Rubidium

For the solution data, sampling and isotope analysis was done in a prior study and is detailed in Britton et al.^{23 (link)}. To summarise, the second and third molars (M₂ and M₃) were extracted from the mandibles, brush-cleaned with water and left to dry overnight. The whole teeth were mechanically abraded to remove surficial enamel. Sampling was done on the buccal face of the anterior loph as it presented a thicker enamel, with the face removed and cleaned from adhering dentine using a tungsten carbide burr. Enamel faces were then marked for horizontal sectioning at ∼ 1.5 mm intervals, ultrasonicated in deionize water (DI H₂O, 18.3 MΩ) and dried, before being cut into strips using diamond-coated superfine circular drill bits. Samples were then individually ultrasonicated in DI H₂O, dried and split longitudinally with ∼ 5 mg of enamel being reserved for ⁸⁷Sr/⁸⁶Sr solution analysis and the remainder being retained for carbon and oxygen isotope analysis. In that study, sections were numerically assigned from the ERJ to the OS (M₂-1, M₂-2, M₂-3,…).
Strontium was isolated from enamel in clean laboratory facilities at the Department of Human Evolution, Max Planck Institute for Evolutionary Anthropology (MPI-EVA), Leipzig, Germany using a modified version of the method from Deniel and Pin^{28 (link)} described in detail in Copeland et al.^{34 (link)}. The following description of the analytical procedure is reproduced from Britton et al.^{23 (link)}, and is also presented there in full. The ∼ 5 mg samples were dissolved in 1 ml 14.3 M high purity HNO₃ then evaporated to dryness. The obtained residue was then re-dissolved in 1 ml 3 M HNO₃ before loading into pre-conditioned columns containing Sr Resin (Eichrom Technologies, Lisle, IL, USA), being passed through three times. Strontium was then eluted using ultrapure deionized water (18.2 MΩ), dried and re-dissolved in 3% HNO₃ and analysis of ⁸⁷Sr/⁸⁶Sr ratios was undertaken using a Thermo Fisher Neptune™ (MC-ICP-MS). All the acids solutions used in the procedure were purified through a PicoTrace double-distilled sub-boiling distillation system. The subsequent ⁸⁷Sr/⁸⁶Sr measurements on standards and samples were corrected for interferences from krypton (Kr) and rubidium (Rb) and normalized for instrumental mass bias to ⁸⁸Sr/⁸⁶Sr = 8.375209 (exponential law). Analysis of the international strontium isotope standard NIST SRM987 (National Institute of Standards and Technology, Gaithersburg, USA) during each analytical session was used for external normalisation of data (long-term ⁸⁷Sr/⁸⁶Sr value = 0.710273 ± 0.000033 (2σ) (n = 97)). All ⁸⁷Sr/⁸⁶Sr values reported here were adjusted so SRM987 = 0.710240⁷², involving a data correction factor of − 0.00002. Strontium concentrations of the enamel samples were determined using the method described in^{34 (link)}, which is accurate to within ± 31 ppm.

We measured 24 urinary metals in urine samples, including lithium, beryllium (Be), aluminum (Al), titanium, vanadium (V), chromium (Cr), manganese, iron (Fe), cobalt, nickel (Ni), Cu, Zn, As, Se, rubidium, strontium, molybdenum (Mo), Cd, indium (In), tin, antimony (Sb), barium, thallium (Tl), and Pb using inductively coupled plasma-mass spectrometry (ICP-MS, NEXION 300X. PerkinElmer Inc. Waltham, Massachusetts, USA). Briefly, urine samples were thawed at room temperature and then centrifuged (4,200 rpm × 10 min) at room temperature. Afterward, 0.5 ml of supernatant from each urine sample was added into a 15 ml polypropylene tube, and then 4.5 ml of 2% nitric acid solution (10 times the diluted urine sample) was added. To assess the accuracy of the measurements, Seronom^TM Trace Elements Urine L-1 (Sero Incorporated Company. Billingstad, Norway), Seronom^TM Trace Elements Urine L-2 (Sero Incorporated Company. Billingstad, Norway), and Trace Elements in Natural Water (SRM1640a) (National Institute of Standards and Technology. Gaithersburg, Maryland, USA) were added as quality control samples for each batch. As shown in Supplementary Table 1, the range values of 66.04–152.86% in urinary metals were considered acceptable for spike recoveries. The limits of detection (LOD) ranged from 0 to 1.66 μg/L for urinary Fe and from 0 to 0.13 μg/L for urinary Zn. The limits of quantification (LOQ) ranged from 0.01 to 5.54 μg/L for urinary Fe and from 0 to 0.44 μg/L for urinary Zn. Urinary concentrations of Be, In, and Sb were excluded from further analysis because the values of Be, In, and Sb in more than 80% of individuals were below the corresponding LOD. Values of urinary metals below the LOQ were replaced by LOQ/2.

Analysis of Spirulina samples’ macro-elemental composition was performed non-destructively by Energy Dispersive X-Ray Fluorescence Spectrometry (EDXRF) to determine the following elements (13): phosphorous (P), titanium (Ti), zinc (Zn), silicon (Si), bromine (Br), sulfur (S), chlorine (Cl), manganese (Mn), rubidium (Rb), strontium (Sr), potassium (K), calcium (Ca), and iron (Fe). Powdered samples were pressed into 0.5–1.0 g pellets for analysis using a pellet die and a hydraulic press. For fluorescence excitation disc radioisotope, excitation sources Cd-109 (20 mCi, Eckert and Ziegler, Berlin, Germany) and Fe-55 (25 mCi, Eckert and Ziegler, Berlin, Germany) were used. An EDXRF spectrometer with a PX5 digital pulse processor (Amptek, Bedford, MA, USA), an XR-100 SDD detector (Amptek, Bedford, MA, USA), and a PC-based, multichannel analyzer software package (DPPMCA) were used for the emitted fluorescence radiation detection. For light element analysis (Si, P, S, and Cl), the spectrometer operating in Fe-55 mode was equipped with a vacuum chamber, and for K, Ca, Ti, Mn, Fe, Zn, Br, Rb, and Sr analysis, measurements in Cd-109 mode were performed in the air. The energy resolution of the spectrometer was 125 eV at 5.9 keV. AXIL Spectral Analysis software was used to analyze the complex X-ray spectra. For quantification, the Quantitative Analysis of Environmental Samples (QAES) software developed in our laboratory was used [35 ,36 (link)]. Method validation was performed using 1573a (tomato leaves) and 1547 (peach leaves) NIST standard reference materials. The EDXRF analysis estimated uncertainty budget was 11% and was incorporated in the QAES software procedure.