Introduction

Stable isotope ratios in animal tissues vary with diet, habitat, and environmental conditions, and are often used to assess the foraging ecology and habitat preferences of living and extinct species (West et al. 2006). These studies have varied methodologically, using a range of tissues. For instance, the diets of wild primates have been assessed using isotope values from hair (e.g., Schoeninger et al. 1997, 2006), tooth enamel (e.g., Codron et al. 2005; Fourie et al. 2008; Smith et al. 2010), bone (e.g., Ambrose and DeNiro 1986; Thackeray et al. 1996; Smith et al. 2010), and feces (e.g., Codron et al. 2006). These data, in turn, have been used to inform paleo-ecological models of extinct species, including early human ancestors (e.g., Thackeray et al. 1996; Codron et al. 2005; Sponheimer et al. 2006, 2010). Due to tissue preservation issues, these studies have frequently had to use different tissues in their modern and ancient comparisons.

Although the availability and state of preservation of specimens are practical constraints, it can be problematic to compare the isotopic composition of different tissues for two reasons. First, carbon and nitrogen isotope values of proteinaceous tissues can differ within an animal because each tissue has a unique amino acid (AA) composition, and the AAs themselves vary isotopically (e.g., Hare et al. 1991; Styring et al. 2010). Second, when studying fossils, researchers generally use the carbonate fraction of biological apatite. The isotopic difference [hereafter termed apparent enrichment, ε*(defined below)] between carbon in organic tissues, such as collagen in bone or dentin, and carbon in the carbonate in bone or tooth apatite varies with both digestive physiology and dietary macromolecular composition (reviewed in Hedges 2003).

A wide range of dietary and gut physiological adaptations among primates could lead to differences in ε* values for both carbon and nitrogen that could in turn confound ecological or paleoecological interpretations. Many experiments have been conducted on rodents and pigs, but most were focused on carbon isotope differences between carbonate and collagen. Scarcely any work has examined the differences between proteinaceous tissues, let alone unconventional taxa. For instance, only one published study has focused on nonhuman primates (O’Regan et al. 2008) (Table 1). Accordingly, we compared the carbon and nitrogen isotope values in keratin, muscle protein, bone collagen, and bone carbonate (carbon only) for a diverse group of primate species (Table 2). We expected that differences in AA composition would drive ε* variation among proteinaceous tissues for both carbon and nitrogen, but that these differences would be small and consistent across individuals and species. We also expected that ε* values for carbonate versus collagen carbon would vary among species as a function of both diet and digestive physiology, and factors that correlate with these variables (body size, habitat, etc.).

Table 1 Tissue-tissue carbon and nitrogen fractionation values from previous research on mammals
Full size table
Table 2 Species, body mass of males and females, and provenience of specimens included in this study
Full size table

Background on variation in ε* values

Variation among proteinaceous tissues

The isotopic values of proteinaceous tissues within an individual could vary because (1) the concentration of different AAs varies among tissues, and (2) the isotopic composition of individual amino acids (AAs) shows considerable variation (19.9‰ average range for C, 24.4‰ average range for N; Fig. 1). This variability relates to isotopic differences among ingested AAs, differences in mammalian biosynthetic pathways for non-essential AAs, and the extent to which a mammal either synthesizes or incorporates a particular AA from its diet. This is a complex subject, but a few patterns have emerged. For carbon, glycine and metabolically-related AAs (serine, cysteine) are often 13C-enriched relative to other non-essential AAs, whereas essential AAs track variation in ingested AAs (Hare et al. 1991; Fogel and Tuross 2003; Jim et al. 2006). For nitrogen, there are a suite of AAs that are 15N-enriched with each trophic step (e.g., glutamate, asparate, alanine, isoleucine, valine, proline) and others that do not enrich (e.g., phenylalanine, lysine, glycine) (McClelland and Montoya 2002; Popp et al. 2007). Muscle (myosin) in humans, and most likely other primates, is dominated by 15N-enriched glutamate and alanine (Bergström et al. 1974). Collagen is mainly composed of 13C-enriched glycine (33%) and 15N-enriched aspartate (5%), glutamate (7%), and proline and hydroxyproline (33%). Primate keratin is dominated by 13C-enriched cysteine (~12–17%) and serine (~10%), and 15N-enriched glutamate (~17%) (Hrdy and Baden 1973; O’Connell et al. 2001).

Fig. 1
figure 1

Fractionation between individual amino acids and bulk collagen and tendon (Bos only) for carbon (a) and nitrogen (b) in mammals. The number of individuals for each taxon is provided in parentheses. Carbon data sources: Bos (1) (Hare and Estep 1983), Rattus (C3 = 2, C4 = 2) (Jim et al. 2006), Sus (C3 = 4, C4 = 4) (Hare et al. 1991), archaeological Bison (1), Oidocoleus(1), and Sylvilagus (1) (Fogel and Tuross 2003), and archaeological Procavia (2), Syncerus (2), Ovis (3), Raphicerus (2), Arctocephalus (2), whale (2), and humans (32) (Fogel and Tuross 2003; Corr et al. 2005). Nitrogen data sources: Bos (1) (Hare and Estep 1983), Sus (C3 = 4, C4 = 4) (Hare et al. 1991), archaeological Syncerus (2), Ovis (2), Raphicerus (2), Arctocephalus (2), whale (1), and humans (11) (Styring et al. 2010)

Full size image

Variation in carbonate-apatite ε* values

The measured mean carbon isotope difference between carbon in carbonate and collagen (ε13*carbonate–collagen) is ~7‰ or greater in wild large-bodied herbivores and ~3‰ in faunivorous animals (Table 1). There are two potential explanations for this difference that are not mutually exclusive. First, it could result from differences in dietary macromolecular composition (i.e., protein, lipid, and carbohydrate), which affect both diet-to-protein and diet-to-carbonate ε* values due to differing δ13C values among macromolecules, and differential routing of macromolecules to particular tissues. Second, it could result from differences in how animals digest plant and animal matter, which only affect diet-to-carbonate ε* values (Hedges 2003).

Apatite carbonate, which likely forms in isotopic equilibrium with blood bicarbonate, reflects carbon in bulk diet (i.e., a proportional mixture of carbon from all assimilated macronutrients) (Ambrose and Norr 1993; Passey et al. 2005). The isotopic composition of consumer proteins reflects that of dietary proteins (Ambrose and Norr 1993; Tieszen and Fagre 1993; Ambrose et al. 1997; Howland et al. 2003; Jim et al. 2004, 2006). Essential AAs must be routed directly from the diet, but depending on dietary protein concentration, non-essential AAs can also be routed into consumer tissues or synthesized using carbon from dietary carbohydrates, lipids and proteins. Theoretically, animals on high protein diets (e.g., faunivores) should route more carbon from dietary protein to tissue protein, whereas animals on low protein diets (e.g., many herbivores) should synthesize more non-essential amino acids de novo, incorporating carbon from lipid and carbohydrate as well as protein into their tissue protein (Fogel and Tuross 2003; Hedges 2003; Martínez del Rio and Wolf 2005). Additionally, because assimilation of 13C-depleted lipids could lower apatite δ13C values without affecting body protein δ13C values (due to routing), faunivores with fat-rich diets (such as seals) should have even smaller ε13*carbonate–collagen values (Krueger and Sullivan 1984; Lee-Thorp et al. 1989; Hedges 2003). Provided that these animals consume monoisotopic diets (e.g., only C3-derived foods), this should result in larger and smaller ε13*carbonate–collagen values in herbivores and carnivores, respectively. Whereas, all primates consume a dominantly vegetarian diet (Milton 1987), some genera such as Cebus, Daubentonia, Galago, and Microcebus can consume considerable amounts of animal matter (Milton and May 1976). Based on these dietary differences, we might anticipate that these taxa should have lower ε13*carbonate–collagen values than more herbivorous species. Importantly, controlled diet studies demonstrate that animals fed a mixture of C3, C4 and marine-derived macronutrients exhibit substantial variation in ε13*carbonate–collagen values (Table 1). Mixed diets are unlikely in the majority of wild primate species. However, this could be important for captive primates if they consume manufactured pellets containing a mix of C3 and C4 foods.

The isotopic composition of carbonate in bone apatite is also predicted to vary with the extent to which complex carbohydrates are fermented in the gut (Hedges 2003). During fermentation, bacteria break down structural carbohydrates, releasing appreciable amounts of hydrogen, CO2, and volatile fatty acids (VFA) (Jensen 1996). Some of the released CO2 can be reduced to form CH4. This process discriminates heavily against 13C, leaving the remaining CO2 13C enriched (Metges et al. 1990; Schulze et al. 1997). If even a small amount of this 13C-enriched CO2 enters the blood bicarbonate pool, it could increase the δ13C value of apatite carbonate which forms from this pool, thus increasing ε13*carbonate–diet and ε13*carbonate–collagen values (Passey et al. 2005). The δ13C value of collagen is not affected by methane production (e.g., Metges et al. 1990).

Ruminants have been shown to produce copious amounts of methane and large Δcarbonate–collagen values (e.g., Crutzen et al. 1986; Metges et al. 1990; Table 1). Although some large, non-ruminant herbivores such as camelids and horses also exhibit high levels of methane production and elevated Δcarbonate–collagen values (Crutzen et al. 1986; Langer 1987; Table 1), methane production in most simple-stomached species is trivial, despite the presence of methanogenic bacteria (Crutzen et al. 1986; Jensen 1996). Acidic conditions in the stomachs and small intestines of simple-stomached animals may prevent methane production, but neutral conditions in the posterior portions of the colon may be more amenable (Jensen 1996). Nevertheless, because gases formed near the end of the gastro-intestinal tract do not likely have time to diffuse into the blood stream, 13C-depleted methane produced in the posterior portions of the colon have a negligible effect on apatite δ13C values. Little is known about methane production in nonhuman primates. For the most part, it is doubtful that nonhuman primates would differ substantially from other simple-stomached animals. However, colobine monkeys could provide a possible exception. This subfamily of Old World Primates, has been likened to ruminants because they have large sacculated stomachs to facilitate microbial fermentation of leaves (Kay and Davies 1994). Primates with adaptations for caeco-colic fermentation, such as Alouatta palliata (Lambert 1998), may also have increased levels of methane production. This possibility is strengthened by the observation that horses, which are also caeco-colic fermenters, have Δcarbonate–collagen values (Sullivan and Krueger 1981; Kellner and Schoeninger 2007, Table 1).

Isotopic terminology

Isotope ratios are typically presented using δ notation, where

$$ \delta^{\text{H}} {\text{X}} = \left({\left({R_{\text{sample}}/R_{\text{standard}}} \right)- 1} \right) \times 1{,}000 $$
(1)

and R is the heavy-to-light isotope ratio in element X. It is expressed in parts per thousand (i.e., per mil, ‰). Carbon isotope values are reported relative to the V-PDB standard (a marine carbonate); nitrogen isotope values are relative to AIR. The offset, or fractionation, between two substances (a and b) is often expressed using Δ notation (Martínez del Rio et al. 2009), where

$$ \Updelta^{\text{H}} {\text{X}}_{{{\text{a}} - {\text{b}}}} = \delta^{\text{H}} {\text{X}}_{\text{a}} -\delta^{\text{H}} {\text{X}}_{\text{b}} $$
(2)

δ values are trivial to calculate and accurate so long as the differences in δ values among tissues are small. However, Δ values become less accurate as the differences in δ values among tissues increase. We choose to use alternative expressions, the fractionation factor (α) and isotope enrichment values (ε), which provide exact solutions and are not limited by the isotopic scale on which they are calculated (e.g., PDB vs. SMOW). Δ and ε values are nearly identical when isotopic differences among tissues are <1–2‰, but the two increasingly differ with increasing isotopic differences among tissues. When tissues are ≥10‰, Δ and ε values can differ by as much as 0.5‰ (Cerling and Harris 1999). To calculate ε, we first calculate α.

$$ \alpha_{{{\text{a}} - {\text{b }}}} = (\delta^{\text{H}} {\text{X}}_{\text{a }} + 1{,}000)/(\delta^{\text{H}} {\text{X}}_{\text{b}} + 1{,}000) $$
(3)
$$ \varepsilon_{{{\text{a}} - {\text{b}}}} = \left({\alpha_{{{\text{a}} - {\text{b}}}} - 1} \right) \times 1{,}000 $$
(4)

In animals, the observed α value between two tissues, or between diet and a tissue, is the net result of a large range of biochemical and transport phenomena, not the simple equilibrium and kinetic reactions for which isotopic fractionation factors are typically measured. We recognize the complexity of these physiological systems by denoting these as apparent fractionation factors (α*) and apparent enrichment values (ε*). When referring to values for a particular element, we will use ε13* for carbon and ε15* for nitrogen. Note that the sign of enrichment is dependent on which substance is in the numerator in Eq. 3. Hence ε* (and α* and Δ) values must always be reported with subscripts or explicitly defined.

Materials and methods

Sample acquisition

Tissues from captive and wild primates were acquired from cadaveric and osteologic collections in museums, universities and research field stations (Table 2). With a few exceptions, the animals were in good health at the time of death. The main manner of death for captive animals was electrocution, drowning, or short-term illness. However, a few individuals endured chronic illness, and some died at an advanced age. The manner of death for wild animals was largely unknown, but we were able to attribute the deaths of several individuals to predation or automobile impact (Electronic supplementary material, ESM, Table S1). The acquisition and analysis of tissues was approved by the Chancellor’s Animal Research Committee, University of California, Santa Cruz (approval nos. DOMIN 07.01 and ZIHL 97.12), and the Institutional Animal Care and Use Committee, Stony Brook University (approval no. 20001142). We combined our data with data from three preexisting datasets (Kibale primates: Carter, 2001; modern humans: O’Connell et al. 2001; Macaca mulatta: O’Regan et al. 2008).

Sample preparation and analysis

For each specimen, soft tissues were separated and lyophylized. Bone was defleshed; 20 mg were ground for the analysis of carbonate in bone apatite and 50 mg were crushed coarsely for extraction of collagen. For protein analysis, bone samples were treated with 5 ml of 0.5 N HCl for 72 h to remove the mineral fraction. Samples were rinsed 5× with water and dried. Lipids were removed from all proteinaceous tissues by repeated rinsing and sonication in 5 ml aliquots of petroleum ether for 15 min intervals until all visible lipids were removed. Samples were then rinsed 5× with ultrapure water and lyophilized.

With the exception of keratin—which was cleaned, cut to 1 mm lengths, and homogenized—all soft tissue samples were powdered using a mortar and pestle. Approximately 700 μg of ground soft tissue, homogenized keratin, or bone collagen were then sealed into tin boats and analyzed for δ13C and δ15N values on a ThermoElectron (Finnigan) Delta + XP continuous flow system coupled to an elemental analyzer (EA) at the University of California, Santa Cruz (UCSC) Stable Isotope Laboratory. Analytical precision (±1SD) based on 33 replicates of IAEA Acetanilide was −29.6 ± 0.1‰ for carbon and 1.1 ± 0.1‰ for nitrogen. We ran replicate samples for a subset of our specimens to determine sample precision. The average difference between the absolute value of 14 duplicate tissue samples was 0.2 ± 0.2‰ for carbon and 0.2 ± 0.2‰ for nitrogen. The average difference between the absolute value of five triplicate samples was 0.3 ± 0.1‰ and 0.3 ± 0.3‰ for carbon and nitrogen, respectively.

Bone carbonate samples were prepared using a modified technique from Koch et al. (1997). To oxidize organic materials, 1 ml of 30% laboratory-grade hydrogen peroxide (H2O2) was added to 20 mg of powdered sample and left for 48 h, then rinsed 5× with ultrapure water. To remove non-lattice bound carbonate, samples were reacted for 24 h with 0.5 ml of 1 M acetic acid (buffered to pH 5.0 with calcium acetate). Samples were again rinsed 5× with ultrapure water and lyophylized. For carbonate samples, 1.5 mg of powdered bone were put into steel cups and dried at 65°C for 1 h under vacuum. The samples were then analyzed on a Micromass Optima gas source mass spectrometer integrated with an Isocarb automated carbonate device. Samples were dissolved in 100% H3PO4 at 90°C, with concurrent cryogenic distillation of CO2 and H2O and automated CO2 admittance to the mass spectrometer for analysis. Reaction time was set at 740 s and blanks were run between samples. Accuracy and precision (±1SD) based on the international NBS 19 standard analyzed with samples was δ13C = 2.1 ± 0.1‰ (n = 18), very close to the known value of 2.0‰. The average difference between the absolute value of 10 duplicate samples was 0.3 ± 0.2‰.

Data analysis

We were not able to assess dietary composition or digestive physiology carefully for primates included in this study. Although it is tempting to divide primates into broad groups such as folivore, frugivore, or trophic omnivore, these dietary categories would likely be inaccurate for four reasons. First, the majority of primates are generalist primary consumers rather than strict folivores or frugivores. For example, the “frugivorous” lemur Varecia variagata can eat substantial amounts of leaves and fungus (A. Baden, personal communication). Conversely, the diet of Piliocolobus badius, a “folivorous” monkey, frequently contains fruit and flowers (Chapman et al. 2002a). Second, all primates have omnivorous tendencies (Fleagle 1999). In particular, many “frugivorous” primate species supplement their predominantly herbivorous diets either intentionally or inadvertently with insects or vertebrates. For example, among the “frugivorous” species, Hylobates lar and Lemur catta spend a substantial amount of time feeding on insects in addition to vegetation (Rowe 1996; Yamashita 2002), and Pan troglodytes consumes termites and red colobus monkeys (Boesch and Boesch-Achermann 2000). Third, primate diets can differ substantially between years and between localities (e.g., Chapman et al.2002a, b; González-Zamora et al. 2009). For example diets ranging from 49 to 87% leaves, and 13–49% fruits have been reported for Mexican populations of A. palliata (Cristóbal-Azkarate and Arroyo-Rodríguez 2007). Finally, we know little about the diets of most of our captive individuals, including the degree to which they were provisioned with chow.

Instead, we used one-way analysis of variance (ANOVA) and Tukey post-hoc tests of honestly significant differences (HSD) to detect differences in ε* values among habitats (e.g., captive, dry, or moist habitat) that may correlate with diet quality. Diet and digestive physiology may covary with two other variables that we were able to assess: body size and phylogenetic relatedness. In general, diet quality decreases with increasing body size (e.g., Kleiber 1961). More folivorous primates have longer and more complex guts than frugivorous or insectivorous primates (Chivers and Hladik 1980), and primates that are more closely related should have more similar digestive physiology. We used Pearson correlation coefficients to determine if ε* values correlate with body mass, and we tested for the potential confounding effects of phylogenetic relatedness by using the primate phylogeny of Bininda-Emonds et al. (2007) and the PDAP module of Mesquite version 2.5 (Maddison and Maddison 2008) to calculate phylogenetic independent contrasts.

Additionally, we used one-way ANOVA and Tukey HSD to detect differences in ε* values among manners of death (grouped into abrupt, short-term illness, long-term illness, and unknown). We used independent sample t tests to detect differences in ε* values between sexes. Detailed age information for strepsirrhines from the Duke Lemur Center allowed us to calculate percent lifespan lived. We grouped these individuals into five equal age classes, and used one-way ANOVA and Tukey HSD to detect differences in ε* values among age classes. Although there are no theoretical expectations for ε* differences among sexes, age classes, or manners of death, we sought to verify that metabolic or dietary differences between these different groups do not affect ε* values. Such comparisons are often missing from tissue fractionation and enrichment studies. Analyses were performed using JMP version 5.0.1a for Macintosh with the significance of all tests set at α ≤ 0.05.

Results

Mean and standard deviations for each species are presented in Table 3, and raw δ13C, δ15N, and ε* values are available in ESM Table S1. Patterns of apparent enrichment varied little within the Strepsirrhini (Fig. 2) and Haplorrhini (Fig. 3). Across primates, the ε13*collagen–keratin, ε13*collagen–muscle, ε13*muscle–keratin and ε13*carbonate–collagen values did not differ (p > 0.05). Whereas, the ε15*collagen–keratin and ε15*muscle–keratin values also did not differ (p > 0.05), ε15*collagen–muscle values did (t = −2.42, df = 18, p = 0.027); however, this result was driven by two Eulemur and Microcebus individuals. The removal of these two individuals resulted in no overall difference among species for ε15*collagen–muscle (p > 0.05).

Table 3 Mean carbon and nitrogen apparent enrichment (ε*) values ± one standard deviation for all species and genera
Full size table
Fig. 2
figure 2

Mean ε*collagen–keratin, ε*collagen–muscle, ε*muscle-keratin, and ε*carbonate-collagen, for carbon (ε13*) and nitrogen (ε15*) ± 1 standard deviation for each strepsirrhine genus. Phylogeny based on Orlando et al. (2008). Illustrations by Stephen D. Nash/Conservation International, used with permission

Full size image
Fig. 3
figure 3

Mean ε*collagen–keratin, ε*collagen–muscle, ε*muscle–keratin, and ε*carbonate-collagen for carbon (ε13*) and nitrogen (ε15*) ± 1 standard deviation for each haplorrhine genus. Phylogeny based on (Groves 2001). Homo sapiens data from O’Connell et al. (2001), Macaca mulatta data from O’Regan et al. (2008), and P. badius, Cercopithecus ascanius, and wild Pan troglodytes data from Carter (2001). Illustrations by Stephen D. Nash/Conservation International, used with permission

Full size image

We found small but significant variation among habitat types for both carbon and nitrogen ε*collagen–keratin values (carbon: F 2,82 = 3.36, p = 0.040; nitrogen: F 2,81 = 6.73, p = 0.020). Captive animals had significantly larger ε*collagen–keratin values than those from moist habitats (Table 4). Our results for ε15*collagen–muscle values showed a similar pattern (F 2,44 = 7.03, p = 0.0023), but ε13*collagen–muscle values did not differ significantly among habitat types (p > 0.05). Mean ε13*muscle–keratin and ε13*carbonate–collagen values also did not differ significantly among habitats.

Table 4 Mean carbon and nitrogen apparent enrichment (ε*) values ± one standard deviation for primates living in dry, moist, and captive settings
Full size table

With the exception of ε13*collagen–muscle, ε* values between proteinaceous tissues did not correlate with body size (p > 0.05; Table 5). If we excluded two captive Microcebus individuals, the relationship between ε13*collagen–muscle and body size was insignificant (r 2 = 0.06, p = 0.10). The relationship between ε13*carbonate–collagen and body mass was significant (r 2 = 0.031, p = 0.038; Fig. 4; Table 5). However, because the slope is near 0 and the r 2 value is low, we suspect that this result is an artifact of sample size. The range in ε13*carbonate–collagen values for the smallest and largest species (Microcebus spp. and Gorilla gorilla) are similar (4.5–6.9 and 5.5–7.1‰, respectively), and the lowest and highest ε13*carbonate–collagen values, 3.6 and 8.6‰, come from two similar-sized species, P. badius and A. palliata (Tables 2 and 3).

Table 5 Regression results for ε* versus the natural logarithm of body mass
Full size table
Fig. 4
figure 4

The relationship between the natural log of body mass (kg) and ε13*carbonate-collagen (ε*carbonate-collagen = 5.24 + 0.163*ln body mass, r 2 = 0.031, p = 0.038)

Full size image

Finally, ε* values did not differ among males and females (p > 0.05; ESM Table S2), manner of death (p > 0.05; ESM Table S3), or age class (p > 0.05; ESM Table S4). Given the overall consistency of our results, we combined data from all individuals and calculated mean primate ε* values between all proteinaceous tissues, and between carbonate and collagen (Table 6).

Table 6 Suggested ε* values for comparing different primate tissue types
Full size table

Discussion

Variation in ε* values among proteinaceous tissues

We expected some variation based on differences in amino acid compositions, but that such differences would be small and consistent across taxa. In line with our expectations, we found small (≤1‰) ε* values between collagen and muscle, collagen and keratin, and muscle and keratin for both carbon and nitrogen (Table 6). These mean values are smaller than the majority of the Δ values reported for captive or wild animals (Table 1). It appears that because each tissue is composed of multiple AAs, the effects of isotopic differences among specific AAs are minimized. For example, relatively 13C-enriched glycine in collagen, serine, and cysteine in keratin, and glutamate in muscle may be driving similar δ13C values in all three tissues (Fig. 1). O’Connell et al. (2001) suggest that the relatively elevated levels of serine and threonine in keratin (6–7% vs. ~2% in collagen) tend to lower keratin δ15N values relative to collagen. The 15N-enriched glutamate in muscle may increase its δ15N values relative to keratin.

Variation in ε13*carbonate–collagen

We had anticipated that differences in diet (e.g., δ13C differences in dietary sources, differences in herbivory vs. faunivory), and digestive physiology (e.g., degree of fermentation) would lead to differences in ε13*carbonate–collagen. Our results, however, suggest that all primates have comparable ε13*carbonate–collagen values regardless of variation in the variables that covary with diet and digestive physiology such as phylogeny, body size, and habitat. Our mean ε13*carbonate–collagen value of 5.6‰ for primates is similar to the mean fractionation factor (Δcarbonate–collagen) for wild omnivores (5.5‰), captive omnivorous rodents fed mixed and uniform diets (5.5 and 5.4‰, respectively), and captive pigs fed uniform diets (6.0‰; Table 1). The mean primate ε13*carbonate–collagen value is larger than the Δcarbonate–collagen value reported for carnivores (3.0‰), and smaller than the reported values for both wild ruminant and non-ruminant herbivores (9.0 and 7.8‰, respectively; Table 1).Footnote 1 Based on the consistency of our results, we conclude that (1) δ13C values for dietary protein did not differ substantially from whole diet δ13C values for either captive or wild primates, and (2) that microbial fermentation, to the extent that it occurred in the primates in our study, failed to significantly label the blood pool with 13C-enriched bicarbonate, irrespective of differences in habitat, gut physiology or body size.

Diet

We had anticipated that more herbivorous primates would have larger ε13*carbonate–collagen values than more faunivorous primates. We found some variation but no consistent trends. We found no differences in ε13*carbonate–collagen values with body size, despite probable dietary differences between the smallest primates, Galago and Microcebus spp., which likely consumed more insect matter, and the largest primates, Gorilla, Pan, and Pongo, which likely consumed more vegetation. A single aye-aye (Daubentonia madagascariensis), which relies largely on invertebrate prey, had a carnivore-like ε13*carbonate–collagen value of 3.7‰. However, ε13*carbonate–collagen values for the white-faced capuchin (Cebus capucinus), which also consumes animal matter, resembled the overall primate mean (5.0 and 5.8‰ in dry and moist habitat, respectively). Our results likely reflect underlying dietary similarities among all primate species. In spite of apparent differences in the consumption of animal matter, all primates have a predominantly vegetarian diet (Milton 1987). These results agree with the recent findings of Smith et al. (2010), who showed that collagen δ13C values did not differ between male and female chimpanzees despite observations that males consumed substantially greater amounts of red colobus meat. These authors speculated that either meat consumption did not noticeably affect male collagen δ13C values, or that consumption of termites elevated female δ13C values (Smith et al. 2010).

With the exception of a few M. mulatta individuals (O’Regan et al. 2008; ESM Table S1), none of the wild primates in our study ate C4 or marine foods (mean δ13C apatite = −16.7‰ ± 1.4, n = 105; mean δ13C collagen = −22.1‰ ± 1.5, n = 110). Conversely, the majority of our captive primates appear to have incorporated some C4 foods into their diets (mean δ13C apatite = −12.7‰ ± 1.6, n = 36; mean δ13C collagen = −18.5‰ ± 1.8, n = 45; ESM Table S1). However, despite this addition of C4 foods, ε13*carbonate–collagen values for captive and wild primates do not differ (Table 4). We cannot assess dietary composition in the captive primates quantitatively. Nevertheless, our results suggest that protein and whole diet δ13C values did not differ substantially for captive animals. We note that the laboratory diets for some of the controlled feeding studies listed in Table 1 were designed to maximize possible isotopic differences among tissues. The majority of these diets were not designed to maintain healthy individuals, and most laboratory animals were sacrificed at a young age. In contrast, captive primates are given balanced diets designed to maintain their health and increase their longevity. As a result, diets for captive primates tend to be much more isotopically restricted than experimental laboratory diets consisting of mixed C3, C4, and marine components. In line with this reasoning, the range in captive primate ε13*carbonate–collagen values (3.8–7.1‰) is similar to the range in Δcarbonate–collagen values reported for captive animals fed isotopically homogenous diets (4.5–6.7‰), but much smaller than the ranges reported for captive animals fed experimental diets incorporating a mix of C3, C4, and marine components (−0.8 to 11.3‰, Table 1).

Physiology and fermentation

Despite differences in diet, all primates ferment their food to some degree. More folivorous, gumnivorous, and faunivorous primates break down the structural carbohydrates in vegetation, plant exudates, and arthropod exoskeletons, respectively (Lambert 1998). Nevertheless, carbohydrate fermentation in primates does not appear to produce enough methane and associated 13C-enriched CO2 to significantly label blood bicarbonate or bone carbonate. We might have anticipated that taxa with long measured retention times such as Gorilla, Pongo, Lophocebus, Chlorocebus, and Cercopithecus might have larger ε13*carbonate–collagen values (Kleiber 1961; Langer 1987). Increased retention time may increase methane production during fermentation, and the degree to which 13C-enriched CO2 diffuses into the blood (Kleiber 1961; Langer 1987). Our results do not support these expectations. Mean ε13*carbonate–collagen values for the hominoids (6.2 and 6.0‰, respectively), and the cercopithecines (5.9, 5.9, and 4.3‰, respectively; Table 1) are comparable to or only slightly larger than our mean primate ε13*carbonate–collagen value (5.6‰). Conversely, the ε13*carbonate–collagen value for Ateles geoffroyi, which has a fast retention time (6.8‰), is substantially larger than the average primate value.

We had also anticipated that colobine monkeys, represented by P. badius and Semnopithecus entellus, and the ateline monkey A. palliata, would have higher ε13*carbonate–collagen values associated with fermentation in their enlarged stomachs and caeca, respectively. Our results do not support these expectations. Despite their potential for increased levels of methane production, both wild and captive colobine monkeys in our dataset had ε13*carbonate–collagen values comparable to other primate species (Fig. 2; Tables 3 and ESM S1). Our lowest reported ε13*carbonate–collagen value (3.6‰) is from a wild P. badius individual. This result is in agreement with the lack of methane production observed in two wild Colobus polykomos individuals (Ohwaki et al. 1974). It appears that, despite their large “ruminant-like” stomachs, colobines produce little to no methane and associated 13C-enriched CO2, and their digestion resembles that of small simple-stomached animals rather than ruminants.

We did find a large mean ε13*carbonate–collagen value for the mantled howling monkey (A. palliata) in a rainforest habitat (7.6‰). However, we also found a large ε13*carbonate–collagen value (8.4‰) for rainforest-dwelling black-handed spider monkeys (A. geoffroyi), which does not have a gut designed for extensive fermentation (Chivers and Hladik 1980). Intriguingly, these two species had comparable but lower ε13*carbonate–collagen values similar to our primate mean in a seasonally dry forest habitat (5.7 and 5.3‰, respectively). Although A. palliata and A. geoffroyi are typically categorized as folivorous and frugivorous, respectively, both of these species have been observed to have highly variable diets (Cristóbal-Azkarate and Arroyo-Rodríguez 2007; González-Zamora et al. 2009). It is possible that they shared dietary items in the rainforest habitat that were rich in non-starch polysaccharides (NPS), the breakdown of which has been associated with increased methane production in pigs (Jensen 1996). Alternatively, it is possible that the two species shared a food item with elevated δ13C values, (e.g., a CAM plant) which increased their whole diet δ13C values without affecting their dietary protein. This result is interesting and suggests that future work examining species-specific ε13*carbonate–collagen values with varying diets could be enlightening. Nevertheless, these are the only two taxa that demonstrate substantial differences in apparent enrichment values among habitats. For example, ε13*carbonate–collagen values for C. capucinus from the same two habitats are much more similar (5.8 and 5.0‰ in the moist and dry habitats, respectively). Pan troglodytes exhibits similar ε13*carbonate–collagen values among captive and moist habitats (6.1 and 6.6‰, respectively), and all Microcebus taxa have similar ε13*carbonate–collagen values in all three habitat types (5.4, 6.0, and 5.7‰ in captive, moist, and dry habitats, respectively). Based on the data available, we therefore advocate using our mean primate ε13*carbonate–collagen value (5.6‰) to compare collagen and carbonate δ13C values among primates.

Verification of ε* values

An important outcome of our analyses is the ability to determine mean apparent enrichment values that can be used in existing and future comparisons based on mixed tissues or samples. To validate primate ε* values, we estimated keratin δ13C and δ15N values by applying mean ε* collagen–keratin values to measured collagen δ13C and δ15N values for wild primate populations not included in our apparent enrichment dataset. We then compared these estimated keratin values to measured keratin values from different individuals within the same wild populations (Table 7). Compellingly, the range of estimated keratin isotope values closely matches the measured keratin isotope values.

Table 7 Comparing estimated keratin δ13C and δ15N values with measured keratin δ13C and δ15N values from wild and captive primate populations
Full size table

Conclusions

We have presented data on the apparent isotopic enrichment in carbon and nitrogen isotopes between collagen and keratin, collagen and muscle, and apatite carbonate and collagen in primates. Primates are an extremely diverse group of animals in terms of diet, body size, and gut morphology, yet ε* values are relatively invariant across the order. We recommend applying our calculated mean ε* values when comparing isotope values from different modern primate tissues. Additionally, using these mean apparent enrichment values will be essential for accurately predicting how the isotopic niches of extinct primates compare with those of modern extant primates.