A chimpanzee enamel-diet δ¹³C enrichment factor and a refined enamel sampling strategy: Implications for dietary reconstructions

Abstract

Reconstructing diets from stable carbon isotopic signals in enamel bioapatite requires the application of a δ¹³C enamel-diet enrichment factor, or the isotopic offset between diet and enamel, which has not been empirically determined for any primate. In this study, an enamel-diet enrichment factor (ε*_enamel-diet) of 11.8 ± 0.3‰ is calculated for chimpanzees (Pan troglodytes) at Ngogo in Kibale National Park, Uganda, based on a comprehensive isotopic assessment of previously analyzed dietary plant data and new isotopic analyses of enamel apatite. Different enamel sampling methods are evaluated to determine the potential influence of weaning on isotopic enamel values and dietary interpretations. The new chimpanzee enrichment factor and a sampling strategy that excludes teeth that formed before weaning completion are applied to all known chimpanzee δ¹³C_enamel data, either previously published or newly derived in this study, resulting in a dietary range of almost 6‰ across all chimpanzees sampled. This new chimpanzee enamel-diet enrichment factor is then used to reassess dietary reconstructions of 12 fossil hominin species whose isotopic enamel signatures have been determined. Results reveal hominin diets that are isotopically more positive than previously reconstructed, highlighting the widespread contribution of ¹³C enriched C₄/CAM (crassulacean acid metabolism) resources in fossil hominin diets and emphasizing the broad use of these resources during human evolution. These findings stress the importance of ascertaining and employing an appropriate enrichment factor for dietary reconstructions of specific taxa as well as standardizing the sampling protocol for tooth enamel in isotopic paleodietary reconstructions.

1. Introduction

The axiom that you are, isotopically, what you eat comes with caveats. As food components are digested and routed to specific body tissues, the carbon isotopic ratio (δ¹³C) is systematically altered. The isotopic offset between the δ¹³C of the diet and tissues is variable, depending on the macronutrients required to build and maintain the destination tissue. This offset is also contingent on the digestive physiology of organisms and the nature of the diet. Quantitatively determining this offset therefore is a complex undertaking, especially for a long-lived, eclectic foraging hominoid such as a chimpanzee. This introduction provides a framework outlining some of the key theoretical and methodological considerations for this research.

1.1. Stable isotopic dietary ecology

Stable carbon isotopic paleoecology represents a developing methodology that can provide insights into past diets, vegetation types, and environmental conditions. This approach is based on the differential uptake by plants of two stable isotopes of carbon (¹³C and ¹²C) during photosynthesis. The ratio of the heavier to the lighter isotope of carbon (¹³C/¹²C) is expressed as δ¹³C in parts per thousand, or parts per mil (‰). Plants that utilize the C₃ pathway include most fruits, trees, shrubs, and temperate grasses and have δ¹³C values ranging from ca. −20‰ to −37‰ (Deines, 1980; Kohn, 2010). This isotopic range for C₃ plants reflects a number of factors, including water stress, irradiance, canopy nature (subcanopy, canopy periphery, height in canopy), and plant component (Kohn, 2010). In East Africa, C₃ plants tend to average ca. −25‰ in savannah/woodland environments, −28‰ in open canopy forests, and < −31‰ in closed canopy forests (Deines, 1980; O’Leary, 1988; Peters and Vogel, 2005). C₄ plants, mainly tropical grasses and sedges, range from ca. −10‰ to −14‰ (Cerling and Harris, 1999). For plants that utilize CAM photosynthesis, δ¹³C values are intermediate, typically closer to the C₄ range (O’Leary, 1988; Farquhar et al., 1989).

The stable carbon isotopic composition of enamel (δ¹³C_enamel) has been used extensively to reconstruct the dietary niches of fossil taxa with several paleoecological goals, including differentiating grazing/browsing/mixed feeding guilds (e.g., Kingston and Harrison, 2007), characterizing niche partitioning (e.g., Bibi, 2007; Macho and Lee-Thorp, 2014), linking diet with locomotor and other adaptive complexes (e.g., White et al., 2009), and establishing evolutionary foraging trends in lineages (e.g., Cerling et al., 1999; Lister, 2013). Central to these strategies is translating the δ¹³C_enamel value to a particular diet, despite the limited resolution afforded by this approach and the complexity of diets. Attributing the enamel isotopic signature to a specific resource or a particular niche is, in part, complicated by an isotopic offset, or fractionation, between the diet and enamel. This offset typically leads to an enrichment of ¹³C in the consumer tissue relative to the diet and is designated as ε*_enamel-diet (Cerling and Harris, 1999; Passey et al., 2005; Sharp, 2007).

The enrichment factor ε between two substrates A (enamel) and B (diet) is defined as:

$$\epsilon {*}_{\text{A}-\text{B}}=\left[{\alpha }_{\text{A}-\text{B}}-1\right]\times 1000$$ (Equation 1)

where the fractionation factor α between A and B is: α_A-B = (1000 + δ_A) / (1000 + δ_B) and δ_A = δ¹³C_enamel and δ_B = δ¹³C_diet. The asterisk for the enrichment factor (ε*) indicates a non-equilibrium fractionation assumed between A and B.

In several studies, this offset has been reported as simply the difference between the isotopic value of the enamel and the diet: Δ_enamel-diet = δ¹³C_enamel - δ¹³C_diet. While this calculation yields a good approximation between the two phases, Δ_enamel-diet is not constant across a range of δ¹³C values, and the difference between ε* and Δ_enamel-diet ranges from about 0.1‰ for C₄ diets to close to 0.5‰ for closed-canopy C₃ herbivores (Cerling and Harris, 1999; Passey et al., 2005). In situations where Δ_A-B is greater than about 10‰, as is the case with the isotopic offset between diet and enamel and when the isotopic compositions of the two phases are not in chemical equilibrium, it is more accurate to use the enrichment factor ε (Sharp, 2006).

1.2. Enrichment variability

Establishing the specific nature and magnitude of this enrichment for different organisms has been problematic because it is difficult to: (1) determine a cumulative isotopic composition of the total diet; (2) measure the δ¹³C of constituent macronutrients in food items; (3) quantify potential fractionation during digestion and fermentation; and (4) track the routing of carbon to various body tissues. This study focuses on enamel bioapatite only, eschewing bioapatite from bone or dentine, in an effort to remove as many sources of potential variability from the ε*_enamel-diet calculation as possible. While several controlled feeding experiments and field studies have provided valuable insights (De Niro and Epstein, 1978; Ambrose and Norr, 1993; Tieszen and Fagre, 1993; Cerling and Harris, 1999; Balasse, 2002; Howland et al., 2003; Jim et al., 2004; Passey et al., 2005; Zazzo et al., 2010; Han et al., 2016), they have also generated an almost 6.0‰ interspecific ε*_enamel-diet range in mammals. Giant pandas (Ailuropoda melanoleuca) may have an ε*_enamel-diet of 10.0‰ (Han et al., 2016), while the ε*_enamel-diet of cattle (Bos taurus) may be as large as 15.7‰ (Passey et al., 2005). It has been suggested that the extent and nature of methanogenesis is a primary factor generating this level of interspecific variation (Hedges, 2003; Passey et al., 2005). The δ¹³C value of methane produced in the forestomaches of ruminants ranges from −50‰ to −80‰ (Rust, 1981; Schulze et al., 1998), requiring that carbon remaining in the digestive tracts be correspondingly enriched in ¹³C. Large bodied ruminants with relatively long and complex digestion are generally characterized by diet-enamel enrichment ranging from 12–15‰ while mammals producing little to no methane, such as voles, have an enrichment of about 10.6‰ (Passey et al., 2005). Tejada-Lara et al. (2018) documented a correlation between body mass and ε*_{bioapatite-diet}, indicating that at least some proportion of these enrichment factors is size dependent. The most widely used and cited isotope enrichment factor, 14.1 ± 0.5‰, is based on isotopic analyses of tooth enamel of a variety of wild and captive large ungulate mammals from Africa, Asia, and North America (Cerling and Harris, 1999). While furnishing an estimate for a ruminant offset, this value is based on averages of δ¹³C signatures of vegetation and mammalian bioapatite and conflates differences in enrichment between and within mammalian species due to factors such as variable foraging strategies, food selectivity, plant isotopic heterogeneity, and digestive physiology. As Cerling and Harris (1999) noted, these considerations indicate that more research is required to fully understand the fractionation of dietary carbon as it is incorporated into body tissue.

Given considerable interspecific δ¹³C_enamel-diet enrichment variation, it is important to characterize the potential enrichment factor as closely as possible when inferring the dietary patterns of fossil taxa. In the case of fossil hominins, an ε*_enamel-diet factor specific to large-bodied ruminants (13–14‰) has been applied in previous research (e.g., Cerling et al., 2013; Wynn et al., 2013; Levin et al., 2015; Garcia et al., 2015; Ludecke et al., 2018), generally with the acknowledgement that it may not be appropriate for reconstructing diets (see Schoeninger, 2014 for critique). Enrichment factors for primates would theoretically be most relevant for reconstructing early hominin diets, given their shared ancestry and the shared tendency for eclectic and flexible feeding behavior. However, controlled feeding studies using primates have not been performed, and diet-enamel enrichment data do not exist from field research conducted on these animals. A cumulative isotope enrichment of 12.8 ± 0.6‰ was calculated for five cercopithecoid primate taxa (n = 6) and one hominoid, in this case Pan (n = 1), from the Ituri Forest in the Democratic Republic of the Congo (Cerling et al., 2004). However, this value was based on enamel isotopic signatures and average δ¹³C values of canopy fruits, nuts and leaves without specific attribution of food types and how frequently individuals consumed them. Several estimated enrichment factors for primates, ranging from 10–13‰, have been suggested given that these animals possess relatively high-quality diets and for the most part display limited rumination (Kellner and Schoeninger, 2007; Smith et al., 2010; Loudon et al., 2016; Nelson and Hamilton, 2017; Quinn, 2019). To date, none of these estimates have been validated through a systematic study of a primate species whose diet is well characterized and for which the δ¹³C values of its foods and tissues have been analyzed.

In developing a comparative approach to interpreting isotopic dietary ecology of extinct taxa, it is also critical to focus specifically on enamel apatite, as other biological substrates typically analyzed in modern organisms (e.g., hair, dentine, bone collagen/apatite, feces) are either not usually recovered from the fossil record or do not retain the intact biogenic isotopic signal during fossilization (Kohn and Cerling, 2002). Conversion of the isotopic signatures of these tissue values to δ¹³C_enamel values is problematic, as dietary fractionation for other tissues has also not been well characterized, and each level of conversion contributes additional amounts of unknown variability. This ambiguity is compounded by uncertainty due to variable diet-tissue enrichment in different species and individuals, seasonal and life history shifts in diet, and differential macronutrient routing (Louden et al., 2016). Finally, it is also important to focus isotopic analysis on enamel forming after weaning that reflects adult dietary signals to prevent conflation of isotopic contributions of pre- and postweaning diets (see section 1.4 below).

1.3. Chimpanzees as an analog for early hominins

Having shared a recent common ancestor (Muller et al., 2017), chimpanzees arguably represent an appropriate extant taxon whose ε*_enamel-diet can be used to reconstruct the diets of fossil hominins. Acknowledging the usefulness of chimpanzee dietary isotopic data for interpreting the δ¹³C_enamel values of fossil hominins, several authors have made efforts to determine the isotopic baseline of potential and/or actual food items of chimpanzees at several sites (e.g., Oelze et al., 2016; Wessling et al., 2019). However, without verifying which potential food items and specific quantities chimpanzees are eating at each site, reconstructing the dietary input value remains elusive. In contrast, long-term field research on chimpanzees conducted at Ngogo in Kibale National Park, Uganda (Mitani, 2020), furnishes an opportunity to estimate a composite δ¹³C_diet that can be linked directly to δ¹³C_enamel values of individual animals. These data, in turn, can be used to determine an ε*_enamel-diet for these chimpanzees. It is important to acknowledge that, given differences in dietary ecology and habitat preferences, chimpanzees are not a direct analog for fossil hominins and differences may be reflected in the enrichment factor(s) as well.

To establish a suitable enrichment factor that can be extended to reconstructing fossil hominin diets, it is also important to use only enamel that forms after weaning completion. In chimpanzees, enamel apatite that mineralizes before weaning is complete will incorporate a nursing signal which can result in δ¹³C_enamel values up to 2‰ more negative relative to enamel forming in equilibrium with the postweaning/adult diet (Smith et al., 2010; Malone, 2019). However, age of weaning is difficult to determine using behavioral observations alone as weaning is a variable and lengthy process in chimpanzees and other primates (Lee, 1996). Recently, a novel method to estimate weaning age, using the timing of the convergence of fecal nitrogen isotope values between mothers and infants, has been applied to the Ngogo chimpanzees and indicates that weaning is complete by 4.0–4.5 years (Badescu et al., 2017). The diets of juvenile chimpanzees (aged 5–10 years) do not differ from those of adults in the Kanyawara chimpanzee community, also in Kibale National Park (Bray et al., 2018), and this finding has been replicated in ongoing research on the Ngogo chimpanzees (Lee, unpublished data). As upper and lower M3 crowns begin forming when chimpanzees are ca. four years old and may continue forming for five years or more (Reid et al., 1998), this tooth provides the only unequivocal source of enamel with a postweaning/adult dietary isotopic signal in this species. The latest forming enamel of chimpanzee upper and lower premolars and M2s also potentially record a postweaning dietary signal, a possibility explored below in documenting intratooth δ¹³C_enamel variation. Throughout this paper, when teeth are referred to without a super- or subscript it denotes a mixture of multiple upper and lower teeth.

1.4. Enamel sampling protocol issues

Although the δ¹³C_enamel value for a given tooth or enamel fragment is typically reported as representative of the adult dietary signal of an individual, this value is influenced by variable aspects of diet and life history. An intertooth δ¹³C_enamel difference of up to 2‰ or more has been documented among chimpanzee molars as described in section 1.3. This variability may be related to seasonal or annual shifts in diet, and perhaps more importantly, and as indicated above, to the weaning process. In acknowledgement of this potentially confounding variation, isotopic studies have attempted to present values of ‘bulk’ dietary signatures, which theoretically provide an average signal of enamel that has formed after weaning completion. A ‘bulk’ sample ideally includes enamel that spans a significant interval of tooth developmental (>1 year), in effect averaging any dietary variation. A ‘bulk’ sample is typically taken by mechanically extracting enamel fragments, or by collecting enamel powder by drilling holes spanning the crown from the occlusal to cervical ends, and then homogenizing the sample, or homogenizing the values obtained from multiple samples. This procedure, however, has never been formally standardized. Given the opportunistic sampling required for valuable museum specimens (including fossil hominins) and the shift to progressively smaller samples (e.g., ~600 μg), data sets may not uniformly provide ‘bulk’ dietary signatures. Additionally, a ‘bulk’ dietary signal obtained in the ways described above may not be equivalent to the postweaning/adult diet if the taxon in question continues nursing during part or all of the formation of the M3 crown. The conflation of pre- and postweaning dietary signals, or of dietary signals derived from rarely consumed fallback foods, can result in considerable differences in δ¹³C_enamel values, and hence variable interpretations. This study therefore uses ‘postweaning/adult diet’ to refer to the signal targeted within the chimpanzee teeth, in acknowledgement that this is not equivalent to a ‘bulk’ M3 value and because the Ngogo chimpanzee juveniles have a diet indistinguishable from the adults.

Extracting postweaning dietary enamel ultimately requires an understanding of the dental developmental timeline of a species as well as behavioral data relating to the weaning process and the achievement of a postweaning/adult diet. Given the complexity of tooth mineralization and timing, which differs among taxa, studies that sample enamel need to consider multiple factors when selecting the appropriate sampling protocol for their question of interest. These include: (1) the orientation of tooth mineralization; (2) enamel growth rates and their relationship to the internal geometry of tooth microstructures; and (3) additional time introduced by the secondary formation stage of maturation.

Tooth development is reflected in the microstructure of the enamel, with variable timing of mineralization and different growth rates characterizing different parts of the tooth. Enamel mineralizes from the occlusal to the cervical end and from the enamel-dentine junction (EDJ) to the outer enamel surface (OES). Growth rates also vary, resulting in different amounts of time represented by cervical versus occlusal volumes of enamel when sampled using drilling methods. A second phase of enamel formation, called maturation, follows a different developmental trajectory, cross-cutting the orientation of initial mineralization (Suga, 1982; Passey and Cerling, 2002; Hoppe et al., 2004; Tafforeau et al., 2007; Green et al., 2017). The majority of the final mineral content of enamel apatite is incorporated into this tissue during maturation (Passey and Cerling, 2002; Smith et al., 2011; Simmer et al., 2012). It may take several months to a year or more after initial mineralization for maturation to be complete, depending on the taxon and the area of the tooth in question, especially as maturation only begins once full enamel thickness has been attained during the initial mineralization phase (Green et al., 2017). However, the amount of additional time represented by the period of enamel maturation in any given drilled sample is currently unknown. Due to the combination of these mineralization variables, it remains difficult to determine whether enamel samples from different species are isotopically equivalent with respect to time represented or developmental stage.

1.5. Early hominin dietary ecology: a case study

Many of the key innovations during human evolution have been linked to shifts in dietary ecology. These innovations include increased terrestriality/bipedality associated with niches in more open environments (Rodman and McHenry, 1980; Lovejoy, 1981; Wheeler, 1992), tool production linked to extractive foraging involved in scavenging and hunting (Wynn, 2002; Pobiner, 2015), and encephalization related to increased dietary quality (Aiello and Wheeler, 1995; Isler and Van Schaik, 2014; Navarette et al., 2011). Furthermore, it has become apparent that early hominins likely occupied multiple dietary niches, as evidenced by biomechanical reconstructions based on craniodental variation (e.g., Daegling and Grine, 2017; Rak, 2014), microwear analyses (e.g., Walker, 1981; Ungar and Sponheimer, 2013), occlusal morphology (e.g., Kay, 1984), and dental structure (e.g., Schwartz, 2000). Isotopic analyses of fossil hominin enamel (δ¹³C_enamel) have also provided key evidence consistent with niche variation, in particular supporting the hypothesis that early hominins consumed ¹³C-enriched foods, including water stressed C₃ or C₄/CAM plants in savannah/open woodland ecosystems or edaphic sedge grasslands, and/or animal foods (see Discussion; Lee-Thorp et al., 1994, 2012; van der Merwe et al., 2008; Ungar and Sponheimer, 2011; Grine et al., 2012; Cerling et al., 2013; Sponheimer et al., 2013; Wynn et al., 2013; Stewart, 2014; Levin et al., 2015; Paine et al., 2018). Much discussion has focused on identifying specific dietary items that may have contributed to isotopic profiles that are significantly ¹³C-enriched relative to extant hominoids. Critical to interpreting these dietary signatures is characterizing the physiological processes that alter the carbon isotopic value of the diet as it is incorporated into enamel bioapatite. In addition, it is important to assess the timing of enamel mineralization to ensure that isotopic sampling of enamel reflects adult dietary signals, because the trophic effects of nursing alter isotopic signatures significantly.

1.6. Overview of research agenda

With these considerations in mind, the goals of this study are to: (1) calculate the carbon isotopic offset between diet and enamel for chimpanzees at Ngogo in Kibale National Park, Uganda; (2) establish a sampling protocol that emphasizes analyses of enamel forming after weaning completion; and (3) apply these findings to interpretations of the δ¹³C_enamel values of modern chimpanzees and fossil hominins. To achieve these goals, an initial estimate of a composite δ¹³C dietary input value for chimpanzees at Ngogo was determined, based on previous studies of diet composition and diversity over 15 years (Watts et al., 2012) and analyses of the isotopic composition of 33 of the top ranked food items. The δ¹³C_enamel values of four natal male chimpanzees from this group are then determined, with special emphasis placed on minimizing potential biases created by different sampling protocols and intra- and intertooth variation. Based on the δ¹³C values of the diet and enamel from the Ngogo chimpanzees, a new diet-enamel enrichment factor is derived for this species. This new ε*_enamel-diet is then used to reconstruct and evaluate the isotopic values of diets of extant populations of chimpanzees throughout equatorial Africa. To do so, previously published δ¹³C_enamel values were used from 44 Pan troglodytes verus individuals and 15 Pan troglodytes schweinfurthii individuals, while new data are generated for eight wild P. t. schweinfurthii individuals from Kibale National Park and Budongo Forest Reserve in Uganda, as well as 15 Pan troglodytes troglodytes individuals from Gabon, Cameroon, Equatorial Guinea, and two unspecified West African locations. The new enrichment factor is then applied to δ¹³C_enamel values reported for 12 fossil hominin taxa to demonstrate the significance of the diet-enamel enrichment factor for paleodietary interpretations and to emphasize the need to refine estimates of this enrichment factor through additional isotopic analyses of the enamel of primate taxa and their corresponding diets.

2. Materials and Methods

2.1. Ngogo chimpanzee dietary inputs

Previously collected data on the feeding behavior of chimpanzees at Ngogo in Kibale National Park and the isotopic values of their foods were used to compute a composite dietary isotopic input value. This value is employed in subsequent ε*_enamel-diet calculations. In this and all succeeding analyses, raw δ¹³C values were standardized to the preindustrial (1750 CE) isotopic value of atmospheric CO₂ (δ¹³C _{atmos-CO2–1750}) of −6.3‰, to correct for changing values through time, and thus allowing for comparisons between modern, historical, and fossil tissues (Francey et al., 1999; Tipple et al., 2010; Graven et al., 2017; Keeling et al., 2017). To make this correction, the differences were calculated between each of the δ¹³C values of atmospheric CO₂ (δ¹³C_atmos-CO2), either at the time of plant collection or at the time of enamel mineralization, and the δ¹³C _{atmos-CO2–1750} value of −6.3‰. These differences were then subtracted from the raw tissue values to normalize all values to 1750 CE. (Tables 1–4, and Supplementary Online Material [SOM] Tables S1–S4).

Table 1.

Ngogo plant part isotopic contributions to composite δ¹³C dietary input value

Plant part and n	Ranka	Percent feeding timea	Feeding time normalized to 100%b	Mean δ13Cplant-2009/2010 valuesc (‰)	Standard deviationd	Mean δ13Cplant-1750 values (‰)	δ13Cplant-1750 proportional contributione–f
Ficus mucuso fig (n = 9)	1	17.97	20.12	−26.94	± 0.3	−24.94	−5.02
Uvariopsis congensis fruit (n = 5)	2	9.98	11.18	−27.37	± 0.6	−25.37	−2.84
Pterygota mildbraedii leaves (n = 9)	3	8.49	9.51	−30.88	± 1.0	−28.88	−2.75
Pseudospondias microcarpa fruit (n = 4)	4	4.90	5.49	−27.28	± 0.6	−25.28	−1.39
Cordia millenii fruit (n = 3)	5	4.82	5.40	−26.43	± 1.5	−24.43	−1.32
Monodora myristica fruit (n = 3)	6	4.60	5.15	−27.14	± 1.9	−25.14	−1.30
Pterygota mildbraedii seeds (n = 6)	7	3.57	4.00	−27.47	± 1.3	−25.47	−1.02
Morus lactea fruit (n = 5)	8	3.26	3.65	−25.42	± 0.5	−23.42	−0.85
Ficus exasperata leaves (n = 12)	9	3.19	3.57	−27.96	± 1.7	−25.96	−0.93
Celtis africana leaves (n = 9)	10	3.17	3.56	−28.51	± 1.6	−26.51	−0.94
Aningeria altissima fruit (n = 5)	11	2.96	3.32	−25.70	± 1.5	−23.70	−0.79
Mimusops bagshawei fruit (n = 6)	12	2.79	3.12	−26.74	± 0.9	−24.74	−0.78
Treculia africana fruit (n = 4)	13	2.62	2.93	−25.97	± 1.4	−23.97	−0.70
Ficus dawei fig (n = 3)	14	2.45	2.74	−26.04	± 0.9	−24.04	−0.66
Chrysophyllum albidum fruit (n = 9)	15	2.34	2.62	−27.26	± 1.0	−25.26	−0.66
Ficus natalensis fig (n = 6)	16	2.26	2.53	−26.94	± 0.8	−24.94	−0.63
Celtis durandii fruit (n = 5)	17	1.59	1.78	−26.23	± 0.8	−24.23	−0.43
Ficus brachylepis fig (n = 5)	18	1.55	1.74	−28.30	± 1.6	−26.30	−0.46
Morus lactea flowers (n = 5)	20	0.92	1.03	−24.53	± 0.8	−22.53	−0.23
Celtis mildbraedii leaves (n = 4)	21	0.75	0.84	−30.19	± 0.7	−28.19	−0.24
Afromomum mildbraedii pith (n = 8)	23	0.70	0.78	−29.51	± 1.8	−27.51	−0.22
Acanthus pubescens pith (n = 7)	24	0.61	0.68	−29.39	± 2.3	−27.39	−0.19
Morus lactea leaves (n = 4)	26	0.56	0.63	−30.09	± 0.7	−28.09	−0.18
Chaetacme aristata leaves (n = 5)	29	0.51	0.58	−28.88	± 1.8	−26.88	−0.15
Warburgia ugandensis fruit (n = 6)	30	0.45	0.50	−24.62	± 2.2	−22.62	−0.11
Cyperus papyrus pith (n = 5)	31	0.42	0.47	−10.63	± 0.5	−8.63	−0.04
Ficus exasperata fig (n = 7)	34	0.38	0.43	−26.28	± 1.4	−24.28	−0.10
Pterygota mildbraedii flowers (n = 10)	35	0.38	0.43	−24.39	± 1.0	−22.39	−0.10
Neoboutonia macrocalyx roots (n = 5)	36	0.35	0.39	−26.14	± 1.3	−24.14	−0.09
Piper capense pith (n = 9)	43	0.24	0.27	−31.34	± 1.4	−29.34	−0.08
Pterygota mildbraedii cambium (n = 5)	44	0.24	0.27	−25.78	± 1.0	−23.78	−0.06
Cordia millenii flowers (n = 2)	50	0.15	0.17	−25.83	± 1.1	−23.83	−0.04
Marantachloa sp. pith (n = 5)	60	0.10	0.11	−32.25	± 1.7	−30.25	−0.03
Totals		89.26	100.00				−25.33 ± 0.24e

Abbreviations: δ¹³C_{plant-2009/2010} = raw δ¹³C values of plant parts sampled during 2009/2010; δ¹³C_plant-1750 = δ¹³C plant part values adjusted to their 1750 equivalents.

^aData from Watts et al. (2012).

^bFeeding times were normalized to 100% given that only 89.3% of the total diet is isotopically characterized. For example, for Ficus mucuso, the adjusted feeding time was 20.1% ([18.0% / 89.3] × 100).

^cData from Carlson and Kingston (2014).

^dIndividual plant values used to calculate standard deviation presented in SOM Table S1.

^eTo calculate δ¹³C₁₇₅₀ input contribution, the mean δ¹³C_plant-1750 value was multiplied by the feeding time normalized to 100%. For example, for Ficus mucuso, the contribution is −5.019067089 ([20.1 / 100] × 24.9‰).

^fComposite diet input was calculated using Equation 2 and the standard deviation is calculated in SOM Table S1.

Table 4.

Single sample δ¹³C values from all new Pan troglodytes troglodytes individuals sampled in this study.

Specimen ID	Pan group location, country	Sex	Year of collection	Age at death	Average δ13Catmos-CO2 at M3 formationa (‰)	δ13Cenamel (‰)	δ13Cenamel-1750 (‰)	δ13Cdiet-1750 with ε* = 11.8 (‰)	Tooth
MCZ 6244	West Africa	f	1879	Adult	−6.44b	−15.43	−15.29	−26.8	M3
MCZ 46416	West coast of Africa	m	1942	Adult	−6.75	−14.96	−14.52	−26.0	M3
MCZ 9493	Cape Palmas, Gabon	f	1847	Adult	−6.46	−16.47	−16.32	−27.8	M3
MCZ 9494	Gabon River, Gabon	f	1847	Adult	−6.46	−15.85	−15.70	−27.2	M3
MCZ 15312	Lolodorof, Cameroon	f	1913	Adult	−6.60	−14.67	−14.38	−25.9	M3
MCZ 17702	Metet, Cameroon	f	1915	Adult	−6.56	−14.23	−13.98	−25.5	M3
MCZ 17685	Metet, Cameroon	f	1917	Subadultc	−6.65	−14.64	−14.29	−25.8	M2
MCZ 19189	Sakbayeme, Cameroon	f	1925	Subadultc	−6.76	−13.82	−13.37	−24.9	M2
MCZ 23167	Sakbayeme, Cameroon	f	1925	Adult	−6.68	−15.42	−15.05	−26.5	M3
MCZ 26847	Cameroon	f	1930	Adult	−6.66	−15.60	−15.24	−26.7	M3
MCZ 26849	Sakbayeme, Cameroon	f	1930	Adult	−6.66	−15.04	−14.68	−26.2	M3
MCZ 20041	Sakbayeme, Cameroon	m	1921	Adult	−6.61	−14.85	−14.54	−26.0	M3
MCZ 23163	Sakbayeme, Cameroon	m	1924	Adult	−6.63	−15.51	−15.19	−26.7	M3
MCZ 23164	Sakbayeme, Cameroon	m	1924	Adult	−6.63	−14.91	−14.59	−26.1	M3
MCZ 48686	Great Forest of Ayamiken, Eq. Guinea	m	1957	Adult	−6.78	−15.58	−15.11	−26.6	M3

Abbreviations: m = male; f = female; δ¹³C_atmos-CO2 = δ¹³C of atmospheric CO₂; δ¹³C_enamel = raw δ¹³C values from sampled enamel; δ¹³C_enamel-1750 = δ¹³C values from sampled enamel adjusted to their 1750 equivalents; δ¹³C_diet-1750 = δ¹³C value of the composite diet adjusted to its 1750 equivalent; ε* = enrichment factor.

^aKeeling et al. (2017).

^bAll values from individuals in Table 4 are from single drilled samples of enamel taken from just under the occlusal-most surface, as in Figure 2E.

^cAs these two individuals were subadults, there were no M3s available to be sampled, so M₂ values are given instead. Values from these two individuals are indicated in Figure 4 as open circles since they do not represent the postweaning diet.

The feeding behavior of chimpanzees at Ngogo was documented from 1995 to 2010 (Watts et al., 2012). As part of that study, feeding times on all dietary items were collected. Fifteen years of data collection has the effect of attenuating potential biases (e.g., fruit masting events, seasonal fluctuations in precipitation, and resource availability) and furnishes a representative average dietary input for the Ngogo chimpanzees. Plant parts consumed by the Ngogo chimpanzees were collected in both the rainy and the dry seasons of 2009 and 2010 and their isotopic values (δ¹³C_{plant-2009/2010}) were subsequently analyzed (Carlson and Kingston, 2014). These isotopic signatures were then used to calculate a composite dietary input value using a mixing model based on relative feeding time and δ¹³C_plant values (Table 1, SOM Table S1). In this model, it is assumed that percentage feeding time provides a rough estimate of the amount each item contributed to the total isotopic intake. As the analyzed plant parts comprised 89.3% of feeding time, the data were normalized to 100%. The composite chimpanzee isotopic dietary input was calculated using the data presented in Table 1 as follows:

$$\text{Total}{\mathrm{\delta }}^{13}{\text{C}}_{1750}\text{dietary}\text{input}=\Sigma {\text{n}}_{1\dots }{\text{n}}_{33}$$ (Equation 2)

where n_x = δ¹³C₁₇₅₀ input contribution = (adjusted % of feeding time / 100) × (mean δ¹³C_plant-1750), adjusted % of feeding time = (% of feeding time / 89.3) × 100, and mean δ¹³C_plant-1750 = mean plant δ¹³C_{plant-2009/2010} + 2.0‰.

The 2.0‰ adjustments to the mean δ¹³C_plant-1750 values were made due to the difference between the δ¹³C_atmos-CO2 in 2009/2010 (−8.3‰) when the plants were collected and that in 1750 (−6.3‰; Keeling et al., 2017). The preceding calculations can be illustrated with Ficus mucuso, the most frequently consumed item in the Ngogo chimpanzee diet (Table 1). With a mean δ¹³C_plant-1750 value of −24.9 ± 0.3‰, this fig represented 18% of chimpanzee feeding time as reported by Watts et al. (2012). This feeding time was normalized to 100% given that 89.3% of the total diet is isotopically characterized, resulting in an adjusted feeding time of 20.1% ([18.0% / 89.3] × 100). The δ¹³C₁₇₅₀ input contribution of this species is then −5.0049 ([20.1 / 100] × 24.9‰).

To calculate the cumulative uncertainty (standard deviation) of the composite isotopic dietary input value for the Ngogo chimpanzees, the variance of each dietary item was multiplied by the square root of the feeding time for that item (Table 1, SOM Table S1). The resulting values for each of the dietary items were summed, and the square root of the total provided the standard deviation for the composite dietary isotopic input value.

2.2. Sampling enamel from chimpanzee teeth

Three chimpanzee data sets, including the enamel from 23 individuals, were generated in this study. Each data set was produced using a different sampling procedure that provides complementary insights into isotopic variability and sampling strategies. The sexes of the specimens and details related to the collection sites and year(s) of collection, where known, are provided in Tables 2–4. Each procedure yielded enamel powder that was pretreated with 5.25% NaOCl for four hours, rinsed three times with Millipore purified water, soaked in 0.1M CH₃COOH for 4 hours, rinsed 3 more times with purified water, and then freeze dried. The samples were analyzed isotopically on a Finnigan-MAT 252 isotope ratio mass spectrometer with a Kiel III carbonate preparation device at the University of Florida Light Stable Isotope Laboratory. Results are reported using the standard ‰ notation where δ¹³C = (R_sample / R_standard - 1) × 1000 and R_sample = (¹³C / ¹²C)_sample and R_standard = (¹³C / ¹²C)_standard. Carbon values are reported relative to PDB (Pee Dee Belemnite). Precision was ± 0.08‰ for δ¹³C ratios for 14 replicate pairs of fossil enamel. International and in-house laboratory standards analyzed with the enamel samples yielded a standard deviation of ± 0.06‰ (n = 36).

Table 2.

δ¹³C microsampled and bulk valúes from new Pan troglodytes schweinfurthii individuals sampled in this study.

Pan ID	Pan group, location	Sex	Name	Year/age at death	Average year at M3 formation	Average δ13Catmos-CO2 during M3 formationa (‰)	Tooth/sample #b	δ13Cenamel (‰) all values	δ13Cenamel-1750 (‰) cervical value	δ13Cdiet-1750 with ε* = 11.8 (‰)
NG003	Ngogo, Kibale National Park, Uganda	m	Grappelli	2002/20	1989	−7.65	M3 #1	−15.86
							M3 #2	−15.35	−14.00	−25.5
							M3 bulkc	−15.61
NG004	Ngogo, Kibale National Park, Uganda	m	Stravinsky	2006/32	1981	−7.42	M3 #1	−15.24
							M3 #2	−14.70	−13.58	−25.1
							M3 bulk	−14.97
NG005	Ngogo, Kibale National Park, Uganda	m	Tatum	2009/23	1991	−7.67	M3 #1	−15.35
							M3 #2	−15.66
							M3 #3	−15.27	−13.90	−25.4
							M3 bulk	−15.43
NG013	Ngogo, Kibale National Park, Uganda	m	Webster	2014/26	1993	−7.80	M3 #1	−14.84
							M3 #2	−15.13	−13.63	−25.1
							M3 bulk	−14.91
NG012	Ngogo, Kibale National Park, Uganda	f	Carmen	2013/51	1969	−7.01	M3 #1	−15.27
							M3 #2	−15.71	−15.00	−26.5
							M3 bulk	−15.49
NG001	Kibu, Kibale National Park, Uganda	m	N/A	2002/Adult	1984	−7.57	M3 #1	−15.45
							M3 #2	−15.70	−14.44	−25.9
							M3 bulk	−15.57
NG002	Wantabu, Kibale National Park, Uganda	m	N/A	2002/Adult	1984	−7.57	M3 #1	−14.53
							M3 #2	−14.87	−13.61	−25.1
							M3 bulk	−14.70
MUZM2625	Budongo Forest Reserve, Uganda	f	N/A	1995/∼20	1982	7.58d	M3 #1	−15.30
							M3 #2	−15.14	−13.86	−25.4
							M3 bulk	−15.22

Abbreviations: m = male; f = female; δ¹³C_atmos-CO2 = δ¹³C of atmospheric CO₂; δ¹³C_enamel = raw δ¹³C values from sampled enamel; δ¹³C_enamel-1750 = δ¹³C values from sampled enamel adjusted to their 1750 equivalents; δ¹³C_diet-1750 = δ¹³C value of the composite diet adjusted to its 1750 equivalent; ε* = enrichment factor.

^aKeeling et al. (2017).

^b‘sample #’ refers to each sample’s developmental order within that tooth, with sample #1 taken more occlusally and #2 or #3 representing the most cervical sample, depending upon the tooth in question (see Fig. 3).

^cBulk is the average of the 2 or 3 samples from each tooth.

^dThe δ¹³C_atmos-CO2 during M₃ formation had to be estimated for this individual based on her year of death/collection of 1997. Her age at death was judged to be ∼20 years, similar to that of the youngest Ngogo male (NG003), based on tooth wear, which places her M₃ formation between 1979–1984, giving her an average δ¹³C _{atmos CO2} during M₃ formation of −7.58. Her more detailed microsampling data are in Table 3.

The first data set consisted of samples extracted from a mix of upper and lower M3s of eight P. t. schweinfurthii individuals, five from Ngogo, two from groups neighboring Ngogo in Kibale National Park, and one from the Budongo Forest Reserve, Uganda (Table 2). The five chimpanzees from Ngogo were well studied during their lifetimes, and much is known about their life history and foraging behavior. The other two males were victims of intergroup encounters with the Ngogo chimpanzees and consequently there are no associated data. All of these Kibale chimpanzees were buried by researchers at Ngogo, subsequently exhumed, and then accessioned at Makerere University’s Zoology Museum (MUZM) in Uganda. The seven Kibale chimpanzees are all accessioned and stored at the MUZM with the NG (Ngogo) prefix to distinguish this collection. The Budongo chimpanzee is also accessioned and stored at the MUZM. Using a Brassler Z500 drill with a carbide-tungsten drill bit (Brassler H23R.11.010), 2–3 microsamples were taken from each M³ or M₃. The δ¹³C value from the latest forming enamel, taken from the cervical end of each crown, was used to generate the enrichment factor (see Table 2 and SOM Table S2). Utilizing the δ¹³C value of the cervical end samples instead of the bulk signals from the M3 minimized any dietary contribution of a nursing signal. These later forming samples each span at least 1–2 years of enamel formation and maturation time, thereby representing at least some interannual dietary variation.

The time in calendar years of M3 mineralization was determined based on the known or approximated year at birth and the year of death (Wood et al., 2017), assuming that M3 crown mineralization and maturation occurred when the chimpanzees were 4–9 years old (Reid et al., 1998). The ages of death for the extragroup males from Kibale (NG 001, NG 002) and the Budongo female (MUZM 2625) were estimated by comparing the level of tooth wear from these individuals with tooth wear from the Ngogo individuals whose approximate ages at death were known. The chimpanzees’ raw δ¹³C_enamel values were adjusted to the 1750 equivalent using the years their M3s were forming, or estimated to be forming, rather than the years they died.

In a second sampling protocol, enamel was sampled using a more invasive, but analytically more informative, procedure from only the Budongo P. t. schweinfurthii skull (MUZM 2625). The crowns of this individual were particularly unworn and well preserved, providing an opportunity to develop a sampling approach to assess inter- and intratooth variation and potential biases introduced by different sampling techniques. A 2 mm wide segment of enamel, about 0.6–0.8 cm long representing the full length of a crown from the occlusal to cervical end, was removed using a Brassler Z500 drill with a diamond disc (11-HP-945B) from the mesiolingual cusps of the P³, P⁴, M¹, M², and M³ (Figs. 1 and 2).

Figure 1.

Intratooth sampling protocol used to resample the teeth of the Budongo chimpanzee MUZM 2625. A). Antimeres of the sampled teeth (left P³–M³), whose mesiolingual (ml) cusp lengths were measured to reconstruct the lengths of the tooth cusps that were actually sampled. In the case of the P³, the mesiobuccal cusp was sampled instead of the mesiolingual cusp. B) Left M¹ depicted in A, showing the different approximate lengths and widths of the individual samples from the length of that crown. C) Actual sampled right M¹ to demonstrate the depth of the sample sawed from this tooth crown. The depth of each sample is used in conjunction with the lengths and widths from B to calculate approximate volumes for each enamel sample once it was reduced to powder.

Figure 2.

Parts and tissues of a tooth as seen in the M³ of MUZM 2625 as well as representations of the three types of sampling described in the Materials and Methods, with the δ¹³C_enamel-1750 values (from the MUZM 2625 M³) indicated on each sampled area, and the average ‘bulk’ values that result. A) Descriptions of the abbreviated features: En = enamel; De = dentine; Ce = cementum; P = pulp cavity; CEn = cuspal enamel; IEn = imbricational enamel; red arrow = direction of enamel growth from the occlusal to cervical end of the crown. B) View of the periodic enamel structures from the M³, in which the orientation and distance between striae of Retzius change (seen in blue) moving from the more occlusal end to the cervical end of the crown. These differences reflect variable mineralization rates. Red arrow = direction of enamel growth from the enamel-dentine junction (EDJ) to outer enamel surface (OES). C) two samples are drilled, one closer to the occlusal surface (depending upon how worn the tooth in question is) and one closer to the cervical end, which was the protocol for sampling the Ngogo chimpanzees (the asterisk indicates where a third sample was taken in several other teeth, but only two were taken from this tooth, so that value is blank here). D) Image of the M³ in which all of the enamel along a ca. 2 mm wide section of the tooth, representing the full length of a crown from the occlusal to cervical ends, was removed and sampled; this was the protocol for the Budongo chimpanzee MUZM 2625. E) Image depicting the method in which only one sample is extracted by drilling near the occlusal surface, which was the protocol for sampling all specimens from the MCZ (Museum of Comparative Zoology, Harvard University).

Any dentine adhering to the inner surface of the enamel segment was gently abraded away using a Brassler carbide-tungsten drill bit (HP-901–018), and the length of enamel was cut transversely into 4–6 fragments, each of which was crushed to form 1–2 mg of powder. In order to avoid partial breaks or splintering of the enamel segment, transverse cuts were first scored, using a 0.22 mm Fisherbrand carbon steel razor blade. A pair of Excel Blades stainless steel-edged wire cutters were used to separate the samples from the enamel segment. This technique maximized the number of samples that could be generated mechanically and yield sufficient powder for isotopic analyses. With increasing samples per tooth, there is a greater likelihood of characterizing intratooth variation (Table 3; Fig. 3). Sampling all five of these teeth from one individual also presents an opportunity to evaluate intertooth variability (Fig. 3) in order to assess whether a postweaning dietary signal could be obtained from later-forming areas of teeth other than the M3 in chimpanzees.

Table 3.

Microsampled and bulk δ¹³C values from the Budongo chimpanzee (MUZM 2625).

Tooth/sample #a	δ13Cenamel (‰)	δ13Cenamel-1750 (‰)b	δ13Cdiet-1750 with ε* = 11.8 (‰)
M1 #1	−16.01	−14.73	−26.22
M1 #2	−16.28	−15.00	−26.49
M1 #3	−16.60	−15.32	−26.80
M1 #4	−16.50	−15.22	−26.70
M1 #5	−16.65	−15.37	−26.85
M1 #6	−16.10	−14.82	−26.31
M1 bulkc	−16.35	−15.07	−26.56
M2 #1	−15.97	−14.69	−26.18
M2 #2	−16.26	−14.98	−26.47
M2 #3	−15.58	−14.30	−25.80
M2 #4	−15.81	−14.53	−26.02
M2 bulk	−15.90	−14.62	−26.11
P3 #1	N/Ad	N/A	N/A
P3 #2	N/A	N/A	N/A
P3 #3	N/A	N/A	N/A
P3 #4	−16.13	−14.85	−26.34
P3 #5	−15.62	−14.34	−25.84
P3 #6	−16.20	−14.92	−26.41
P3 bulk	−15.98	−14.70	−26.19
P4 #1	N/A	N/A	N/A
P4 #2	−15.59	−14.31	−25.81
P4 #3	−16.19	−14.91	−26.40
P4 #4	N/A	N/A	N/A
P4 #5	−15.67	−14.39	−25.88
P4 #6	N/A	N/A	N/A
P4 bulk	−15.82	−14.54	−26.03
M3 #1	−15.26	−13.98	−25.48
M3 #2	−15.40	−14.12	−25.62
M3 #3	−14.86	−13.58	−25.08
M3 #4	−14.56	−13.28	−24.79
M3 #5	−14.43	−13.15	−24.66
M3 bulk	−14.90	−13.62	−25.12

Abbreviations: δ¹³C_atmos-CO2 = δ¹³C of atmospheric CO₂; δ¹³C_enamel = raw δ¹³C values from sampled enamel; δ¹³C_enamel-1750 = raw δ¹³C values from sampled enamel adjusted to their 1750 equivalents; δ¹³C_diet-1750 = δ¹³C value of the composite diet adjusted to its 1750 equivalent; ε* = enrichment factor.

^a‘sample #’ refers to each sample’s developmental order within that tooth, with sample #1 taken more occlusally and #4, 5, or 6 representing the most cervical sample, depending upon the tooth in question (see Fig. 3).

^bThe average δ¹³C value of atmospheric CO₂ during years of mineralization was −7.58, and the difference between this value and −6.3, or the δ¹³C value of atmospheric CO₂ in 1750 CE, is −1.28 for all samples

^cBulk values are averages of the 3–6 values from each tooth. Individual values are also depicted in Figure 2D.

^dN/A = samples that did not produce enough CO₂ to obtain δ¹³C values

Figure 3.

Inter- and intratooth δ¹³C_enamel-1750 values for P³, P⁴, M¹, M², and M³ from a single chimpanzee (MUZM 2625). Intratooth variation in individual δ¹³C_enamel-1750 values is shown as colored circles connected by solid-colored lines. Three samples from the P³ and three from the P⁴ are not depicted here because they did not produce sufficient CO₂ for isotopic analyses. Horizontal grey dashed lines extending from each δ¹³C_enamel-1750 value represent a model of overlapping approximate age ranges for successive samples, given their location within the crowns (e.g., Fig. 2D), but not including maturation time (see text for more complete description). All 1750-adjusted values are listed in Table 3.

Finally, a third data set and sampling strategy included the M3 enamel of 15 West African P. t. troglodytes individuals from collections at the Museum of Comparative Zoology (MCZ) at Harvard University. For these individuals, single samples were drilled laterally perpendicular to the buccal surface directly under the occlusal-most area of the M3 crowns (Fig. 2E). The goal was to maximize time represented by drilling through the cuspal enamel and minimize any damage to diagnostic features. Imbricational enamel is thin in some areas and presents the risk of incorporating dentine when sampling by drilling. Specimens in this third data set are part of historical MCZ collections accessioned between 1847 and 1957 and all δ¹³C_enamel signatures were normalized to a 1750 value (Table 4). To do this, the year of collection was assumed to be the year of death. The ages of adult chimpanzees in the MCZ data set are unknown. To estimate them, an expected range of ages for these individuals was constructed using data from studies of wild chimpanzees. The average age at first reproduction was used to define the start of adulthood and the life expectancy at birth to define its end. Because there are sex differences in both of these age ranges, separate ranges for females and males were calculated. The range of ages for adult females was 16.21–37.3 years (Wood et al. 2017; Walker et al. 2018). The range of ages for adult males was 17.74–29.0 years (Wood et al. 2017; Ngogo Chimpanzee Project, unpublished data). To estimate the years that the M3s of the MCZ individuals formed, six years were subtracted from the preceding maxima and minima of the ranges, since age six is the midpoint in the age range during which the M3 mineralizes in chimpanzees (Reid et al., 1998). The resulting differences were then subtracted from collection years. The average δ¹³C_{atmos CO2} value of those from the two dates derived from this calculation (using the maximum and minimum) was then subtracted from the δ¹³C_{atmos CO2–1750} value of −6.3‰, and the difference was added to the δ¹³C_enamel value to normalize it to 1750. For example, the possible range of M3 mineralization years for an adult female specimen (MCZ 15312) with a δ¹³C_enamel signature of −14.7‰ and collected in 1913 was estimated to be between 1875 (=1913 – 37.3 yr) and 1897 (1913 – 16.21 yr). The average of the δ¹³C_{atmos CO2} values for these years was −6.6‰. Therefore, 0.3‰ (−6.3‰ - −6.6‰) was added to the δ¹³C_enamel value of the M3 to normalize it to 1750, resulting in a δ¹³C_enamel-1750 value of −14.4‰ (Table 4). Two specimens (MCZ 17685, MCZ 19189) were from females who had not yet reached adulthood based on lack of M3 eruption. Their long bone epiphyses, however, had fused, and they we near adult size. We used 12.5 years as the estimated age for these two samples based on the reported age at which most long bone epiphyses fuse and M3s emerge in female chimpanzees in the wild (Zihlman et al., 2007).

The ε*_enamel-diet standard deviation was calculated in R using a Monte-Carlo simulation following the approach of Tsutaya and colleagues (2017).

3. Results

3.1. Ngogo chimpanzee dietary inputs

Data for 33 dietary items consumed by the Ngogo chimpanzees, accounting for 89.3% of their total feeding time, are shown in Table 1. 1750-adjusted C₃ plant item δ¹³C values range from −30.3‰ (Marantachloa sp. pith) to −22.4‰ (Pterygota mildbraedii flowers). These values are consistent with those derived from C₃ vegetation in closed canopy forest as well as values from plants that may be water-stressed or subject to high irradiance in forest gaps. The pith of Cyperus papyrus, the only dietary plant that utilizes the C₄ photosynthetic pathway, yielded a value of −8.6‰. Cyperus papyrus is ranked 26^th in the list of 33 plants and represented only 0.5% feeding time in the Ngogo chimpanzee diet. Calculation of the relative isotopic contributions of the 33 dietary items resulted in a cumulative 1750-adjusted dietary input value (δ¹³C_diet-1750) of −25.3 ± 0.2‰.

3.2. Inter- and intratooth variation

Detailed sampling of the teeth of one chimpanzee from the Budongo Forest Reserve (MUZM 2625; Figs. 1 and 2D) provides insight into the isotopic variation inherent in the enamel of an individual chimpanzee.

The δ¹³C_enamel-1750 values of 3–6 serial samples from the P³ to M³ of this specimen (Table 3) are depicted in Figure 3 along with the approximate timing of mineralization for each of these tooth crowns based on previous studies of crown formation in chimpanzees (Reid et al., 1998). Isotopic values of enamel from different teeth known to form at the same time (M², P³, and P⁴) have similar values, as would be expected given the developmental overlap. Overall, there is a 2.2‰ increase in intertooth δ¹³C_enamel-1750 values through time, with the most ¹³C depleted value from early forming enamel in the M¹ (−15.4‰) and the most ¹³C enriched value from the latest forming enamel in the M³ (−13.2‰). The overall pattern suggests a systematic increase in the δ¹³C_enamel-1750 values as the individual starts to consume the average adult diet. However, there are several deviations from this general trend, including the initially more positive values in earliest-forming M¹ enamel, and the nearly simultaneous decrease in values occurring in the M², P³, and P⁴ around age 3. Intertooth variation, at the level of bulk tooth analyses, indicates that the earliest forming enamel of the five teeth sampled, the M¹, has a bulk value of −15.1‰. Bulk values of the M², P³, and P⁴, mineralizing at roughly the same time as one another, are −14.6‰, −14.7‰, and −14.5‰ respectively. There is a 1.0‰ increase in bulk value from the M² to M³ (−13.6‰), double the increase from M¹ to M², possibly reflecting a more significant shift from a weaning to a postweaning/adult diet or dietary proportions.

There was also considerable intratooth variation, with a maximum 0.9‰ difference between enamel at the occlusal (−14.1‰) and cervical (−13.2‰) ends of the M³ crown. These results reinforce the need to consider both intra- and intertooth variation when sampling teeth to obtain a postweaning/adult dietary signal. In this case, for example, sampling the M¹ yields a bulk isotopic value of −15.1‰, 1.5‰ more negative than the bulk M³ dietary signature of −13.6‰. The nearly identical values from the later forming enamel of the M², P³, and P⁴ are still 0.2‰ lower than the lowest M³ value (−14.1‰), as much as 0.7‰ lower than the calculated bulk M³ value (−13.6‰), and as much as 1.1‰ lower than the final, most enriched M3 value (−13.2‰). Taken together, this precludes later forming enamel from the M², P³, or P⁴ being considered equivalent to the latest forming M³ enamel values, assumed here to represent the postweaning/adult diet. Collectively, these results highlight the need to sample strategically and consistently to maximize the probability of capturing the postweaning/adult dietary signal.

3.2. Establishing an enrichment factor for the Ngogo chimpanzees

The δ¹³C_enamel values of four males (NG 003–005, and NG 013; Table 2) were used to calculate an enrichment factor for the Ngogo chimpanzees. Because male chimpanzees are philopatric (Nishida and Kawanaka, 1972), these individuals spent their entire lives at Ngogo, and it can be assumed that their diet consisted of food items, approximately in the proportions listed in Table 1, while their M3s were forming. Females, who typically emigrate, and extragroup males were excluded from this calculation, as their diets during M3 development are unknown. This conservative approach is warranted given, for example, that the most consumed food at Ngogo, fruit of Ficus mucuso, is rare and thus infrequently consumed by the Kanyawara chimpanzees only 10 km away (Potts et al., 2011). The feeding behavior data in Table 1 are biased toward males (Watts et al., 2012), increasing their applicability to determine an enrichment factor using enamel samples from natal males. The four Ngogo males had a relatively limited δ¹³C_enamel-1750 range from −13.6‰ to −14.4‰ (Table 2) with an average value of −13.8 ± 0.2‰. Using Equation 1, the composite δ¹³C_diet-1750 value (−25.3 ± 0.2‰) was combined with the average δ¹³C_enamel-1750 value of the Ngogo male chimpanzees to yield a mean ε*_enamel-diet of 11.8 ± 0.3‰ (see Tables 1–2 and SOM Tables S1–S2). The relatively low ε*_enamel-diet standard deviation of ± 0.3‰ reflects limited isotopic variability in major contributing dietary items (see Table 1) as well as minimal isotopic differences between the four natal male chimpanzees (SOM Table S2). In other studies that report ε*_enamel-diet, standard deviations range from ± 0.8‰ to ± 2.2‰ for a variety of wild and captive mammals from Africa and South America (Table 1, Cerling and Harris, 1999), ± 0.6‰ for primates in the Ituri Forest, DRC (Cerling et al., 2004), and ± 1.1‰ for giant pandas in Foping National Nature Reserve, China (Han et al., 2016). Controlled feeding studies have yielded ε*_enamel-diet standard deviations of ± 0.3‰ to ± 0.7‰ (Passey et al., 2005) and ± 0.68‰ to ± 1.03‰ (Tejada-Lara et al, 2018). Even though chimpanzees have a more complex and varied diet than almost all the taxa described in these studies, the standard deviation for the ε*_enamel-diet of these chimpanzees is lower than essentially all known taxa, possibly suggesting that much of the variability inherent in the dietary ecology of Ngogo chimpanzees may have been accounted for in this study.

3.3. Reconstructed diets of other chimpanzees using the Ngogo enrichment factor

The ε*_enamel-diet of 11.8 ± 0.3‰ calculated from the four Ngogo male chimpanzees was used to reconstruct the isotopic dietary signals of chimpanzees sampled from different subspecies, sites, and groups. Figure 4 depicts the δ¹³C_diet1750 values from the 23 chimpanzees sampled in this study along with isotopic signatures of 44 P. t. verus individuals and 15 P. t. schweinfurthii individuals analyzed in previous studies (see also SOM Table S1). Values from all the new P. t. schweinfurthii and P. t. troglodytes individuals and 30 of the P. t. verus individuals were from M3s. The 15 previously published P. t. schweinfurthii values and 14 of the P. t. verus values taken from the literature are derived from M3s and other sources. The latter include: (1) molars other than M3s; and (2) combinations of δ¹³C_enamel values from multiple teeth. As discussed above, values derived from non-M3 samples (open circles) are unlikely to reflect the diet after weaning.

Figure 4.

δ¹³C_diet1750 values for Pan troglodytes subspecies from regions and groups in equatorial Africa. Mean annual precipitation (MAP) estimates (summarized in Loudon et al., 2016; Schoeninger et al., 2016) indicate that, for the sites represented here, there is a non-uniform gradual decrease in MAP moving eastward from the Atlantic coast towards the rift-related uplift of East Africa. M³ values (sampling the postweaning/adult diet) are indicated by filled circles, while data from other teeth (sampling preweaning dietary signals) are indicated by open circles. * = new data from this study; ** = previously published data.

Figure 4 reveals a range of δ¹³C_diet values that differ by almost 6‰ across the entire chimpanzee sample. Although values vary within communities and sampling locales, they are relatively low in West African chimpanzees, Pan troglodytes verus. Isotopic signatures start to increase as one moves eastward, with higher values displayed by the two other subspecies, P. t. troglodytes and P. t. schweinfurthii. In general, the changing δ¹³C_enamel values displayed across the geographic distribution of chimpanzees vary with corresponding changes in rainfall and habitat (Fig. 4).

3.4. Application of the new enrichment factor to fossil hominins

Figure 5 depicts how the application of the 11.8‰ chimpanzee-specific enrichment factor, and exclusion of juvenile/infant dietary signals, affects the distribution of δ¹³C _enamel signals for fossil hominin taxa. The difference between the 11.8‰ chimpanzee-derived ε and the 14.1‰ generalized large mammalian herbivore ε is 2.3‰, hence all hominin values are shifted to be 2.3‰ more positive. The resulting reconstructed diets are more ¹³C-enriched than previously assumed.

Figure 5.

δ¹³C_diet1750 values from the teeth of 12 fossil hominin taxa (SOM Table S4) and M³s from Pan (SOM Table S3). Using the corrected enrichment factor and focusing on only those values that likely reflect a postweaning diet (black box-and-whisker plots) results in all fossil hominins occupying, at least in part, the mixed C₃/C₄ diet or exclusive C₄ ranges. The Paranthropus boisei-like specimen is indicated by a star next to the Pa. boisei box-and-whisker plots. Since age at weaning is unknown for extinct hominins, and some aspects of nursing, including the length of time of exclusive nursing, may have resembled humans more than great apes for some australopiths (Joannes-Boyau, 2019), it is important to remain open to the possibility that weaning may have been completed earlier relative to dental development in some extinct hominin species. Consequently, data from premolars and M²s were included for the fossil hominins. MAP = mean annual precipitation; ε* = enrichment factor

The effects of removing unidentified teeth or teeth likely to incorporate a pre-weaning dietary signature (SOM Table S2) warrant additional examination. For seven of the twelve species considered, the median ¹³C_enamel value increased, including Homo sapiens (by 2.8‰), Paranthropus robustus (by 0.3‰), Paranthropus aethiopicus (by 1.8‰), Australopithecus bahrelghazali (by 1.6‰), Kenyanthropus platyops (by 0.2‰), Australopithecus afarensis (by 0.2‰), and Ardipithecus ramidus (by 0.9‰). An increase in the median values accords with studies indicating a general trend of ¹³C enrichment in the postweaning/adult diet relative to the pre-weaning diet (Smith et al., 2010; Malone, 2019; Wright and Schwarcz, 1998). For three of the hominin species (Australopithecus sediba, Paranthropus boisei, Australopithecus anamensis), previous studies largely sampled only teeth formed after weaning, so there was no appreciable change in the median ¹³C_enamel values, as indicated in Figure 5. In the remaining two taxa, Homo spp. and Australopithecus africanus, there was an approximately 1‰ decrease in the ¹³C_enamel signatures when samples are restricted to teeth formed after weaning. It is important to acknowledge that samples for these taxa (and others) span significant intervals of time and space, generating variability that may effectively obscure any trends related to consideration of adult-forming enamel only. A. africanus specimens reported here, for example, derive from two different sites dated to different intervals of time (Sponheimer and Lee-Thorp, 1999; Lee-Thorp et al., 2000; van der Merwe et al., 2003; Sponheimer et al, 2005, 2013), and Homo spp. potentially represents different species from different sites (Cerling et al, 2013; Ludecke et al., 2018). Finally, an additional consequence of removing teeth with non-adult or preweaning dietary signals is a decrease in variability for several species. While a decreased range typically relates to elimination of more ¹³C depleted signatures associated with preweaning/juvenile diets, it is also apparent that more positive ¹³C_enamel values were also removed from consideration. The latter situation almost always involved removal of unidentified teeth (SOM Table S4).

4. Discussion

Isotopic analyses of the enamel of four adult males from the Ngogo chimpanzee community, integrated with a composite δ¹³C value of their diet, provides the first reliable estimate of an enrichment factor for any hominoid primate. This finding emerges from long-term study of the dietary ecology of wild chimpanzees, providing a template to assess the enrichment factor of a long-lived species. The enrichment factor calculated here includes information regarding dietary input, digestive physiology, and enamel formation. Because these factors vary among individuals and between and within populations, it is important to minimize, as much as possible, the sources generating variability. A critical step involves sampling and analyzing the enamel of individuals whose diet is known. In addition, it is necessary to sample enamel forming in isotopic equilibrium with the adult diet because adult dentition in chimpanzees represents an archive of dietary input spanning the weaning process and establishment of an adult diet. In chimpanzees, only the M3s reflect the relevant adult composite δ¹³C dietary signature.

4.1. Extant chimpanzee comparisons

The combined δ¹³C_diet1750 values for all of the chimpanzee samples analyzed in this paper and in previously published studies reveal a gradient that increases moving from the west coast of equatorial Africa to East Africa (Fig. 4). This gradient mirrors a general trend of decreasing mean annual precipitation (MAP; Nicholson, 2017) and a shift from lowland, closed canopy to more seasonal/montane forests across Africa. Humid air masses from the Atlantic driven by the West African monsoon system result in MAP of 1900–2600 mm/yr along west coastal regions (Nicholson, 2017). Rainfall decreases moving inland to the east, as the influence of the Atlantic Ocean diminishes and rift-related topography affects air circulation and precipitation, creating seasonal heterogeneity (Feakins et al., 2010). Isotopic dietary differences (Fig. 4) related to proximity to the Atlantic coast also roughly parallel the geographic distribution of chimpanzee subspecies. This suggests that the slight subspecific isotopic gradient seen in Figure 4 may reflect differences in dietary items available to chimpanzees at these sites, as well as different plant isotopic baselines in ecosystems moving from west to east where those subspecies are found.

The reconstructed diets of chimpanzee communities based on the ε*_enamel-diet of 11.8 ± 0.3‰ (Fig. 4) reveal variation within communities, between communities, and between subspecies. Several factors may contribute to this variation, including differences in feeding behavior, female dispersal, seasonality, MAP, elevation, and δ¹³C dietary plant variability due to local and regional effects of canopy architecture. Taken together, the reconstructed diets of modern chimpanzees indicate a species capable of occupying diverse habitats, ranging from closed-canopy forest settings to broken or open-canopy habitats, reflecting their ecological flexibility. The degree of this flexibility, however, does not appear to extend to significant consumption of ¹³C enriched savannah resources, seemingly exploited by early hominins. While chimpanzees remain an appropriate modern analog to estimate an enrichment factor for early hominins from a phylogenetic perspective, the preceding considerations indicate that using them for this purpose requires caution.

4.2. Inter- and intratooth variation

The findings presented here represent the only comprehensive assessment of both intra- and intertooth δ¹³C_enamel variation in a modern non-human primate. Previous studies have focused on intratooth variation in fossil primates, including A. africanus (Lee-Thorp et al., 2010), Pa. robustus (Sponheimer et al., 2006), Homo heidelbergensis (Garcia et al., 2015) and Indopithecus and Sivaladapis (Patnaik et al., 2014). One isotopic study revealed more than 6‰ intratooth variation in a single Australopithecus africanus M³ (Lee-Thorp et al., 2010). This range of intratooth variation, however, may have been influenced by analytical uncertainty associated with the laser ablation used in that study. Lee-Thorp et al. (2010) note that values generated using laser ablation relative to acid digestion were offset by −1.3 ± 1.5‰.

Primate intertooth variation has been documented for P. t. verus (Smith et al., 2010) and for M1–M3 differences in a mandible of Sivaladapis (Patnaik et al., 2014). The variation revealed in these studies has been hypothesized to reflect differences in pre- vs. postweaning diets, in juvenile, subadult, and adult diets, and/or seasonal/annual dietary fluctuations in foraging strategies. Isotopic analyses of the enamel of chimpanzees from Ganta, Liberia (Smith et al., 2010), for example, provide information on intertooth variation based on a single ‘bulk’ value for both the M¹s and M³s of 30 individuals (14 females, 16 males). ¹³C enrichment of the M³ relative to the M¹ is more common in the females (93%) than males (63%). This suggests sex-based weaning differences, with females consistently weaned and consuming the postweaning/adult diet by the time that their M³ enamel was forming, and roughly half of the males continuing to nurse during M³ formation. This asymmetry may also have implications for interpreting variation documented in early hominin data sets.

The intra- and intertooth δ¹³C_enamel variation documented in this study (Table 3; Fig. 3) suggests the detection of inter-annual dietary changes or weaning behavior-based fluctuations, such as the simultaneous decrease and then increase in δ¹³C_enamel values at the end of M², P³, and P⁴ crown formation (ca. three years old; Reid et al., 1998). However, it is important to recognize that age estimations for each value (gray dashed horizontal lines in Fig. 3) are based upon incremental structures laid down during the initial mineralization phase of enamel formation, and there is evidence that the majority of the final mineral component of enamel is incorporated into the crown during maturation, the second phase of formation (Simmer et al., 2012; Green et al., 2017). Unlike the well-documented incremental structures laid down during the initial mineralization phase (e.g., Boyde, 1989; Smith, 2006; Smith and Tafforeau, 2008), the orientation of the maturation front and the rate of enamel maturation remains unstudied for most taxa, including chimpanzees. Prior research suggests that the innermost and mid-crown regions of enamel mature first after mineralization (Suga, 1982; Tafforeau et al., 2007; Blumenthal et al., 2014), with the outer layers maturing later (Traylor and Kohn, 2017). If this is the case for chimpanzees and other hominoids, then age estimates attributed to the individual values in Figure 3 likely underestimate ages of each sample. To resolve this, additional research, like that of Green et al. (2018) and Traylor and Kohn (2017), needs to be conducted on dental remains of chimpanzees, gorillas, orangutans, and other primates with the aim of characterizing the unknown geometry and pace of enamel maturation in hominoids.

4.3. Recommendations for achieving a bulk vs. postweaning/adult dietary signal

If the goal is to produce a δ¹³C_enamel value representative of the isotopic dietary niche of an organism, while also minimizing the destruction of enamel, it is important to strategically analyze enamel that represents and averages variation in the postweaning/adult diet. In primates, the M3 should be targeted for sampling in any taxon for which it is available. In chimpanzees, despite the fact that the P3, P4, and M2 crowns continue forming beyond the age of 4 years, when the M3 begins to form (Reid et al., 1998), the isotopic values yielded by sampling late forming areas of these teeth are depleted relative to the M3 values which are likely to reflect the postweaning/adult diet (Table 3; Fig. 3). With these considerations in mind, achieving a ‘bulk’ dietary signal, while also accessing the range of variation in the represented in the M3, can be best implemented by averaging the values obtained from a sample drilled close to the occlusal surface and one from close to the cervical end on the M3 (such as the first and third samples in Fig. 2C). These samples are temporally distinct and yet representative of the full range of M3 intratooth δ¹³C_enamel-1750 variation that can be detected using such methods. In fact, if samples #1 and #5 in Figure 2D are averaged, they yield the same bulk value as when all five values are averaged (−13.6‰). In other primates, including fossil hominins, weaning completion may occur relatively early or dental formation may extend over a relatively longer period of time. In these cases, a postweaning/adult signal could conceivably be found in late forming P3, P4, or M2 enamel, so sampling these crowns at their cervical ends, when no M3s are available, could also yield a postweaning/adult dietary signal in such taxa.

As mentioned previously, there is some evidence suggesting that the outermost layers of imbricational enamel may take the longest to reach full maturation (Traylor and Kohn, 2017). If so, sampling from the outer layers of M3 enamel where the crown is thickest could yield dietary signals from even later in development than can be obtained from the cervical enamel. Given this possibility, it may also be possible to access a postweaning/adult dietary signal from late-maturing outer layers of P3, P4, or M2 imbricational enamel if those teeth are all that are available.

4.4. Early hominin dietary reconstructions

The reinterpretations of fossil hominin diets presented here are based upon the use of the new chimpanzee ε*_enamel-diet of 11.8 ± 0.3‰. Future careful isotopic analyses of the diet and enamel δ¹³C of additional hominoids will, no doubt, help refine this enrichment factor. At present, however, it potentially represents the best available value for application to the reconstructions of fossil hominin diets.

Isotopic analysis of the enamel bioapatite of early hominins has revealed ¹³C enriched diets relative to modern chimpanzees, indicating a fundamentally different niche that includes resources from open savannah/woodland habitats. δ¹³C_enamel-1750 values range from −12‰ to 0.9‰, reflecting a dietary spectrum of C₃ forest/woodland plants to savannah C₄ vegetation. This variation is partitioned differentially among hominin taxa with implications for characterizing species-specific niches (Fig. 5). Much speculation has taken place regarding the ¹³C enriched dietary items that may have contributed to these novel isotopic profiles, including C₄ grass products (seeds, rhizomes, coms, tubers, bulbs, blades; e.g., Paine et al., 2018; Cerling et al., 2011; Dominy et al., 2008), C₄ sedge components (e.g., Cyperus papyrus; e.g., Hatley and Kappelman, 1980; van der Merwe et al., 2008), C₄ dicots (e.g., Peters and Vogel, 2005), invertebrates (termites, mollusks, crustaceans; Sponheimer et al., 2005; Stewart, 2014), small vertebrates (Sponheimer et al., 2013), large grazing mammals (Sponheimer and Lee-Thorp, 1999), birds and eggs (Peters and Vogel, 2005), fruits or nuts (Lee-Thorp et al., 2000), reptiles (Peters and Vogel, 2005), fish (Stewart, 2014), CAM plants (Peters and Vogel, 2005), or even highly water stressed C₃ plants (Kohn, 2010). It remains difficult to reconcile the potential ¹³C enriched resources that were utilized with other lines of evidence such as microwear or masticatory morphology (Grine et al., 2012; Sponheimer, 2013). Identifying which of these resources, or others not yet considered, are responsible for the unique hominin dietary signatures remains beyond the scope of this study, but a key to unraveling this issue is translating enamel isotopic values into postweaning/adult dietary signatures in a biologically meaningful way.

As emphasized throughout this paper, reconstructing the diets of our hominin ancestors using δ¹³C dietary signatures will be best accomplished by excluding teeth that form before weaning from analysis and applying an appropriate enrichment factor to maximize the likelihood of deriving adult dietary isotopic signatures. Implementing these two considerations to obtain the preceding results produces several insights.

First, assuming that the 11.8 ± 0.3‰ offset is appropriate for hominins, there is no δ¹³C_diet-1750 overlap between chimpanzees and hominin species shown in Figure 4, generally supporting a dietary distinction between the hominin lineage and Pan. The enamel of chimpanzees living in savannah/woodland communities (e.g., Fongoli, Ugalla, Ishasha) has yet to be analyzed isotopically, but isotopic analyses of hair samples indicate that δ¹³C_diet of savannah chimpanzees are ca. 1‰ more positive relative to chimpanzees in forested environments (Schoeninger et al., 1999, 2016; Sponheimer et al., 2006). Thus, the extreme ¹³C-enriched range of savannah chimpanzee δ¹³C_enamel values might approach isotopic signatures of A. anamensis and A. sediba (Fig. 5). If modern chimpanzee dietary (or forest) niches are hypothetically assumed to be representative of the last common ancestor (LCA) of the chimpanzee-hominin clade, the small difference in isotopic signatures between chimpanzees and some early hominins is important as it hints at an isotopic continuum, that is, a more gradational difference in diets rather than an abrupt niche shift. Paleoecological reconstructions of the only known chimpanzee fossil site, ca. 500 ka in the Kapthurin Formation, Baringo Basin (McBrearty and Jablonski, 2005), indicate a wooded grassland potentially similar to modern ‘savannah’ chimpanzee habitats (Leslie et al., 2016). While an association of open habitats with the only fossil chimpanzee assemblage may reflect a taphonomic artifact, this finding raises the possibility that chimpanzees in the recent past also had a relatively broad niche or wide habitat tolerance.

Second, dietary reconstructions based on the 11.8 ± 0.3‰ enrichment factor derived in this paper indicate that hominins began incorporating non-C₃ dietary components as early as 4.4 Ma with Ar. ramidus. Previous assessments have attributed a transition to significant C₄/CAM-derived dietary resources to A. afarensis ca. 3.45–3.0 Ma (Sponheimer et al., 2013) or A. anamensis at 4.2–3.9 Ma (Quinn, 2019). The revised dietary reconstruction here indicates that Ar. ramidus represents the earliest known instance of hominin foraging in more open woodland habitats, a behavior consistent with the wooded savannah reconstructions for Ar. ramidus at Aramis (Cerling et al., 2010) instead of a woodland-to-forest habitat (White et al., 2009).

Third, when considering only enamel potentially representing postweaning/adult diets (black box-and-whisker plots in Fig. 5), 22 out of 24 Pa. boisei, 12 out of 19 Pa. aethiopicus, and the only A. bahrelghazali specimen, yielded values consistent with an exclusive C₄/CAM dietary niche (δ¹³C_diet >14‰). This suggests obligate foraging on the most ¹³C enriched dietary resource(s) and considerable feeding specialization in these lineages. However, a reconstructed δ¹³C_diet-1750 of −18.3‰ derived from a single Pa. boisei-like maxillary fragment from Malawi (Ludecke et al., 2018; Kullmer et al., 1999) (depicted as a star in Fig. 5 Pa. boisei group) indicates a more eclectic mixed C₃/C₄ foraging niche similar to Pa. robustus from South Africa. Craniodental morphology and dental microwear studies of Pa. boisei have been interpreted to indicate repetitive loading of abrasive and fibrous foods, possibly including C₄ grass/sedge parts (Grine et al., 2012). To the extent that these may represent low-quality dietary foods, it is possible that these C₄ grass/sedge-consumers possessed a digestive physiology and/or a microbiome to facilitate fermentation. Enhanced fermentation would be associated with increased ¹³C-depleted methane production, resulting in potentially a larger enrichment factor ε and therefore dietary signatures that would be shifted to more negative values than depicted in Figure 5 (Passey et al., 2005). If this were the case, the δ¹³C_diet values for these taxa would not differ as much from those previously estimated using the 14.1‰ enrichment factor based on ruminant herbivore physiology. Thus, while the new hominoid enrichment factor of 11.8 ± 0.3‰ calculated here may be more suitable than a ruminant enrichment value, caution is warranted.

Fourth, all hominin taxa include isotopic values that indicate some proportion of C₄/CAM/¹³C-enriched dietary components, suggesting that a diet including resources from open woodland or grassland habitats may be a synapomorphy of the hominin clade. This is notable as previous reconstructions of hominin evolution have suggested that the exploitation of C₄/CAM/¹³C-enriched foods occurred later, in more dentally (and postcranially) derived taxa. A. sediba (n = 1) is a possible exception, with a δ¹³C_diet value of ca. −23‰, consistent with either a minor C₄ dietary component or consumption of highly xeric C₃ vegetation characteristic of habitats not occupied by primates today (MAP <500 mm; Kohn, 2010). However, the small sample of A. sediba, and arboreal postcranial features compatible with evolutionary reversals (Zipfel et al., 2011), preclude making any firm conclusions and underscore the need for additional isotopic sampling of this species.

Fifth, while this study examines δ¹³C_enamel data from all available hominin teeth whose values have been published (red and blue box-and-whisker plots in Fig. 5), the isotopic profiles developed in the black box-and-whisker plots in Figure 5 exclude all unidentified teeth as well as those that have been tentatively or definitively identified as deciduous teeth, incisors, canines, or M1s, since all of these teeth may incorporate non-adult dietary signals. For chimpanzees, there was ample justification to exclude data from other teeth, except those from M3s, but δ¹³C_enamel values from premolars, M2s, and M3s of the extinct hominins were included. Although dental developmental parameters such as crown formation times have been assessed for several fossil hominin taxa (Smith et al., 2015), and were found to develop more quickly relative to chimpanzees, it remains unknown whether the age at weaning completion for fossil hominins was most similar to that determined for chimpanzees at 4.0–4.5 years old (Badescu et al., 2017). Alternatively, fossil hominin weaning completion could have been closer to the mean age of weaning completion determined for 113 modern, non-industrialized human populations (2.4 ± 0.8 years; Sellen, 2001) or the mean of 2.80 ± 1.32 years for 39 archaeological human populations (Tsutaya and Yoneda, 2013). Recently, Joannes-Boyau et al. (2019) have shown that the length of time of exclusive nursing in A. africanus may have resembled the nursing duration in humans more than great apes. This, in turn, is compatible with the hypothesis that weaning may have been completed earlier relative to dental development in some extinct hominin species. Consequently, data from premolars and M2s for the fossil hominins were included in this study as potentially representative of postweaning/adult diets. Further work using trace elemental evidence for weaning-related dietary changes in the teeth of fossil hominin taxa (e.g., Austin et al., 2013; Joannes-Boyau et al., 2018; Smith et al., 2018) will be required to resolve this issue.

Sixth, previous dietary reconstructions using enrichment offsets derived from herbivores make the assumption that individuals occupy an exclusive herbivorous dietary niche. However, most hominins have been suggested to be more omnivorous than chimpanzees, although the extent of an eclectic diet that extends beyond plant components remains unknown for the different hominin lineages. Feeding on variable combinations of the flesh, organs, marrow and other tissues of large and small vertebrates, arthropods such as termites and grasshoppers, eggs, honey, and aquatic invertebrates alters the assumption that there will be a simple dichotomy with members of some hominin species ingesting C₃ foods, while others rely on a C₄ diet. The intra- and interspecific isotopic variation evident in the fossil hominin record displayed in Figure 5 likely reflects multiple dietary niches. Distinct dentognathic morphologies, variable tooth wear patterns, and sympatry in early hominins collectively support significant niche separation and trophic flexibility, as well as the idea that dietary sources of ¹³C-enrichment are likely to be different within and between the hominin lineages. Thus, it remains critical to continue to combine multiple lines of evidence, including isotopes, when reconstructing hominin diets (e.g., Grine et al., 2012).

Seventh, several hominin species, including A. afarensis and A. africanus, exhibit considerable isotopic variability relative to chimpanzees. In the case of A. afarensis, this variability has been attributed to dietary flexibility linked to unique landscape use patterns and malleable foraging strategies (Wynn et al., 2013), although short term habitat change (ca. 3.4–3.0 Ma) was acknowledged as a potential confounding variable. Of the 20 teeth of A. afarensis analyzed isotopically thus far, two represent unequivocal M3 samples, with the remainder a scattering of M1s, M2s, molar fragments, premolar fragments, P3s and P4s. Wynn et al. (2013) noted that intertooth variability may be due to feeding on different plant foods within and between years, a proposal consistent with the chimpanzee data presented here. However, to date, most of the more ¹³C-depleted values (six out of the nine values < −7.0‰) reported for A. afarensis derive from enamel that likely includes preweaning signals, indicating that more remains need to be analyzed to characterize the isotopic dietary ecology of this taxon. Recent analyses of Pa. aethiopicus enamel from the Lower Omo Valley (Ethiopia) considerably extend the range of variation previously reported for this taxon (Wynn et al., 2020). This range now reflects a dramatic shift in the median δ¹³C_enamel-1750 values from −7.8‰ to −1.0‰ at 2.37 Ma (Wynn et al., 2020). These data indicate that evolution of the robust dentognathic morphology of Pa. aethiopicus predates the reliance on C₄ foods documented after 2.37 Ma, suggesting that this adaptive complex initially evolved for feeding primarily on C₃ dietary items, with perhaps fallback feeding on lower quality, fibrous ¹³C enriched dietary items. Sex differences in weaning behavior should also be considered as a source of δ¹³C_enamel variability seen within hominin taxa. If sex-based weaning behavior differences, like those suggested by the values recorded in the teeth of chimpanzees from Liberia (Smith et al., 2010), were present in fossil hominins, this could account for some of the δ¹³C_enamel variation, with males potentially exhibiting relatively more ¹³C depleted M3 values than the females (Fig. 5). However, since sex has not been definitively assessed for any fossil hominin dental specimens that were isotopically analyzed, it remains unclear whether delayed weaning of male offspring by high-ranking females, as may occur at Taï (Boesch, 1997; Boesch and Boesch-Achermann, 2000), could be responsible for some of the variation in hominin dietary signals in Figure 5. As outlined in section 4.3, it may be possible to obtain a slightly later dietary signal by sampling the late-maturing outer layers of thick M3 imbricational enamel. If so, this should be attempted for future sampling of fossil hominin teeth to maximize the chances of obtaining a postweaning dietary signal.

Eighth, it is important to acknowledge that body size or factors associated with body size such as retention time or gut size, may play a significant role influencing the ε*_enamel-diet for the various hominin taxa and chimpanzees. Tejada-Lara et al. (2018) developed a regression equation for body mass and ε*_enamel-diet for non-ruminant taxa. The relationship is positive but with a relatively low correlation coefficient (r² = 0.62), suggesting there are other factors related to digestive physiology that influence the enrichment factor. Applying the minimum and maximum hominin body size estimates of 29 kg for a female A. afarensis and 60 kg for a male Homo rudolfensis (McHenry and Coffing, 2000) to this equation yield enrichment factors of 12.4‰ and 12.7‰, respectively. Using the average body mass of an adult male P.t. schweinfurthii of 43 kg (Carter et al., 2008) predicts an enrichment factor 12.5 ‰. The enrichment factor for P.t. schweinfurthii based on this regression is 0.7‰ greater than the ε*_enamel-diet of 11.8 ± 0.3‰, calculated in this study. This difference can be accommodated by the fact that nearly 40% of the variance in enrichment factors is not directly explained by the regression based on body size. In addition, the appropriate regression for hominoids (Tejada-Lara et al., 2018) utilizes a composite of herbivore taxa with unspecified digestive fermentation, presumably incorporating a number of additional variables. This degree of uncertainty easily encompasses the enrichment factor of 11.8 ± 0.3‰ yielded here by careful consideration of the factors inherent in characterizing a composite dietary signal and appropriate enamel sampling.

5. Conclusions

5.1. Summary findings and implications

Reconstructing fossil hominin diets from stable carbon isotopic signals in enamel bioapatite requires the application of a δ¹³C enamel-diet enrichment factor, which has never before been empirically determined for any hominoid primate. This study calculates the 1750-adjusted dietary isotopic input value (−25.3‰) for chimpanzees at Ngogo in Kibale National Park, Uganda, and presents a new average δ¹³C_enamel-1750 signature of −13.8 ± 0.3‰ for four natal male chimpanzees. These values are then used to estimate the first ε*_enamel-diet for a hominoid of 11.8 ± 0.3‰. This study also documents the largest collection of wild chimpanzee δ¹³C_enamel values thus far, including the first data from P. t. troglodytes. These data reveal significant isotopic dietary differences between chimpanzee subspecies, reflecting geographic variation in rainfall and relationships with habitat types that accord with subspecies attributions.

The results presented here indicate that utilization of a general herbivore isotopic enrichment factor and inclusion of unidentified teeth and teeth that formed before weaning completion, have led to dietary reconstructions for fossil hominin taxa that are more ¹³C-depleted and varied than may be the case. Application of the new chimpanzee-specific enrichment factor to 12 fossil hominin taxa effectively extends the hominin niche even further towards open, water-stressed landscapes that include C₄/CAM resources, increased omnivory including vertebrates and invertebrates, or some combination of these. Although these isotopic data corroborate previous interpretations of hominins as primarily savannah foragers, the extreme ¹³C-enriched components of hominin dietary signatures remain difficult to interpret. The convergence of δ¹³C_enamel hominin values with obligate C₄ grazers, for example, requires consideration of unique hominoid niches that have no modern analog.

It is important to note that enamel sampling protocols that do not account for intra- and intertooth δ¹³C_enamel variation can introduce errors up to 2.2‰, as seen in Table 3. This error could be even greater in taxa with longer periods of crown formation or inhabiting more seasonally variable environments. The cumulative effect of these considerations, in combination with an inappropriate enrichment factor, can result in significant differences in dietary interpretations. For example, using the generalized herbivore ε of 14.1‰ and sampling a chimpanzee M1, with a pre-weaning signal, (e.g., −15.4‰ from the M¹ of MUZM 2625 in Table 3), yields a reconstructed δ¹³C_diet value −29.5‰. This would be 4.5‰ more depleted relative to a δ¹³C_diet value of −25.0‰ reconstructed using the most ¹³C enriched value from that individual’s M³ enamel, −13.2‰, and, of course, an ε of 11.8 ± 0.3‰ (Table 3). While the quantitative adjustments for each of these considerations may be small, in some cases on the order of tenths of a per mil, the cumulative effect can critically alter the location of the reconstructed diet on the isotopic spectrum as well as the range of variation attributed to any hominoid species or community.

5.2. Limitations and future directions

The results presented here emphasize the need to carefully consider and integrate multiple factors to estimate a δ¹³C diet-enamel enrichment factor. The single numerical value represented by this offset, ε*_enamel-diet, distills numerous components contributing to isotopic variability in enamel, including diversity and relative abundance of food items over multiple years, seasonal cycling of resource base and isotopic values, intra- and interpopulation differences in foraging strategy, digestive physiology, variable fractionation during routing, bioavailability of dietary nutrients, enamel developmental considerations, intra- and inter-ecosystem isotopic variability, and shifting patterns of δ¹³C_atmos-CO2. Developing a framework to comprehensively assess all these variables is challenging. In addition, given inter- and intraspecific ranges of ca. 6‰ and 1‰, respectively (Passey, et al, 2005; Cerling and Harris, 1999), in the ε*_enamel-diet for mammals analyzed, and only limited studies of the underlying mechanisms mediating this variability, identifying an appropriate offset value remains problematic for most taxa.

While the new chimpanzee enrichment factor of 11.8 ± 0.3‰ calculated here may be more suitable than a ruminant enrichment value to reconstruct the diets of fossil hominins and hominoids, it is possible, even likely, that intra- and interspecific variation in the enrichment factor existed among fossil hominins due to differing digestive physiology, body size, diet nutrient/fiber quality, microbiome nature/activity, basal metabolic rates, and energy budgets. The inherent complexities involved in computing an enrichment factor mean that there is likely no single value that can encompass the eclectic foraging strategies of hominoids in general. Further refinement of the enrichment factor most appropriate for fossil hominins or other primates will require additional quantitative studies on living primates that include determining: (1) isotopic values of individual food items; (2) relative contribution of each item to the total diet; and (3) correlative δ¹³C_enamel values for multiple taxa spanning a range of body sizes, digestive systems, and ecological niches, particularly those of other wild great ape populations.