Hydrogen and Oxygen Isotope Ratios in Body Water and Hair: Modeling Isotope Dynamics in Nonhuman Primates

Abstract

The stable isotopic composition of drinking water, diet, and atmospheric oxygen influence the isotopic composition of body water (²H/¹H, ¹⁸O/¹⁶O expressed as δ²H and δ¹⁸O). In turn, body water influences the isotopic composition of organic matter in tissues, such as hair and teeth, which are often used to reconstruct historical dietary and movement patterns of animals and humans. Here, we used a nonhuman primate system (Macaca fascicularis) to test the robustness of two different mechanistic stable isotope models: a model to predict the δ²H and δ¹⁸O values of body water and a second model to predict the δ²H and δ¹⁸O values of hair. In contrast to previous human-based studies, use of nonhuman primates fed controlled diets allowed us to further constrain model parameter values and evaluate model predictions. Both models reliably predicted the δ²H and δ¹⁸O values of body water and of hair. Moreover, the isotope data allowed us to better quantify values for two critical variables in the models: the δ²H and δ¹⁸O values of gut water and the ¹⁸O isotope fractionation associated with a carbonyl oxygen-water interaction in the gut (α_ow). Our modeling efforts indicated that better predictions for body water and hair isotope values were achieved by making the isotopic composition of gut water approached that of body water. Additionally, the value of α_ow was 1.0164, in close agreement with the only other previously measured observation (microbial spore cell walls), suggesting robustness of this fractionation factor across different biological systems.

Introduction

The hydrogen and oxygen isotope ratio (²H/¹H, ¹⁸O/¹⁶O expressed as δ²H and δ¹⁸O) analyses of hair and other keratin-based tissues have been used in studies of movement and resource use in the fields of anthropology [Bowen et al., 2009; Ehleringer et al., 2008; Nardoto et al., 2006; O’Connell & Hedges, 1999; Sharp et al., 2003; Thompson et al., 2010], ecology [Brattström et al., 2008; Cerling et al., 2006; Chamberlain et al., 1997; Hobson & Wassenaar, 1997; Hobson et al., 2004; Ramos et al., 2009; Rubenstein & Hobson, 2004; Wassenaar & Hobson, 1998], and more recently in human forensic investigations [Bol et al., 2007; Fraser et al., 2006; Meier-Augenstein & Fraser, 2008; Mützel Rauch et al., 2009]. The link between isotopic composition and geographic location exists because the hydrogen and oxygen isotope composition of environmental water varies predictably across the landscape [Bowen et al., 2007; Craig, 1961; Dansgaard, 1954; Kendall & Coplen, 2001]. As water is first incorporated into plants (as water or in organic matter) and later into primary and secondary consumers in the ecosystem, the isotopic composition of water is integrated into the food web, resulting in an isotopic “signature” of geography [Bowen, 2010; Bowen et al., 2005a; West et al., 2006, 2010]. Animal tissues commonly used in geo-location studies, such as hair, teeth, feathers, and nails, incorporate the isotopic signature of the water and food consumed at the time of synthesis – thus these tissues provide geographical information about location during these intense periods of growth.

Most ecological studies of primates have focused on the use of carbon and nitrogen isotopes to learn about diet and infer habitat preferences, while oxygen (O) isotopes have only recently been used to address ecological questions [Bedaso et al., 2010; Cerling et al., 2003, 2004, 2011; Chenery et al., 2011; Fourie et al., 2008; Levin et al., 2006; Quade et al., 1995; Smith et al., 2010; Sponheimer & Lee-Thorp, 2001; White et al., 2009; Yeakel et al., 2007]. Furthermore, the use of stable isotopes of hydrogen (H), at natural abundances, in primate research is mostly limited to geo-location of modern humans [Bowen et al., 2009; Ehleringer et al., 2008; Fraser et al., 2006; Meier-Augenstein & Fraser, 2008; Mützel Rauch et al., 2009; Sharp et al., 2003; Thompson et al., 2010]. The analysis of H and O isotope ratios has the potential to clarify several key ecological questions important for primatologists such as niche partitioning and dietary specialization, dispersal, residence patterns and social organization. Given the many environmental, nutritional and metabolic variables that influence the isotopic composition of animal tissues, it is important to develop a thorough understanding of hydrogen and oxygen isotope dynamics.

An offset between the isotopic composition of source water and tissue occurs due to inputs of H and O atoms from other sources (e.g., atmospheric oxygen, diet) as well as because of the impact of physiological processes associated with water metabolism (e.g., water flux, metabolic rate). Predictive models that incorporate fluxes and fractionation events have been created for the H and O isotopic composition of phosphate in bones and teeth [Ayliffe & Chivas, 1990; Bryant & Froelich, 1995; Daux et al., 2008; Luz & Kolodny, 1985; Podlesak et al., 2008] and more recently, keratin in hair [Bowen et al., 2009; Ehleringer et al., 2008; Podlesak et al., 2008]. While these are endpoint models for the tissue of interest (e.g., hair keratin), calculations necessarily incorporate the body water pool (i.e., free water in the body) [Gretebeck et al., 1997] and thus produce an intermediate prediction for the isotopic composition of body water.

The complexity of models describing the H and O stable isotopes in body water and proteinaceous tissues [Bryant & Froelich, 1995; Gretebeck et al., 1997; Kohn, 1996] can vary greatly. While some models only require the isotopic values of diet, water, and atmospheric O₂ to generate predictions [Ehleringer et al., 2008], others incorporate physiological processes relative to the organism of interest (i.e., mass, metabolic rate, water economy index), as well as climactic and dietary variables. Few modeling studies [Podlesak et al., 2008] have collected both body water and tissue samples from individual animals in a controlled setting to test the accuracy of the modeled values for not only the endpoint tissue of interest, but also the intermediate body water prediction. Here we used a nonhuman primate system to test the robustness of mechanistic models of both the δ²H and δ¹⁸O values of body water [Bryant & Froelich, 1995; Kohn, 1996] and hair [Bowen et al., 2009; Ehleringer et al., 2008; Podlesak et al., 2008]. The goal of this research was to first, test the accuracy of both modeling approaches using a controlled system and second, use data from this system to improve estimates of model parameters.

Methods

Ethical use of Animals

All procedures involving animals were conducted in compliance with state and federal laws of the US Department of Health and Human Services and guidelines established by the Wake Forest University. All animal procedures were approved by the Wake Forest Institutional Animal Care and Use Committee (IACUC). The research and procedures adhered to the American Society of Primatologists Principles for the Ethical Treatment of Nonhuman Primates.

Animal System

Samples of serum and hair were obtained from nonhuman primates (cynomolgus monkey, Macaca fascicularis) housed at the Wake Forest Primate Center (Wake Forest, NC, USA). Serum and hair samples were obtained from 10 adult females and 10 adult males. As part of an on-going study at the Wake Forest Primate Center, animals were divided into two treatment cohorts (5 females and 5 males each) and fed a Typical American Diet (TAD) with 35% of calories from fat, a moderate amount of dietary cholesterol, and less carbohydrate than standard monkey chow. One cohort consumed a diet with the protein derived from soy (TAD soy; LabDiet 5L0R) and the other group consumed a diet with the protein derived from casein (TAD casein; LabDiet 5L0P). Both TAD diets satisfy National Research Council recommendations for nonhuman primates and were custom formulated by Purina LabDiet®. Dietary macronutrient content of the diets has been published previously [Wagner et al., 2009].

Isotope analysis

We determined the stable hydrogen (δ²H) and oxygen (δ¹⁸O) isotope ratios of water cryogenically extracted from serum (hereafter referred to as body water) and hair of TAD soy (n=10) and TAD casein (n=10) monkeys. We also determined the δ²H and δ¹⁸O values of drinking water (n = 12) and of the casein (n = 2) and soy (n = 2) diets. Following extraction, aliquots of 0.4 mL of body water were transferred to 1.8-mL crimp-top gas chromatography (GC) vials and sealed until analysis. Body water and drinking water samples were prepared and analyzed alongside a set of three laboratory water reference materials previously calibrated to the Vienna Standard Mean Ocean Water (VSMOW) scale. The hydrogen and oxygen isotope ratios of two primary laboratory reference waters (DI and ZE), were −123 and −16.5 ‰ for DI and −0.1 and −0.2 ‰ for ZE, when calibrated against international standards (GISP, SLAP and VSMOW). As a Quality Control, a third laboratory water reference was used (values −72.5 ‰ for δ²H and −10.2 for δ¹⁸O). Both hair samples and food samples were ground into a homogenous powder. There is a partial isotopic exchange of H atoms in keratin with environmental water (e.g., the atmosphere); thus, all samples were analyzed together with hair reference materials for which the δ²H of nonexchangeable H had been determined using the principle of identical treatment [Bowen et al., 2005b; Chesson et al., 2009]. Both reference material and samples were equilibrated with water vapor in the laboratory atmosphere and then desiccated under vacuum for 5 days. Samples of 150 μg of ground hair and food were weighed and analyzed.

The δ²H and δ¹⁸O values of body water, drinking water, hair samples and food samples were analyzed at the University of Utah Stable Isotope Ratio Facility for Environmental Research (SIRFER; http://sirfer.utah.edu) on a ThermoFinnigan-MAT Delta Plus XL isotope ratio mass spectrometer (Bremen, Germany) with an attached high temperature conversion elemental analyzer (TC/EA). Samples were pyrolyzed at 1400°C to produce H₂ and CO. Resultant gases were separated on a 1-m, 0.25 in (outer diameter) molecular sieve 5Å gas chromatography column (Costech Analytical, Valencia, CA, USA). Water samples were introduced to the pyrolysis column using a PAL autosampler (LEAP Technologies, Carrboro, NC, USA). For hair and food, samples were stored on a zero-blank autosampler (Costech Analytical, Valencia, CA, USA), to ensure that water vapor did not exchange with the samples during analysis. Raw δ²H and δ¹⁸O measurement values for unknown samples were corrected to the Vienna Standard Mean Ocean Water (VSMOW) scale using a two-point correction with laboratory reference materials. The hydrogen and oxygen isotope ratios of our laboratory reference materials, ground horse hair (FH and UH), were −76.3 and 1.49 ‰ for FH and −141.6 and 5.7 ‰ for UH when calibrated in 2005 [Bowen et al., 2005b]. The analytical process included analyses of QC keratin specimens of known isotopic composition, analyzed also as unknown (its isotope values were −114 ‰ for δ²H and 12.8 ‰ for δ¹⁸O). All QC analyses met the required specifications. The overall analytical precision for water analysis was ±1.55 ‰ and ±0.17 ‰ for H and O, respectively. Analytical precision for hair analysis, calculated from the measured hydrogen and oxygen isotope ratios of powdered keratin used as an internal secondary reference material was ±1.1 ‰ for H and ±0.23 ‰ for O (n = 18).

Stable Isotope Notation

Stable isotope abundances are reported in δ-notation as parts per thousands (‰), where

$${\delta }_{\text{sample}}=({\text{R}}_{\text{sample}}/{\text{R}}_{\text{standard}}-1)\times 1000$$ Eqn 1

and R_sample and R_standard are the molar ratios of the rare to abundant isotope (e.g., ²H/¹H) in the sample and standard, respectively [Coplen, 2011]. The international standard for both hydrogen and oxygen stable isotope analysis is VSMOW.

Modeling

Body water modeling

We applied a mechanistic model describing the isotopic composition of body water to determine if we could accurately predict the H and O isotope ratios of body water of M. fascicularis (hereafter referred to as the Primate Body Water Model). Similar models have been constructed and then applied to reconstruct body water in both modern and paleoclimate conditions [Bryant & Froelich, 1995; Kohn, 1996; Luz & Kolodny, 1985] and to determine energy expenditure via the doubly labeled water technique [Gretebeck et al., 1997; Schoeller et al., 1986]. We constructed a body water model by modifying the parameters of Bryant & Froelich [1995] and Kohn [1996] to most closely approximate the physiological traits of nonhuman primates.

In brief, given the assumption that an animal is in isotopic equilibrium, the general mass balance equation of influxes and effluxes can be written as

$$R_{dw}^O{F}_{dw}^O+R_{fw}^O{F}_{fw}^O+R_{\mathit{food}}^O{F}_{\mathit{food}}^O+R_{\mathit{wvg}}^O{F}_{\mathit{wvg}}^O+R_{\mathit{air}}^O{F}_{\mathit{air}}^O={\alpha }_{co2}^O{R}_{bw}^O{F}_{co2}^O+{\alpha }_{\mathit{bwv}}^O{R}_{bw}^O{F}_{\mathit{bwv}}^O+{\alpha }_{\mathit{twv}}^O{R}_{bw}^O{F}_{\mathit{twv}}^O+R_{bw}^O{F}_{rw}^O$$ Eqn 2

for ¹⁸O, and similarly

$$R_{dw}^H{F}_{dw}^H+R_{fw}^H{F}_{fw}^H+R_{\mathit{food}}^H{F}_{\mathit{food}}^H+R_{\mathit{wvg}}^H{F}_{\mathit{wvg}}^H={\alpha }_{\mathit{bwv}}^H{R}_{bw}^H{F}_{\mathit{bwv}}^H+{\alpha }_{\mathit{twv}}^H{R}_{bw}^H{F}_{\mathit{twv}}^H+R_{bw}^H{F}_{rw}^H$$ Eqn 3

for ²H. Isotope abundances are expressed as ratios (R) of heavy to light isotopes, fluxes (F) are in units mole/day, and isotopic fractionation relative to body water is expressed as α. Oxygen and hydrogen influxes and effluxes considered include drinking water (dw), free water in food (fw), oxygen and hydrogen bound in food (food), water vapor gain (wvg), atmospheric oxygen (air), carbon dioxide (CO₂), breath water vapor (bwv), transcutaneous water vapor (twv), and unfractionated remaining water loss (rw) which includes urine, fecal water, and sweat. Solving for the isotope ratio of body water yields

$$R_{bw}^O=\frac{R_{dw}^O{F}_{dw}^O+R_{fw}^O{F}_{fw}^O+R_{\mathit{food}}^O{F}_{\mathit{food}}^O+R_{\mathit{wvg}}^O{F}_{\mathit{wvg}}^O+R_{\mathit{air}}^O{F}_{\mathit{air}}^O}{{\alpha }_{co2}^O{F}_{co2}^O+{\alpha }_{\mathit{bwv}}^O{F}_{\mathit{bwv}}^O+{\alpha }_{\mathit{twv}}^O{F}_{\mathit{twv}}^O+F_{rw}^O}$$ Eqn 4

for oxygen, and

$$R_{bw}^H=\frac{R_{dw}^H{F}_{dw}^H+R_{fw}^H{F}_{fw}^H+R_{\mathit{food}}^H{F}_{\mathit{food}}^H+R_{\mathit{wvg}}^H{F}_{\mathit{wvg}}^H}{{\alpha }_{\mathit{bwv}}^H{F}_{\mathit{bwv}}^H+{\alpha }_{\mathit{twv}}^H{F}_{\mathit{twv}}^H+F_{rw}^H}$$ Eqn 5

for hydrogen.

Influxes and effluxes of oxygen and hydrogen fall into two categories, those associated with free water (dw, fw, wvg, bwv, twv, rw) and those derived from the metabolism of carbohydrates, fats, and protein (food, air, co2). Oxidation reactions for glucose, palmitic acid, and alanine are considered first order representatives of carbohydrate, fat, and protein substrates [Bryant & Froelich, 1995]. The flux terms associated with metabolism are therefore functions of diet composition, specifically the fraction of energy derived from carbohydrates, fats, and protein, as well as total energy expenditure.

All parameters were either measured directly or estimated from directly measurable parameters (Table I). We used the equations described by Bryant & Froelich [1995] and Kohn [1996] to estimate the molar influxes and effluxes H₂O. Where it was possible, we used calculations specific to nonhuman primates to estimate physiological variables. For example, total energy expenditure (TEE) was estimated using the following equation, formulated using the TEE for 19 species of primates [Key & Ross, 1999; Pontzer & Wrangham, 2004]:

$$\text{TEE}=93.3{\text{M}}^{0.75}$$ Eqn 6

where TEE is kCal/day and M is body mass in Kg. Five of the females considered in this study were nursing during sample collection and thus TEE during lactation was calculated as 1.39×TEE [Key & Ross, 1999]. Animals were housed in an indoor-outdoor facility with temperature-controlled indoor housing areas. Facility-recorded averages were used as a best guess estimate for temperature (21°C) and humidity (40%) inputs in the model. These estimates for temperature and humidity are not perfect, as animals are given access to a large outdoor enclosure when weather is good and outdoor measures of these variables change seasonally.

Table I.

Values used in body water modeling. Flux value ranges represent those calculated for the actual mass range in the sample set (2 kg – 13 kg).

Influx (moles/day)	Oxygen Flux	Hydrogen Flux	Source
dw – drinking water	8.3–36.7	16.6–73.3	[Bryant & Froelich, 1995]
fw – food water	0	0	Measured
food – bound O and H	0.4–1.6	1.2–5.0	[Bryant & Froelich, 1995]
wvg – water vapor gain	0.2–0.8	0.4–1.5	[Bryant & Froelich, 1995]
air – atmospheric O	1.4–5.7	n/a	[Bryant & Froelich, 1995]
Efflux (moles/day)	Oxygen Flux	Hydrogen Flux	Source
CO2 – carbon dioxide	1.2–4.9	n/a	[Bryant & Froelich, 1995]
bwv – breath water vapor	1.3–5.1	2.5–10.2	[Kohn, 1996]
Twv – transcutaneous water vapor	0.8–2.9	1.6–5.8	[Schoeller et al., 1986]
rw – remaining water	7.0–32.0	14.0–64.0	calculated
Fractionation Factors	Oxygen	Hydrogen	Source
aco2	1.038	n/a	[Chamberlain et al., 1997; Pflug et al., 1979]
abwv	0.991	0.946	[Hobson & Wassenaar, 1997; Horita & Wesolowski, 1994]
atwv	0.981	0.935	[Hobson & Wassenaar, 1997; Horita & Wesolowski, 1994]
Isotope Ratio (‰)	Oxygen casein, soy	Hydrogen casein, soy	Source
dw – drinking water	−5.8	−32.9	measured
fw – food water	n/a	n/a	n/a
food – bound O and H	27.4, 26.8	−116.1, −119.8	measured
wvg – water vapor gain	−15.3	−80.5	[Kohn, 1996]
air – atmospheric O	15.1	n/a	[Kohn, 1996]

Hair modeling

The semi-mechanistic model for the hydrogen (H) and oxygen (O) isotopic composition of human hair was developed by Ehleringer et al. [2008] and later modified slightly by Bowen et al. [2009] to incorporate additional dietary considerations (hereafter referred to as “Hair Model”). These models estimate δ²H and δ¹⁸O of H and O atoms fixed in vivo by calculating an isotopic mass balance of H and O influxes (i.e., a, b and c in Fig. 1) and effluxes (i.e., h, j and k in Fig. 1). The models assume that the non-exchangeable H atoms in hair amino acids are a mixture of H atoms from dietary amino acids and amino acids synthesized internally which would have H isotopes reflecting body water. Oxygen atoms in hair are assumed to be the result of isotopic exchange with gut water during hydrolysis of dietary protein.

Figure 1.

Schematic of the flow of oxygen atoms into hair keratin as described by the Ehleringer et al. [2008] hair model. Model parameters are: a = Proportion of body water O derived from drinking water; b = Proportion of body water O derived from diet; c = Proportion of body water O derived from O₂; g₁ = Proportion of gut water derived from body water; α_ow = Fractionation of O between carbonyl oxygen and water; h = Proportion of O in body water that is lost as fractionated water associated with breathing and evaporation; j =Proportion of O in body water lost as CO₂; k = Proportion of O in body water that is lost as unfractionated water.

The controlled diet and environmental study system allowed us to constrain the ranges of possible values for two critical variables presented in the hair model for the estimation of the oxygen isotope ratios of hair (δ¹⁸O_h): the fractional contribution of body water to gut water (g₁) and the oxygen isotope fractionation associated with a carbonyl oxygen-water exchange in the gut (α_ow) (Fig. 1). Such calculations are possible for the flow of O atoms into hair keratin because of the (assumed) primary contribution of body water to the system. Similar calculations for hydrogen are currently not possible because the primary contribution of H atoms to keratin come from both essential (diet-derived) and non-essential (synthesized in vivo) amino acids. The hair model assumes that oxygen atoms flow from the body water pool, to gut water (where isotopic exchange can occur), and into hair follicles where amino acids are condensed to produce hair keratin (Fig. 1). Body water is calculated as an essential intermediate in the hair model and is estimated based on data collected for modern humans [Gretebeck et al., 1997]. The model calculates the isotopic composition of gut water as a mixture of body water and water in food. The estimated contribution of body water to gut water (g₁) is poorly constrained by existing data, but a previous study [Malagelada et al., 1976] suggested a value of g₁=0.57. Using that value in the hair model developed by Ehleringer et al. [2008] provided reasonable results. Using the oxygen isotopic composition of body water and hair of nonhuman primates, we explore how variable settings of g₁ affect estimated values relative to measured hair stable isotope values.

If amino acids remain in a neutral pH environment following protein hydration in the gut, the O atoms in the carbonyl O should reflect the isotopic composition of gut water during digestion. This relationship is illustrated in the following equation from Ehleringer et al. [2008]

$${\delta }^{18}{\text{O}}_{\text{h}}={{\alpha }_{\text{ow}}}^{\ast }(1000+{\delta }^{18}{\text{O}}_{\text{wg}})-1000$$ Eqn 7

where α_ow is the oxygen isotope fractionation associated with a carbonyl oxygen-water interaction in the gut. Based on experimental observations, [Kreuzer-Martin et al., 2003] calculated αow as 1.0164 for microbial spore cell walls during non-log phase growth.

Assuming a rapid water exchange across intestinal walls and making g₁ equal to one, (therefore the isotopic composition of body water (δ¹⁸O_wb) is equal to gut water (δ¹⁸O_wg)), we are able to use the data collected for nonhuman primates to calculate the value of α_ow in this system.

$${\alpha }_{\text{ow}}=({\delta }^{18}{\text{O}}_{\text{h}}+1000)/({\delta }^{18}{\text{O}}_{\text{wg}(\text{wb})}+1000)$$ Eqn 8

Statistical analysis

Data are presented as means ± 1 standard deviation unless stated otherwise. Welch two sample t-test was used to compare stable isotope values of samples from groups [Welch, 1947]. Data analysis and statistical calculations were conducted using R (include reference). Statistical significance is reported as exact P values; significance lever was set at α = 0.05.

Results

Measured Parameter Values

The δ²H and δ¹⁸O values of treatment diets were similar. The δ²H values of the casein and soy diets were −116.1 ± 0.7 ‰ and −119.8 ± 0.5 ‰, respectively, and the δ¹⁸O values of the casein and soy diets were 27.4 ± 0.2 ‰ and 26.8 ± 0.7 ‰, respectively. Animals in both dietary groups had ad lib access to the same water source (δ²H value of −32.9 ± 2.1 ‰ and δ¹⁸O value of −5.8 ± 0.2 ‰).

The δ²H values of the body water of primates fed TAD-casein and TAD-soy were −28.7 ± 3.0 ‰ and 30.8 ± 2.6 ‰, respectively. These δ²H values were not statistically different (t = 1.7, P = 0.11, df = 17.7). The δ¹⁸Ovalues of the body water of primates fed TAD-casein and TAD-soy were −3.9 ± 0.5 ‰ and −4.2 ± 0.7 ‰, respectively. These δ¹⁸O values were not statistically different from each other (t = 1.02, P = 0.32, df = 17.0).

As would be predicted by the hair model, given similar body water values of animals in the two dietary groups, there was no statistically significant difference in the isotopic composition of hair between the two dietary groups. The δ²H values of the hair of primates on the casein diet were −90.4 ± 2.8 ‰ and −89.5 ± 1.6 ‰ for those on the soy diet (t = −0.83, P = 0.42, df = 15.0). The δ¹⁸O values of hair of primates fed TAD-casein and TAD-soy were 12.1 ± 0.6 ‰ and 12.4 ± 0.5 ‰, respectively (t = −1.40, P = 0.18, df = 16.4).

Model predictions

Body water and hair model predictions

The δ²H and δ¹⁸Ovalues of body water, predicted by the Primate Body Water Model, for animals on the casein and soy diet were not statistically different from the measured values (Fig. 2; Table II; TAD-casein measured δ²H vs. predicted δ²H values t = −1.1, P = 0.30, df = 9.3; TAD-casein measured δ¹⁸O vs. predicted δ¹⁸O values t = −0.41, P = 0.70, df = 9.71; TAD-soy measured δ²H vs. predicted δ²H values t = −3.76, P = 0.05, df = 9.2); TAD-soy measured δ¹⁸O vs. predicted δ¹⁸O values t = −1.70, P = 0.12, df = 9.3). The body water values predicted as an intermediate step in the Hair Model were contained within the 95% confidence interval (CI) of the observed isotope distributions (Fig. 2); although, for the casein diet the predicted value was not contained within the 90% CI (Fig. 2A).

Figure 2.

Measured and predicted δ²H and δ¹⁸O values (‰) of body water from nonhuman primates fed two different diets, (A) TAD casein and (B) TAD soy. Open circles represent measured values. Ellipses are the normal-probability contours of the isotope distributions for the measured (dashed lines) and predicted (solid line) values by the Primate Body Water Model. Filled circle represents the body water value predicted as an intermediate step in the Hair Model.

Table II.

Measured and predicted δ²H and δ¹⁸O values (‰) of body water from nonhuman primates.

Diet	Measured Body Water		Predicted using Body Water Model		Predicted using Hair Model
	δ2H	δ18O	δ2H	δ18O	δ2H	δ18O
Casein	−28.7 ± 3.0	−3.9 ± 0.6	−27.6 ± 0.4	−3.8 ± 0.1	−34.9	−4.1
Soy	−30.8 ± 2.6	−4.2 ± 0.7	−27.6 ± 0.3	−3.8 ± 0.1	−35.6	−4.2

The fractional contribution of body water to the gut water pool is represented in the hair model by the term g₁. Ehleringer et al. [2008] set the value for g₁ to 0.57 in agreement with a previously published observation [Malagelada et al., 1976]. Here we tested Hair Model predictions for g₁ = 0.57 and g₁ = 1.0, which seemed more biologically reasonable given the rapid exchange across intestinal walls. Finally, leaving g₁ = 1.0, we combined the body water and hair models, using body water model estimates to predict the isotopic composition of hair. The δ²H and δ¹⁸O values predicted by the Hair Model using g₁ = 0.57 were not contained within the 95% CI of the observed isotope distributions (Fig. 3; Table III). When g₁ was set to 1.0, the predicted values for both diets were contained within the 90% CI of the distributions and were similar to the measured observed isotope values (Fig. 3; Table III). The δ²H and δ¹⁸O values of hair predicted using the combined Hair Model and Body Water Model for animals on the two different diets were not statistically different from the observations (Fig. 3; Table III; TAD-casein measured δ²H vs. predicted δ²H values t = −0.93, P = 0.37, df = 9.0; TAD-casein measured δ¹⁸O vs. predicted δ¹⁸O values t = −2.10, P = 0.06, df = 9.6; TAD-soy measured δ²H vs. predicted δ²H values t = 4.25, P = 0.048, df = 12.0); TAD-soy measured δ¹⁸O vs. predicted δ¹⁸O values t = −0.88, P = 0.40, df = 8.3).

Figure 3.

Measured and predicted δ²H and δ¹⁸O values (‰) of hair from nonhuman primates fed two different diets, (A) TAD casein and (B) TAD soy. Open circles represent measured values. Ellipses are the normal-probability contours of the isotope distributions for the measured (dashed lines) and predicted (solid line) values using the Body Water Model estimates as an intermediate step. The filled circle and triangle represent the hair values predicted by the Hair Model with g₁ = 0.57 and 1.0 respectively.

Table III.

Measured and predicted δ²H and δ¹⁸O values (‰) of hair from nonhuman primates.

Diet	Measured Hair		Predicted using Hair Model			Predicted using Body Water Model intermediate value
				g1 = 0.57	g1 = 1.0
	δ2H	δ18O	δ2H	δ18O	δ18O	δ2H	δ18O
Casein	−90.4±2.8	12.1±0.6	−88.9	14.2	12.2	−89.5 ± 0.1	12.5 ± 0.1
Soy	−89.5±1.6	12.4±0.5	−91.9	13.9	12.1	−92.4 ± 0.1	12.5 ± 0.1

Calculating the carbonyl oxygen-water fractionation factor in the gut (α_ow)

In setting g₁ = 1, we effectively are stating that the isotopic composition of body water (δ¹⁸O_wb) was equal to gut water (δ¹⁸O_wg). This assumption allowed us to calculate α_ow, or the ¹⁸O fractionation associated with a carbonyl oxygen-water interaction in the gut. We calculated α_ow to be 1.0164 ± 0.0005 in this nonhuman primate system, which is in close agreement with the calculation of Kreuzer-Martin et al. [2003] in microbial spore cell walls. This suggests robustness in this fractionation value across biological organisms.

Discussion

Using body water and hair samples from a nonhuman primate system, we evaluated the robustness of semi-mechanistic models of the δ²H and δ¹⁸Ovalues of body water [Bryant & Froelich, 1995; Kohn, 1996] and hair [Bowen et al., 2009; Ehleringer et al., 2008; Podlesak et al., 2008]. Both models performed well (Tables II and III; Fig. 2 and 3), suggesting that broader applications of the models to other primates and to reconstructing human dietary patterns and movements may be justified. A significant improvement in model predictions came as a result of describing gut water as isotopically similar to that of body water, which makes biological sense because of rapid water exchanges across the intestinal walls, and in this system because of the low water content of the food. There were no notable improvements to the δ²H and δ¹⁸O estimates of hair keratin when the outputs of the Primate Body Water Model were used in place of body water calculations within the Hair Model (when g₁ was equal to 1).

The incorporation of physiological (i.e., mass, metabolic rate, water intake) and dietary details (i.e., macronutrient density, food water content) into body water modeling allows for the impact of variables to be isolated. For example, O’Grady et al. [2010] demonstrated elevated water fluxes associated with diabetes could be detected through measurement of the isotopic composition of body water. In this study, the isotopic values of body water from streptozotocin (STZ)-induced diabetic mice approached that of drinking water, due to abnormally high rates of water flux. O’Grady et al. [2010] were able to predict the H and O isotope values of body water in control versus diabetic mice modifying only water flux rates in the body water model. Defining the influence of particular variables, such as water flux, on the isotopic composition of body water may improve explanations of within-population variance, where diet and water inputs are similar.

In addition to evaluating the robustness of the body water and hair models, the results of this study indicate that estimates of the δ¹⁸O values of hair were improved by increasing the fractional contribution of body water to gut water (g₁) from 0.57 to 1.0. This model parameter adjustment suggests that the gut water pool is similar in composition to body water in nonhuman primates. Given the physiological similarities between human- and nonhuman primates, it is probable that model predictions for human hair would be improved if g₁ were close to or equal to 1.0. By defining g₁ as 1.0, we were able to calculate the oxygen isotope fractionation associated with the carbonyl oxygen-water interaction in the gut, α_ow. Interestingly, our calculated value for this nonhuman primate system (1.0164) agreed well with the calculation of α_ow for microbial spore cell walls [Kreuzer-Martin et al., 2003].

We found that current models for both the isotopic composition of body water and hair are accurate for the population of nonhuman primates considered. Investigating within-population variance in the isotopic composition of primate body water and hair would require more detailed information about the individuals studied and their environment. While much of isotopic modeling has been done using unconstrained systems, controlled study systems can shed light on many parameter estimates that cannot be measured directly. Further constraining model variables will only improve model estimates and simultaneously increase knowledge of the metabolic processes involved in shaping the flow of hydrogen and oxygen atoms into biologic tissues.