separate red blood cells and plasma. Red blood cells and plasma were dried to a constant mass in an oven at 50°C, ground into a fine powder and loaded into silver capsules (<0.1–0.2 mg). Samples, packed in capsules, were equilibrated alongside protein reference materials (two horse hair keratins) with known non-exchangeable δD values (horse hair keratin 1: δD=142.0±2.0‰ and δ¹⁸O=4.5±0.3‰; horse hair keratin 2: δD=–76.0±2.0‰ and δ¹⁸O=13.1±0.3‰) for at least 1 week, then dried under vacuum for 1 week prior to analysis. The D/H and ¹⁸O/¹⁶O ratios of the samples were analyzed at the Purdue University Stable Isotope Facility (West Lafayette, IN, USA) using a Zero Blank Auto-sampler (Costech Analytical, Valencia, CA, USA) interfaced with a thermo-chemical elemental analyzer (Thermo Fisher Scientific, Waltham, MA, USA) and a Delta Plus XP isotope-ratio-monitoring mass spectrometer (IRMS; Thermo Fisher Scientific). Co-analyzed reference materials were used to calibrate the measured values to the Vienna standard mean ocean water–Vienna standard light Antarctic precipitation (VSMOW–VSLAP) reference scale and achieve an approximate correction for exchangeable hydrogen present in the sample matrix, following the guidelines proposed by Wassenaar and Hobson (Wassenaar and Hobson, 2003) and Bowen et al. (Bowen et al., 2005). We followed these guidelines strictly to avoid the differences between laboratory results reported by Smith et al. (Smith et al., 2009).

Once the sparrow blood samples had reached constant asymptotic values, we plucked several tail feathers from each bird and allowed them to re-grow. These new-grown feathers were sampled to determine the contribution of drinking water to the D and O¹⁸ signatures of feathers, the most widely used tissue in isotope-based avian movement studies. Feathers were cleaned with distilled water and dried [see Paritte and Kelly (Paritte and Kelly, 2009) for discussion of the effect of cleaning methods on feather δD values]. We used surgical scissors to remove the feather barbs from the rachis. The barbs were then minced and homogenized. The rachis was not included in the isotope analyses. Feathers were prepared and analyzed for D/H and ¹⁸O/¹⁶O ratios using the same methods as used for the blood samples. Replicate analyses of a commercial reference keratin (δD=–115.0‰ and δ¹⁸O=11.1‰) were run throughout the sample analyses and yielded an accuracy and precision of δD=–114.9±2.0‰, δ¹⁸O=11.1±0.3‰.

To assess whether the delta values of bird tissues had achieved a relatively constant value, we used repeated measures ANOVA on the δD and δ¹⁸O measurements of red blood cell samples collected on weeks 13–16 after the drinking water shift. Before doing these analyses we tested whether the covariance matrix satisfied sphericity assumptions using Mauchly's criterion (Mauchly, 1940). We used standard least squares linear regression to establish the relationship between the isotopic composition of drinking water and the asymptotic value (week 16) of tissues. Unless otherwise stated, average values are accompanied by standard errors. All statistical analyses were performed using JMP 8 (SAS Institute Inc., Cary, NC, USA).

There were no significant differences in red blood cell δD or δ¹⁸O for weeks 13–16 (δD: F_(3,48)=0.26, P=0.85; δ¹⁸O: F_(3,48)=0.52, P=0.67). Two birds died before week 13. Consequently, we assumed that the red blood cell δD and δ¹⁸O values for these birds had not reached asymptotic values, and did not include any data from these birds in our analyses. With only one exception (δ¹⁸O in plasma), the δD and δ¹⁸O values of bird tissues were linearly related to the hydrogen and oxygen isotopic values of drinking water (Fig. 1). The slope of the relationship between the delta value of a tissue and that of drinking water has been interpreted as a first order approximation of the percentage contribution of drinking water to the tissue hydrogen and oxygen content (see Ehleringer et al., 2008). The estimated percentage contributions (±s.e.m.) of drinking water to the δD values of plasma, red blood cells and feathers were 17±4%, 14±2% and 18±3%, respectively (Fig. 1). For oxygen, only red blood cell and feather δ¹⁸O were significantly linearly correlated with the δ¹⁸O values of drinking water (Fig. 1). The estimated percentage contributions (±s.e.m.) of drinking water to the δ¹⁸O values of blood cells and feathers were 15±1% and 27±1%, respectively (Fig. 1).

Feathers were significantly enriched in deuterium compared with red blood cells and plasma (δD_{feathers–rbc}=21.0±1.5‰ and δD_{feathers–plasma}=18.6±4.4‰; paired t₁₂=13.70 and paired t₁₂=4.20, P<0.05, respectively), but the δD value of plasma and red blood cells did not differ significantly (paired t₁₂=0.72, P=0.75; Fig. 1). By contrast, feathers were significantly depleted in ¹⁸O relative to blood cells and plasma (δ¹⁸O_{feathers–rbc}=–1.4±0.3‰ and δ¹⁸O_{feathers–plasma}=–3.5±0.7‰; paired t₁₂=4.71 and paired t₁₁=5.29, P<0.01, respectively; Fig. 1). Additionally, plasma was significantly enriched in ¹⁸O relative to blood cells (δ¹⁸O_plasma–rbc=2.0±0.6‰; paired t₁₁=3.33, P<0.01). δD and δ¹⁸O were significantly correlated in feathers and red blood cells (r²=0.46 and 0.65, P<0.05, respectively; Fig. 2). However, there was no correlation between δD and δ¹⁸O in plasma (r²=0.00, P=0.95; Fig. 2).

The δD of house sparrow plasma proteins, red blood cells and feathers, and the δ¹⁸O of house sparrow red blood cells and feathers varied linearly with drinking water δD and δ¹⁸O, despite the provision of identical diets. Following the model of Ehleringer et al. we used the slope of these relationships as a ‘first-order estimate’ of the fractional contribution of drinking water to the hydrogen and oxygen content of tissues (Ehleringer et al., 2008). In the first section of this discussion, we use Ehleringer et al.'s model to examine the factors that might influence the value of these estimates. Our estimates were lower than those for similar tissues from previous studies (Hobson et al., 1999; O'Brien and Wooller, 2007; Ehleringer et al., 2008; Solomon et al., 2009). In the second section, we consider potential explanations for this observation. Our experiment also revealed a positive correlation between δD and δ¹⁸O in feathers and red blood cells, but not in plasma. In the third section, we examine the possibility of using the δD/δ¹⁸O relationship in keratin-containing tissues, such as hair and feathers, as a potential source of ecological information and discuss the implications of our results for the interpretation of animal movements that rely on δD and δ¹⁸O as markers for site of origin. Finally, in the fourth section, we consider possible reasons for the variations in isotopic compositions we found between tissue types.

Can the slopes of tissues versus drinking water regression lines be used to estimate the contribution of drinking water to the hydrogen and oxygen content of tissues? To answer this question, we consider the models proposed by Ehleringer et al. (Ehleringer et al., 2008). For a detailed description of these models, we refer readers to Ehleringer et al. (Ehleringer et al., 2008) and to Bowen et al. (Bowen et al., 2009). Briefly, the models account for all possible sources of hydrogen and oxygen in hair keratin, and identify possible steps in the incorporation of these elements that can lead to fractionation of an isotope. The models predict a linear relationship between the δD and δ¹⁸O values of drinking water and those of protein. Our results verified these predictions (Fig. 1). Furthermore, these models allow for the derivation of explicit expressions to describe the factors that influence the slope of the relationship between the D and ¹⁸O values of drinking water and tissues.

Protein is digested into its constituent amino acids in the small intestine, and most of the water in the lumen of the small intestine is derived from body fluids (Soergel and Hofmann, 1972). Consequently, the value of g₁ should be close to 1, and the value of the slope should be ≈a (namely, the proportion of body water oxygen derived from drinking water). Ehleringer et al.'s model assumes that the contribution of drinking water and food to the oxygen content of tissues is exclusively through their effect on the isotopic composition of body water (Ehleringer et al., 2008). In the following paragraphs, we use these interpretations to examine why the D and ¹⁸O isotopic composition of drinking water has a seemingly lower effect on the isotopic composition of sparrow tissues than that reported before for other species.

The percentage contribution of drinking water to the hydrogen content of house sparrow tissues was lower than previous estimates for similar tissues in other bird species [20–32% (Hobson et al., 1999)] and humans [36% (O'Brien and Wooler, 2007) and 27% (Ehleringer et al., 2008)], but higher than those estimated by Solomon et al. for salmonid fishes (Solomon et al., 2009). Granivorous passerines, such as house sparrows, have relatively low drinking water requirements (MacMillen, 1990). Therefore parameters d in Eqn 2 and a in Eqn 4 are likely to have a low values in our study system. Thus, we should expect lower contributions of drinking water in these animals than in obligate drinkers or aquatic organisms. The low value estimated by Solomon et al. for salmonids is perplexing (Solomon et al., 2009). We also expect the contribution of drinking water to the oxygen content of tissues to be higher in aquatic animals than in terrestrial animals with low drinking water requirements. We emphasize that tests of this prediction must account for potential differences in tissue types because the sources of oxygen and hydrogen in the different biomolecules that constitute different tissues might differ (Nielson and Bowen, 2010).

Eqn 5 establishes that the slope of the relationship between the δD and δ¹⁸O of tissues equals the product of the slope of the relationship for drinking water and the ratio of the fractional contributions of drinking water to the hydrogen [ed(1–fd)] and oxygen (a) content of tissues. For example, the slope of the δD and δ¹⁸O relationship for tap water in the United States is 7.9 and the ratio of the fractional contributions of drinking water to the hydrogen and oxygen content of hair equals 0.27/0.35=0.77 (Ehleringer et al., 2008). Thus, the expected slope of the δD and δ¹⁸O relationship for hair is approximately 6.1, which is close to the 5.7 value reported by Ehleringer et al. (Ehleringer et al., 2008). In our experiment, the δD/δ¹⁸O relationship for drinking waters had a slope of 15.0. For feathers and blood cells, the observed slopes of the δD and δ¹⁸O relationships were 6.4±2.2 and 7.2±1.6 (±s.e.m.), respectively. Although these values differed somewhat from the expected slope values (9.9 and 14.0) for feathers and blood cells, respectively, they were within the 95% confidence intervals of these expected values (5.6–19.8 for feathers and 7.1–80.9 for blood cells).

Feathers are inert after they finish growing, and, as a result, their isotopic composition reflects that of the drinking water and food ingested before the molt. By contrast, blood cells and plasma are constantly renewed (Wolf et al., 2009). In house sparrows, the average residence time of carbon in plasma solids is approximately 10 days, whereas that of blood cells is approximately 29 days (Carleton et al., 2008). The turnover of D and ¹⁸O in these tissues is unknown. By measuring the isotopic composition of different tissues, we can examine the resources (and potentially the habitats) used by animals at different times (see Martínez del Rio et al., 2009a). This application of stable isotopes demands that we know whether tissues differ in isotopic composition in a single animal.

In our experiment, feathers were significantly enriched in D relative to red blood cells and plasma. Feathers were also significantly depleted in ¹⁸O relative to blood cells and plasma. A D enrichment of approximately 20.0‰ or a depletion of approximately –1.4‰ in ¹⁸O in feathers relative to blood cells does not indicate a change in habitat or diet (see Results). We emphasize that the magnitude of the inter-tissue discrimination factors for D and ¹⁸O reported here must be considered preliminary as they are based on a single study on one species. However, these values suggest that there are significant differences in D and ¹⁸O composition in different tissues, and that these differences are large enough to complicate comparisons between the δD and δ¹⁸O values of different tissues. The δD of different amino acids in avian tissues can vary widely (M. L. Fogel and S. D. Newsome, unpublished data). Hence the differences in the δD of different tissues probably results from differences in the amino acid composition of the tissues (Martínez del Rio et al., 2009b).

Our results point to three areas that field researchers must consider when interpreting field data. These areas also represent potentially fruitful arenas of experimental research. First, drinking water has a significant effect on both the δD and δ¹⁸O values in animal tissues. This effect seems to depend on the contribution of drinking water to the animal's body water and hence probably differs between aquatic and terrestrial animals. In terrestrial animals it probably depends on water intake. Available sources of drinking water sometimes differ in δD and δ¹⁸O values from food; for example, when animals feed on crops irrigated with ground water or snow melt or feed on marine sources but drink freshwater (Lott et al., 2003). In these cases, it is unsafe to assume that there is a direct relationship between the isotopic composition of precipitation and that of animal tissues. Differences in the δD and δ¹⁸O values of food and water can confound the assignment of animals to a site of origin. In addition, the δD and δ¹⁸O values of different tissues might differ in animals fed on identical diets and drinking the same water. Newsome et al. have argued that comparing tissues with different time courses of isotope incorporation can yield insights on both the degree and timing of resource specialization of individual animals (Newsome et al., 2007). Comparisons of tissues must be informed by differences in the time course of isotopic incorporation (see Wolf et al., 2009) and account for differences in discrimination factors in the different tissues. We know little about the time course of incorporation of hydrogen and oxygen into animal tissues (see Podlesak et al., 2008) and our results suggest that different tissues will have different D and ¹⁸O discrimination factors. Interpreting differences in δD and δ¹⁸O in the tissues of a single animal requires more experimental data on the magnitude of inter-tissue discrimination factors. We speculate that these differences are the result of variation in δD and δ¹⁸O values in different amino acids and in the amino acid compositions of tissues (Martínez del Rio et al., 2009b). Future explorations of inter-tissue discrimination factors may benefit from the insights provided by compound specific analysis (Howland et al., 2003). Finally, there seems to be a clear relationship between the δD and δ¹⁸O values of different body tissues (Ehleringer et al., 2008; Bowen et al., 2009). The δD/δ¹⁸O meteoric water line has proved invaluable as a reference in hydrology and geochemistry (Clark and Fritz, 1997). We hypothesize that similar lines will be found for animals and will prove to be important tools in animal isotopic ecology.

We thank Scott Carleton, Seth Newsome and Jacob Peters for helpful comments on this manuscript.

 
• a

  proportion of body water oxygen derived from drinking water (^*)

 
• ANOVA

  analysis of variance

 
• d

  proportion of body water hydrogen derived from drinking water (^*)

 
• D

  hydrogen isotope (1 proton, 1 neutron; moles)

 
• D₂O

  99 atom percent deuterated water

 
• e

  proportion of follicle water hydrogen derived from body water (^*)

 
• f_d

  fraction of hydrogen atoms derived from diet (^*)

 
• g₁

  proportion of gut water derived from body water (^*)

 
• g₂

  proportion of gut water derived from drinking water (^*)

 
• h

  proportion of oxygen in body water that is lost as fractionated water associated with breathing and/or evaporation (^*)

 
• H

  hydrogen isotope (1 proton, 0 neutrons; moles)

 
• H₂O¹⁸

  98 atom percent oxygen ¹⁸O water

 
• IAEA-GNIP

  International Atomic Energy Agency – Global Network of Isotopes in Precipitation

 
• j

  proportion of oxygen in body water lost as CO₂ (^*)

 
• k

  proportion of oxygen in body water that is lost as unfractionated water (^*)

 
• m

  proportion of hydrogen in body water that is lost as fractionated water associated with breathing and/or evaporation (^*)

 
• n

  proportion of hydrogen in body water that is lost as unfractionated water (^*)

 
• SMOW

  standard mean ocean water

 
• α_CO2

  fractionation of oxygen between body water and CO₂ (^*)

 
• α_fwlh

  fractionation of hydrogen in water lost through breathing and/or evaporation (^*)

 
• α_fwlo

  fractionation of oxygen in body water lost through breathing and/or evaporation (^*)

 
• α_hw

  hydrogen isotope fractionated during protein synthesis (^*)

 
• α_ow

  fractionation of oxygen between carbonyl oxygen and water (^*)

 
• δD

  [(²H/¹H_sample–²H/¹H_standard)²H/¹H_standard–1]×10³ (‰, SMOW)

 
• δ¹⁸O

  [(¹⁸O/¹⁶O_sample–¹⁸O/¹⁶O_standard)¹⁸O/¹⁶O_standard–1]×10³ (‰, SMOW)

 
• Δ¹⁸O_{feather–precipitation}

  δ¹⁸O_feather–δ¹⁸O_{precipitation} (‰, SMOW)

 
• ΔD_{feather–precipitation}

  δD_feather–δD_{precipitation} (‰, SMOW)

 
• ¹⁶O

  oxygen isotope (8 protons, 8 neutrons; moles)

 
• ¹⁸O

  oxygen isotope (8 protons, 10 neutrons; moles)