Pre-Clovis occupation of the Americas identified by human fecal biomarkers in coprolites from Paisley Caves, Oregon

All of the coprolites analyzed are identifiable as intact structures (figs. S4 and S5); this is critical as it constrains the interpretation of observed biomarkers to those consistent with the observed morphology, i.e., a species assignment of a rat (for example) would not be viable, as it would be inconsistent with the shape of the coprolite. Comparative assessment of lipids derived from coprolite 283 and a sediment sample collected directly underneath the coprolite (table S2) reveal distinctly different distributions where sediments are differentiated by a high level of n-alkanols (C₁₄-C₃₄) with an even-over-odd predominance, characteristic of an input from higher plants (27). Sterols, including sitosterol and stanols (5α-cholestanol and 5α-stigmastanol), microbially transformed in the environment, are also dominant within the lipid extract (28). Crucially, this shows that the lipids are representative of the matrix from which they are derived and have experienced minimal leaching or lateral movement within the sedimentary profile. Fecal biomarkers predominate in lipid extracts from coprolites recovered from the younger phases of the cave, and there is a general trend toward the concentration of fecal sterols decreasing as radiocarbon age of the coprolite increases from 573 μg g⁻¹ in the youngest coprolite to 9.5 μg g⁻¹ in the oldest coprolite. Lipids (and hence fecal biomarkers) are chemically stable and, although some microbial or diagenetic alterations do occur with time, the resulting structures can be linked back to their original source (29). Environmental conditions can greatly influence the preservation of fecal sterols (30–33). The slow deposition of sediment and aerobic conditions of the burial environment is likely the cause for the greater degradation of lipids in the oldest coprolites.

Fecal biomarker interpretations for 13 of 21 coprolites analyzed agree with mtDNA classifications reported previously (Table 1) (3, 5). Of these, the majority indicate a human origin (both Homo sapiens mtDNA haplogroups A and B) and range in age from 1308 ± 28 to 12,260 ± 60 ¹⁴C yr B.P. Two of the 13 coprolites reported as containing mtDNA from Lynx rufus (216) and Panthera leo (271) yielded fecal biomarker distributions consistent with those of a carnivore. A single coprolite (195) reported to contain mtDNA from H. sapiens and Canis lupus/familiaris also yielded fecal biomarker distributions indicative of a mixed human/carnivore origin. One explanation for these results from a single specimen is coprophagy of human feces by carnivores. Coprophagy in canids is a well-recognized phenomenon (34) and could explain both the presence of carnivore-derived lipids and human mtDNA, but it is unlikely that humans were eating carnivores as suggested by Gilbert and coworkers (6). This is supported by the fecal biomarker evidence, i.e., only coprophagy would result in a human/canid mix of biomarkers in a discrete coprolite. Moreover, not only carnivores have a lower biomass transfer efficiency but also consumption of carnivore meat can be hazardous to human health due to parasitic infections and the bioaccumulation of toxins (35–37). Of the remaining eight coprolites, two did not yield enough lipid to be viable for biomarker analysis. At 12,125 ± 30 ¹⁴C yr B.P. (283) and 12,345 ± 55 ¹⁴C yr B.P. (185), these both represent coprolites from the oldest deposits in the cave and a lower recovery of lipid is therefore not unexpected.

Six of the coprolites classified as human using mtDNA contain fecal biomarkers indicative of carnivores. Sterols derived from canid feces are characterized by very high levels of cholesterol, and only trace amounts (if any) of cholesterol-derived 5β-stanols (coprostanol and epicoprostanol) as canids lack the gut microbiota responsible for producing these compounds (38). Carnivores differ from other animals in their bile acid composition and are identifiable from the higher levels of cholic and chenodeoxycholic acids in addition to deoxycholic acid (39–42). Conversely, human feces are usually characterized by high levels of coprostanol and higher amounts of deoxycholic acid and lithocholic acid (43). Previous work has reported that about a quarter of North American subjects in a study exhibited little to no conversion of cholesterol to coprostanol (44). This may potentially result in false negatives when differentiating human coprolites from carnivores and may explain the lack of agreement between mtDNA and fecal biomarker results in these six coprolites.

mtDNA analyses rely upon amplification of ultratrace amounts of extant mtDNA using the PCR, thereby making such an approach not only more sensitive but also more susceptible to contamination (45). Conversely, fecal biomarker analyses are reliant upon coprolites containing enough of each compound to enable detection and quantification (typically in the region of >0.4 μg g⁻¹_coprolite), with the concentration of compounds remaining being a direct reflection of age and depositional environment (30–33). However, because of their source specificity, contamination is very unlikely. Ideally, both techniques should be used together in a complementary fashion for the assessment of feces from both single and multiple sources. While fecal biomarker analysis lacks the detailed information that mtDNA may provide, it does deliver a more rapid assessment of coprolites that, as in this study, can provide robust supporting evidence where results obtained by mtDNA-based approaches are open to criticism. It also represents an ideal screening method before attempting mtDNA analyses that are far more laborious and require a greater degree of verification, i.e., Gilbert and coworkers (6) used several different DNA typing and sequencing methods across six independent laboratories. Where no fecal biomarker results are available, the matrix either is of a nonfecal origin or is sufficiently old and/or has been environmentally exposed such that interpretations of a fecal origin using mtDNA only should be, at best, speculative.

The peopling of the Americas is a complex story. The continents are large, and early populations were likely to be sparsely distributed. The lack of early human skeletal remains, and the ethical issues associated with destructive analysis of the handful that have been discovered (46, 47), means that coprolites are the best direct archaeological evidence that people occupied a particular site. Human occupation at Paisley Caves is now proven to 12,200 ¹⁴C yr B.P. using fecal lipid biomarkers. Coprolites 194 and 280 are the oldest coprolites determined to be unequivocally human based on fecal biomarker analysis, a methodology that bypasses current uncertainties surrounding mtDNA. These coprolites also show agreement between both mtDNA and fecal biomarker analyses. Fecal lipid biomarkers offer access to hitherto unknown (or at best, uncertain) information, adding to the growing body of evidence that is helping to build a picture of not only when and by what route people arrived but also how these people adapted to diverse landscapes across the continent (48–50).

Approximately 0.5 g of a coprolite was crushed using a mortar and pestle and then passed through a 2-mm sieve. Suitable amounts of two internal standards (hyocholic acid and preg-5-en-3β-ol; 50 μl, 0.1 mg ml⁻¹ solution) were added to the powdered samples. The lipids were then microwave-extracted into 10 ml of 2:1 dichloromethane/CH₃OH (v/v). The total lipid extract obtained was subsequently hydrolyzed, and the sterol and bile acid fractions were isolated as outlined by Bull et al. (51) and Elhmmali et al. (43), respectively. For the methylation of bile acids, BF₃-methanol was used as an alternative methylation reagent. The fractions containing the target biomarkers were then analyzed by GC and GC-MS.

Initial screening of biomarker isolates was performed using an HP 5890 Series II gas chromatograph. Trimethylsilylated sterols and methylated and trimethylsilylated bile acids were introduced (1.0 μl) via an on-column injector. The analytical column was a 50 m × 0.32 mm fused silica capillary column coated with a 100% dimethylpolysiloxane nonpolar stationary phase (HP-1, 0.17 μm; Agilent). For the sterol analyses, the GC temperature program was set to hold at 50°C for 2 min, followed by a gradient increase to 200°C at 10°C min⁻¹, and then to 300°C at 3°C min⁻¹ before a final isothermal at 300°C for 20 min. For the analyses of bile acids, the GC temperature program was set to hold at 40°C for 1 min, followed by a gradient increase to 230°C at 20°C min⁻¹, and then to 300°C at 2°C min⁻¹ before a final isothermal at 300°C for 20 min. Helium was used as the carrier gas set to constant flow of 2.0 ml min⁻¹, and the flame ionization detector (FID) used to monitor column effluent was kept at a constant temperature of 300°C. Data were acquired using DataApex Clarity (version 2.6.2.226).

GC-MS analyses of sterols and bile acids were performed using an Agilent 7890-7200B GC/Q-TOFMS. Samples (1.0 μl) were injected using a 7693 autosampler and a multimode inlet (set to track the oven temperature) onto a 50 m × 0.32 mm (Agilent) fused silica capillary column coated with a 100% dimethylpolysiloxane nonpolar stationary phase (HP-1, 0.17 μm; Agilent). The GC temperature programs used were the same as for the corresponding GC analyses. Helium was used as the carrier gas, set to a constant flow of 2 ml min⁻¹. The GC transfer line, ion source, and quadrupole were set to 320, 230, and 150°C, respectively. The MS was set to acquire in the range of m/z (mass to charge ratio) 50 to 1050 at a scan rate of 0.2 Hz in extended dynamic range mode.