Stable oxygen isotope compositions from vertebrate tooth enamel are commonly used as biogeochemical proxies for paleoclimate reconstructions. However, the utility of enamel isotopic values across species varies due to differences in rates of enamel deposition and mineralization as well as sources of ingested water, body water residence times, and species’ physiology. We evaluate the use of stable oxygen isotope compositions from pronghorn (Antilocapra americana Gray, 1866) enamel for the amplitude reconstruction of terrestrial paleoclimate seasonality. We serially sampled the third lower molars of pronghorn from Wyoming for oxygen isotope composition in phosphate (δ¹⁸O_PO4) and compared patterns to: (1) interpolated and (2) measured yearly variation in central Wyoming environmental waters (δ¹⁸O_w) as well as to (3) δ¹⁸O values from sagebrush leaves and stems in the same region. Although we recognize the numerous factors influencing the composition of mammal body water, our null hypothesis was that δ¹⁸O_PO4 enamel values reflect δ¹⁸O_w with a constant offset due to mammalian physiology. We set up our null hypothesis by converting δ¹⁸O_w values to δ¹⁸O_PO4 using a published regression based on empirical results from mammals. Pronghorn δ¹⁸O_PO4 values from enamel are consistently enriched in ¹⁸O relative to the predicted values. We hypothesized that pronghorn δ¹⁸O_PO4 values might also reflect dietary water and therefore also converted δ¹⁸O_leaf values from plants into predicted δ¹⁸O_PO4 values. We infer that pronghorn obtain at least some of their water from ¹⁸O-enriched plants because pronghorn enamel δ¹⁸O_PO4 values are more similar to predicted δ¹⁸O_PO4 values from plants than from meteoric waters. Modeling of source body water δ¹⁸O values show amplitudes between 70% and 95% of seasonal variation from Wyoming δ¹⁸O_w values. Collectively, our findings establish that modern seasonality in source water is reliably reflected in modern pronghorn enamel providing the basis for exploring changes in the amplitude of seasonality of ancient climates using archived tooth collections.

Modern and archaeological pronghorn specimens were acquired from the University of Wyoming Anthropology museum and Department of Archaeology (Appendix I). All specimens were collected from wild populations in Wyoming during 1970-1972 and 2010 following deaths that were not related to this study. The archaeological specimens date to 1720 ± 100 years (Frison 1971), thus pre-dating the rapid climate change typical of the late 20^th century (Mann et al. 1998, Jones et al. 2001). The lower third molar (m3) is one of the last to complete enamel mineralization and erupt in pronghorn (Dow and Wright 1962), therefore, we included only individuals with erupted third lower molars. To recover the most complete isotopic time series, we included only individuals showing no or little wear of the m3. We also excluded individuals with abscesses or obvious abnormalities of the dentition or jaw bone. We extracted lower third molars using a Dremel diamond cutting wheel and serially sampled the enamel at ~2 mm intervals using a Dremel tool with a diamond taper point bit (part #7144). We collected 2-3 mg of powdered enamel for each serial sample. Further, we took bulk samples (~4-6 mg) of bone from the mandibular angle just posterior to third lower molar for each individual.

To analyze the oxygen isotope composition of phosphate (δ¹⁸O_PO4), we weighed 1.5-2.0 mg of enamel and 3-4 mg of bone from each specimen. Preparation procedures for the modern specimens are from a combined approach based on Bassett et al. (2007) and Weidemann-Bidlack et al. (2008). We pre-treated all samples with 300 μl of 2.5% NaOCl for approximately 20 hours to remove organics. Bone samples were usually pre-treated twice to ensure complete organic removal (or more if there was continued gas production). Samples were then rinsed with deionized (DI) water 5 times and dried overnight at 50˚C. We then dissolved the remaining powder in 100μl of 0.5M HNO₃ overnight. To neutralize the solution and precipitate CaF₂, 75μl of 0.5 M KOH and 200μl of 0.36 M KF were added. Samples were centrifuged to pellet the CaF₂ and the supernatant was transferred from the vials to reaction vessels. We precipitated silver phosphate with 250µl of silver amine solution (0.2M AgNO₃, 0.35M NH₄NO₃, 0.74M NH₄OH) plus 3-6 drops of 0.1M AgNO₃ to initiate the precipitation. Samples were placed in a heat block at 50˚C overnight in a fume hood to allow for maximum crystal growth. The silver phosphate crystals were rinsed five times with ~2 mL of DI water to remove residual silver amine solution. After the samples dried overnight at 50˚C, 200 – 300 μg were weighed into pressed silver capsules and stored in an oven flushed with N₂ until isotopic analysis. Preparation of historical specimens used a rapid silver phosphate technique from Mine et al. (2018), which demonstrated d¹⁸O value fidelity to the slow precipitation method in Weidemann-Bidlack et al. (2008). Briefly, the historical samples were similarly treated for organic removal, but hydroxyapatite was dissolved in 50 mL 2M HNO₃ while CaF₂ precipitated with 30 mL 2.9M HF and neutralized with 50mL of 2M NaOH. Steps to pellet and isolate CaF₂ were similar between methods. To precipitate silver phosphate, we added 180 mL silver amine solution (0.37M AgNO₃ and 1.09M NH₄OH) and adjusted pH to 5.5-6.5 using dilute HNO₃, which was shown to maximize phosphate recovery (Mine et al., 2018). The silver phosphate crystals were rinsed and dried similar to the modern samples as outlined above.

The δ¹⁸O value of silver phosphate [SK1] [DF2] was measured after conversion to CO in a Temperature Conversion Elemental Analyzer (TC/EA, Thermo Scientific) coupled with a Conflo IV (Thermo Scientific) to a continuous flow isotope ratio mass spectrometer (CF-IRMS, Thermo Scientific Delta V). Three in-house reference materials (two silver phosphate and one benzoic acid) were used to normalize isotopic values and check the effectiveness of pyrolysis within and between runs (modern samples at UWSIF: ARCOS [N=4 per run], UWSIF21 [N=5 per run], UWSIF33[SK3] [DF4]  [N=6 per run]; historical samples at SIELO UCM: USGS80 [N=8], USGS 81 [N=8], IAEA601 [N=8]). Variation in d¹⁸O values exhibited by these reference materials was < 0.4‰. In addition, we monitored the potential for isotopic alteration during sample preparation by precipitating silver phosphate from a synthetic hydroxyapatite and NIST 120c (N=3 and 3 with 1s < 0.3‰). All samples were analyzed in triplicate. All d¹⁸O_PO4 values are reported relative to the standard V-SMOW.

The spacing between enamel carbonate and phosphate δ¹⁸O values is often used as a check for diagenesis (Koch et al. 1997) To check the carbonate-phosphate isotopic spacing, we also analyzed a subset of enamel samples for carbonate d¹⁸O values as per Koch et al. (1997) and Kohn and Cerling (2002). We weighed 1 mg of enamel and 5 mg of bone for each sample analysis of δ¹⁸O_CO3 values. To remove organic matter from the bone samples, we used 2-3% H₂O₂ at a ratio of 1 ml per 25mg of sample, leaving the caps of the microcentrifuge tubes open to allow the escape of gas for 24 hours. We did not pre-treat the enamel samples to remove organic matter due to the minimal organic content of enamel. Only the bone was pre-treated for carbonate analysis. Similar to the phosphate preparation, pre-treatment was repeated until gas production ceased. We rinsed the bone samples 5 times with DI water to remove H₂O₂ from solution. We then added 1M CH₃COOH with Ca acetate buffer (pH=4.5) to remove non-lattice bound carbonates using the same ratio as the preceding step. Samples were rinsed five times with DI water, dried for 24 hours in a freeze drier, and ~1 mg weighed into exetainer tubes for isotopic analysis. Once all samples and reference materials were weighed into exetainer tubes, they were dried overnight at 50°C, the headspace flushed with He, and 100-200 mL of  >100% H₃PO₄ was added to react for 24 hours (room temperature). The CO₂ within the headspace was sampled for isotopic composition measurement using a gas bench (Thermo Scientific) coupled to a CF-IRMS (Thermo Scientific Delta Plus) at UWSIF. Three in-house carbonate reference materials were used to normalize isotopic values and check for variation within and between runs (UWSIF18 [N=4 per run], UWSIF06 [N=4 per run], UWSIF17 [N=4 per run]).Variation in d¹⁸O_CO3 values of these reference materials was < 0.2‰. In addition, we monitored the potential isotopic alteration during sample preparation with one lab bioapatite (MSW0479, ashed manatee bone, N = 1 per run, d¹⁸O_CO3 = 19.0‰) and variation in d¹⁸O_CO3 values of a phosporite reference standard (NIST 120c, N = 5, d¹⁸O_CO3 = 28.4 ± 0.3‰).  All d¹⁸O_CO3 values are reported relative to the standard V-SMOW.

Recovering primary isotope time series

Modelling is the only means by which the primary isotopic time series can be recovered (Green et al. 2018), even for mammal taxa with relatively fast rates of enamel mineralization and low rates of isotopic overprinting. To recover the primary isotopic input signals from our pronghorn tooth isotopic time series, we used the mathematical model from Passey and Cerling (2002) and Passey et al. (2005) as a transfer function. The Passey and Cerling (2002) model takes primary isotope series using measured d¹⁸O values from body water and reconstructs tooth enamel d¹⁸O values, while also accounting for time averaging due to amelogenesis and variation in enamel maturation from the crown to the root. The Passey et al. (2005) method incorporates the Passey and Cerling (2002) time-averaging model and uses an inverse linear system to recover the input signals from enamel isotopic time series (see Matlab code associated with Passey et al., 2005). The model terms for the Passey and Cerling (2002) and Passey et al. (2005) models are 1_a and 1_m, which are the length of apposition (distance along the tooth from where a new enamel layer contacts the enamel-dentine junction and the external layer of the tooth; 7 mm) and length of maturation (the length of the tooth that is mineralizing at a given time; 13.3 mm assuming an enamel maturation time of one month as in small bovids such as sheep; Zazzo et al. 2010), respectively. The l_a parameter was estimated using values reported for sheep, which range from 2 mm to 12 mm, depending on whether l_a was measured on the buccal, mesial, or lingual side of the tooth and crown height (Zazzo et al. 2012). We used the midpoint of these values (7 mm) as a conservative estimate of l_a in pronghorn. We used a constant sample depth (l_s) of 75% of enamel thickness because we drilled through approximately 75% of the enamel when serial sampling. The Passey and Cerling (2002) and Passey et al. (2005) models were run in R version 3.5.1 (Appendix I).

To compare pronghorn enamel δ¹⁸O_PO4 values to seasonality of environmental waters, we downloaded interpolated monthly average isotope values for precipitation from central Wyoming (43˚N, 107.5 ˚W at 6,700 ft of elevation) from waterisotopes.org (Bowen et al. 2005, Bowen 2014). These interpolations for the entire United States are based in large part on Welker’s USNIP (United States Network for Isotopes in Precipitation; Welker 2000, 2012) for the years 1989-1994 and scarce IAEA GNIP (Global Network for Isotopes in Precipitation) data from a few years in the 1960’s for 6 sites across the United States (Rozanski et al. 1993). Environmental water δ¹⁸O_w values were converted into predicted phosphate values using the following equation from Kohn and Cerling (2002), which is based on empirical data from mammalian enamel:

δ¹⁸O_PO4 = (0.9 x δ¹⁸O_w) + 23  (2)

We use equation 2 as a means of setting up our hypothesis, that pronghorn δ¹⁸O_PO4 values reflect only δ¹⁸O_w and physiological offsets. Given that equation 2 was originally derived for evaporation insensitive species (i.e., those that rely primarily on meteoric waters), underestimation relative to measured δ¹⁸O values can reveal whether a focal species is evaporation sensitive.

To validate the interpolated δ¹⁸O_w values we used 1000+ measured δ¹⁸O values from precipitation at 9 sites in Wyoming (Appendix II) from USNIP for the entire USNIP record 1989-2006 (Welker 2000, Vachon et al. 2010b, Welker 2012). The high density modern precipitation network (USNIP) provides the only site, sub-state, regional and continental record of actual meteoric water values that are becoming increasing valuable in revealing the range of seasonality in modern precipitation (Vachon et al. 2007, Vachon et al. 2010b, a). We made comparisons between the interpolated δ¹⁸O_w values and measured values using monthly averages across all sample sites and at the site closest to the Laramie and Rawlins region (Albany site), the site closest to the area where our modern pronghorn teeth were collected. To evaluate the relative contributions of meteoric and river waters, we also extrapolated maximum and minimum δ¹⁸O values from river waters in Wyoming (Kendall and Coplen 2001). As with the meteoric waters, we converted river water δ¹⁸O values to enamel phosphate values.

We also obtained published δ¹⁸O values for water in sagebrush leaves and stems, rabbitbrush leaves, and pronghorn incisor enamel from Fenner and Frost (2008) for comparison to δ¹⁸O_PO4 values from pronghorn molar enamel (this study) as well as δ¹⁸O_w values (from interpolated and measured meteoric precipitation). All plant tissues were sampled by Fenner and Frost (2008) during the months of June and July; we converted these δ¹⁸O values into phosphate enamel values using equation 2. This conversion of plant leaf water values to enamel phosphate values allowed us to set up our alternative hypothesis: pronghorn δ¹⁸O_PO4 values from enamel reflect a combination of environmental water, evaporated leaf water, and physiological mechanisms. Similarity of pronghorn δ¹⁸O_PO4 values to δ¹⁸O_w from precipitation, lakes, and rivers or δ¹⁸O_leaf values should provide information on their relative inputs.

Appendix I. Raw δ¹⁸O data from serial samples of modern and archaeological pronghorn from Fraser et al. (2021) Pronghorn (Antilocapra Americana) enamel phosphate δ¹⁸O values reflect climate seasonality: implications for paleoclimate reconstruction. Ecology and Evolution, In Press. [Not sure if 2021 will be the correct year or if you want to put the entire citation here]

Appendix II. R translation of the function Emeas1_1 from Passey et al. (2005) from Fraser et al. (2021) Pronghorn (Antilocapra Americana) enamel phosphate δ¹⁸O values reflect climate seasonality: implications for paleoclimate reconstruction. Ecology and Evolution, In Press..

Appendix III. R translation of the function mSolv1_1 from Passey et al. (2005) from Fraser et al. (2021) Pronghorn (Antilocapra Americana) enamel phosphate δ¹⁸O values reflect climate seasonality: implications for paleoclimate reconstruction. Ecology and Evolution, In Press..

Appendix IV. Running the R translations of Emeas1_1 and mSolv1_1 with example data from Passey et al. (2005) from Fraser et al. (2021) Pronghorn (Antilocapra Americana) enamel phosphate δ¹⁸O values reflect climate seasonality: implications for paleoclimate reconstruction. Ecology and Evolution, In Press..

Appendix V. Summary data of Wyoming sites for which USNIP data were extracted from Fraser et al. (2021) Pronghorn (Antilocapra Americana) enamel phosphate δ¹⁸O values reflect climate seasonality: implications for paleoclimate reconstruction. Ecology and Evolution, In Press..