Although its composition can vary slightly, bioapatite is generally composed of calcium, phosphate, and a hydroxy group [Ca₁₀(PO₄)₆(OH)₂; Pate et al., 1994]. While an organism is alive, the compounds and elements of bioapatite are readily substituted by others (such as carbonates, strontium, barium, and magnesium), reflecting nutritional intake and conditions experienced by the individual as its bones and teeth matured (Pate et al., 1994; Pfretzschner, 2004; Wopenka and Pasteris, 2005; Kremer et al., 2012). Elements and compounds in bioapatite are also easily exchanged after burial (Wopenka and Pasteris, 2005; Keenan, 2016). The extent of diagenetic alteration to bones and teeth depends heavily on the timing and environment of burial (Trueman and Benton, 1997; Trueman et al., 2003; Piga et al., 2011). Biological alteration, generally caused by bacteria, leads to the loss of organics. Removal of organics leaves space for potential replacement and infilling by exogenous phosphates, carbonates, and other minerals (Shemesh, 1990; Smith et al., 2002; Kremer et al., 2012). Diagenesis also results from abiotic processes, such as the dissolution of bioapatite under acidic conditions.

As noted above, there are several widely used methods for evaluating specimen preservation. The amount of rare earth elements (e.g., uranium), or halogens (e.g., fluorine; chlorine) in bioapatite has been frequently used as an indicator of alteration (Pate et al., 1994; Trueman, 1999; Smith et al., 2002). These elements, which substitute for other elements in bioapatite, are only incorporated in enamel and bones in relatively small concentrations during life from diet and drinking water. However, they can be considerably more concentrated in archeological and paleontological specimens due to the replacement of biogenic material and the infilling of voids after the loss of organics (Alonzi et al., 2019). In some cases, researchers compare the elemental composition of a specimen to the surrounding sediment or rock. If the composition of a specimen closely resembles that of its matrix (and diverges from values observed in modern bones or teeth), then the specimen is likely altered (Roche et al., 2010).

Carbonates and phosphates precipitate into gaps left by the loss of organics. They readily replace compounds in bioapatite both in life and post-mortem (Wopenka and Pasteris, 2005). This can change both the relative abundance of calcium and phosphorous (Ca/P) and CO₃ and PO₄. Ca/P in unaltered bioapatite is 2.2 based on stoichiometry (Sotiropoulou et al., 2015), but values ranging from 1.3 to 2.3 have been measured for modern human femoral heads (Zaichick and Tzaphlidou, 2002). Researchers typically assume that altered specimens exhibit Ca/P outside of this expected range (Pate et al., 1994; Berna et al., 2004; Zaichick and Zaichick, 2010). Researchers also generally assume that CO₃/PO₄ for unaltered bioapatite will be lower than 0.5 based on stoichiometry and observation. A lower bound is not well established (Piga et al., 2011; Dal Sasso et al., 2018). In theory, Ca/P and CO₃/PO₄ for a specimen should covary. However, this expected relationship breaks down if a carbonate other than CaCO₃ is precipitated (e.g., MgCO₃, FeCO₃). Further, the addition of secondary phosphate minerals, which generally also contain calcium, would lower CO₃/PO₄ while potentially maintaining or even increasing Ca/P.

Crystallinity measures the size of carbonate crystals within a specimen. Biogenic carbonate crystals are relatively small in comparison to the more stable diagenetic carbonates and apatites (e.g., fluorapatite) that can replace them (Posner, 1969; Shemesh, 1990; Surovell and Stiner, 2001).

Lastly, sometimes researchers use the δ¹³C or δ¹⁸O values of the specimens themselves to gauge if diagenetic alteration has occurred. The isotopic composition may be altered to reflect post-burial conditions (Zazzo et al., 2004b; Liu et al., 2008; Alonzi et al., 2019). In particular, extremely homogenous δ¹³C and δ¹⁸O values, or specimens with abnormally high or low isotope values, may be cause for concern (e.g., Koch et al., 1997; Sharp et al., 2000; Zazzo et al., 2004a). Carbon isotopes can be altered by deposition and replacement of carbonates (Koch et al., 1997; Lee-Thorp and Sponheimer, 2003; Trueman et al., 2003), and partial oxygen isotopic exchange can occur between bioapatite and groundwater early in diagenesis (Sharp et al., 2000; Zazzo et al., 2004a,b). Both of these processes are expected to decrease isotopic variability among specimens. However, because alteration may vary among specimens at a locality, some specimens may remain isotopically unaltered while others from the same locality may have abnormally high or low isotope values (Chillón et al., 1994; Zazzo et al., 2004a,b). Researchers typically rely on expected biology (diet and habitat) to determine if specimens with unusual δ¹³C or δ¹⁸O values have been isotopically altered (Zazzo et al., 2004b). It is also possible to use the offset in δ¹⁸O values between the carbonate and phosphate portions of bioapatite to assess preservation. The oxygen isotope composition of carbonate and phosphate can be impacted by different processes, leaving one of these compounds altered and the other untouched (Zazzo et al., 2004b; Martin et al., 2008).

All of the methods for checking preservation outlined above have some drawbacks due to the amount of sample required, the amount of time and training needed to conduct analyses, instrument availability, or the associated cost. Moreover, the instrumentation used to measure these preservation proxies is generally destructive to some degree. Researchers sometimes assess if specimens retain biogenic isotopic values by looking at the variability of the isotope values themselves using methods such as inductively coupled plasma mass spectrometry (ICP-MS). The most common analytical methods for measuring elemental abundance, including rare earth elements, halogens, and Ca/P, are benchtop XRF (bXRF), X-ray diffraction (XRD), and scanning electron microscopy (SEM). These methods, like hXRF, displace electrons and measure the energy difference that results to identify elements. Neutron activation analysis and flame atomic absorption spectroscopy were used to established the accepted range for unaltered Ca/P (Zaichick and Tzaphlidou, 2002; Roche et al., 2010; Sotiropoulou et al., 2015), but these methods are infrequently used in archeology and paleontology. Fourier Transfer Infrared spectroscopy (FTIR) is typically used to characterize CO₃/PO₄. This measures the intensities of carbonate and phosphate peaks. It can also characterize crystallinity because crystal size has a direct relationship with the relative intensities of peaks and valleys observed in FTIR spectra (Piga et al., 2011; Dal Sasso et al., 2018). These commonly used methods generally require powdered, homogenized samples or a small fragment of a specimen. FTIR can be used on whole specimens, but powdering samples aligns the orientation of the bioapatite crystals and gives considerably more accurate readings. Lastly, isotope ratio mass spectrometry (IRMS), which measures δ¹³C and δ¹⁸O values, also typically requires homogenization, and ultimately dissolution, of a sample. Laser ablation may also be able to measure elemental and isotopic composition. It is less destructive than other methods, but does cause a small amount of damage, takes a lot of time and resources, and has trouble accurately measuring some values, including oxygen isotopes (Passey and Cerling, 2006). It also only provides data about a tiny portion of a specimen, which will only capture a limited time window, and cannot remove any diagenetic material (Lewis et al., 2014; Limbeck et al., 2015; Willmes et al., 2016; Rey et al., 2022). To avoid sampling specimens with altered bioapatite, most researchers preferentially analyze tooth enamel (Cerling and Sharp, 1996; Lee-Thorp and Sponheimer, 2003). Enamel is more crystalline and has far less organic material than bone (roughly 5 vs. 30%), making it more resistant to alteration (Lear et al., 2008).

Benchtop XRF requires ~0.5 grams of homogenized sample pressed into a pellet with a binding agent such as cellulose powder, which prevents analyzing the same powdered sample for isotope data. Handheld XRF (hXRF) instead detects the composition of a specimen's surface (rather than homogenized powder; Byrnes and Bush, 2016) by exciting the specimen surface with primary X-rays and measuring the energy of the secondary X-rays emitted in approximately a minute. While FTIR does not have to be destructive, powdering sample randomizes the orientation of what is being measured. Unfortunately, even though only enough powder is needed to cover the detector eye (<10 mg) this powder can be difficult to recover without concern about contaminating samples prior to isotope analyses. In contrast to FTIR, the orientation of a specimen is not expected to matter with hXRF, giving hXRF the potential to assess the preservation of whole specimens without damaging their structural integrity. If there is a reason to measure powdered sample, hXRF also requires far less material than bXRF, and that powder is relatively easy to recover for further analyses, as we discuss below. The ability of hXRF to measure the surface of whole specimens, combined with the small size and portability of hXRF spectrometers, makes this method an attractive option for analyzing museum specimens that are rare or unique and may be difficult to move. Handheld XRF spectrometers also require minimal training and analytical time, and are inexpensive to use after initial purchase.

However, the capacity for hXRF to assess diagenetic alteration of fossil bones and teeth has not been previously tested. This tool was originally developed and used for in situ detection of the presence of uranium and other minerals in mining operations and other industries, but has also been used to identify metals and minerals and expanded into other arenas, including soil and waste analysis, art conservation, and food safety studies (Bosco, 2013). It has also been used to assess the burial history of modern human remains by measuring the composition of matrix on bone surfaces (Byrnes and Bush, 2016) and evaluate diet by measuring elemental abundance of tooth enamel (Hamilton et al., 2019, 2022). We suspect that it will be similarly useful for assessing alteration of fossil specimens. If Ca/P_surface (measured with hXRF) and other widely used methods, such as CO₃/PO₄ and δ¹³C and δ¹⁸O values, are comparable in their diagnoses of altered specimens, then measuring Ca/P_surface could reduce the time and cost associated with evaluating preservation and help prevent unnecessary damage to specimens that are too altered to have biogenic isotopic data.

We used a Bruker Tracer 5i hXRF spectrometer to measure Ca/P_surface, Ca/P_subsurface, and Ca/P_powder in bones and teeth. The spectrometer was set to the mudrock duo calibration, and data were collected and viewed using the Bruker Instrument Tools program 1.7.0.128. For measuring Ca/P_surface, a flat surface of each specimen was scanned while in direct contact with the protective window over the detector of the hXRF spectrometer. When specimens were small enough to rest on the spectrometer's platform (4 inch-squared), the spectrometer was attached to a stand that directs the X-ray source and detector toward the ceiling. Specimens were placed directly on the platform over the detector window, and when possible, a lead cup was placed over them. If specimens were unable to be placed on top of the platform due to their large size or arrangement in a museum exhibit, they were set on a flat surface, and the hXRF spectrometer was removed from the stand and held by hand, directed downward.

Measuring Ca/P_powder required some modifications to the setup described above (Figure 2). Powdered samples are generally measured with an hXRF spectrometer by placing them in plastic sample cups and scanning them through 4 μm-thick Prolene sample support windows. To reduce the amount of powdered enamel needed, we placed powdered sample between two Prolene sample support windows. This assemblage was then backed by a single, open-ended sample cup containing powdered cellulose (Figure 2). Having powdered cellulose behind the sample while scanning helped better cover the detector of the spectrometer, preventing the laser from scattering and allowing for more accurate measurements. The cellulose itself does not have heavy enough elements in high enough concentrations to be measured by hXRF, and thus did not interfere with measuring our samples. We found by iteratively measuring smaller amounts of an internal reference material (powdered bentonite) that spectrometer precision decreased for samples smaller than 250 mg. We therefore worked with 250 mg for powdered enamel samples.

We measured CO₃/PO₄ of untreated powder on a Nicolet™ iS™ 10 FTIR Spectrometer with OMNIC 9 software at the Chemistry Department at Northern Kentucky University. All values were measured relative to a baseline based on local minima for each peak, but additional spectra manipulation, such as peak deconvolution, was avoided (Supplementary Table 4). To characterize CO₃/PO₄, we divide the sum of the carbonate peaks at 1,400 and 1,500 nm⁻¹ by the sum of the phosphate peaks at 1,000 and 604 nm⁻¹. The 565 nm⁻¹ phosphate peak is also frequently used to characterize CO₃/PO₄ (Dal Sasso et al., 2018); however, it was not considered in this analysis as the spectrometer was less accurate near and below 550 nm⁻¹.

Lastly, we measured δ¹³C and δ¹⁸O values in pretreated enamel carbonate on a Finnigan-MAT 252 isotope ratio mass spectrometer (IRMS) equipped with a Thermo Scientific Kiel II carbonate device at the Light Stable Isotope Mass Spec Lab at the University of Florida. We report both δ¹³C and δ¹⁸O data relative to Vienna Pee Dee Belemnite (VPDB). Data were calibrated using the international standard NBS-19. Precision (based on 8 replicates of this standard), was 0.024‰ for carbon and 0.049‰ for oxygen. Accuracy was confirmed using NBS-18. Corrected values for δ¹³C (−5.01±0.02%) and δ¹⁸O (−23.08±0.02%) were very close to accepted values (−5.01% and −23.2%).

In Part 1, means and ranges for various groups were visually compared to the expected range for unaltered Ca/P. We used Wilcoxan matched pairs signed ranks test to compare Ca/P_surface for bone and associated teeth from individual mandibles.

In Part 2, we used Kruskal Wallis tests followed by Dunn Bonferonni post-hoc comparisons to compare Ca/P_surface, Ca/P_subsurface, and Ca/P_powder for specimens identified as altered and unaltered based on Ca/P_surface, as well as δ¹³C and δ¹⁸O values among the four alteration groups (both altered Ca/P_surface and CO₃/PO₄, unaltered Ca/P_surface but altered CO₃/PO₄, altered Ca/P_surface but unaltered CO₃/PO₄, and both unaltered Ca/P_surface and CO₃/PO₄). We used linear regression to determine if a relationship exists between Ca/P_surface and CO₃/PO₄ and Levene's tests for homogeneity of variance to compare variances in δ¹³C and δ¹⁸O values among alteration groups. All statistical tests were run in SPSS 25 with significance set at a = 0.05.

Most of the 30 points scanned on the modern human femora had Ca/P_surface within the expected range of 1.3–2.3 (Figure 3, Supplementary Table 1). Measurements taken using the Brucker Tracer 5i used in this study generally have a 3σ limit of < 1%, and reported values for calcium and phosphorous were always >20,000 ppm, with a margin of error below 800 ppm, in this study. However, several points had values ranging from 2.3 to 2.8, especially those that were near metaphyses. Among the other modern human skeletal elements scanned, one cranium and one talus had Ca/P_surface >2.3 (2.4 and 2.6, respectively). All of the non-human specimens (both bones and teeth) had Ca/P_surface between 1.3 and 2.3 (Figure 3, Supplementary Tables 2, 3). There were no differences in Ca/P_surface between mandibles and their associated teeth (N = 8 individuals, W = 326, DF = 15, p = 0.9; Supplementary Table 5).

Ca/P_surface was higher than Ca/P_subsurface or Ca/P_powder for both altered and unaltered specimens (altered Ca/P_surface: χ² = 22.2, df = 2, p < 0.001; unaltered Ca/P_surface: χ² = 8.0, df = 2, p = 0.02). With the exception of two specimens (2022-1950-57b and 2022-1950-23), which had Ca/P_subsurface values of 2.5 and 2.4, respectively, all specimens had Ca/P_subsurface and Ca/P_powder below 2.3, regardless of the original Ca/P_surface value (Figure 4, Table 1).

The relationship between Ca/P_surface and CO₃/PO₄ was weak and insignificant (R² = 0.1, p = 0.3; Figure 5). Of the twenty teeth scanned, eight were identified as altered and six as unaltered by both Ca/P_surface and CO₃/PO₄; four were identified as altered by CO₃/PO₄ despite being identified as unaltered by Ca/P_surface; and two were identified as altered by Ca/P_surface but unaltered by CO₃/PO₄ (Table 1, Figure 5). There were no significant differences in median δ¹³C or δ¹⁸O values among alteration groups (δ¹³C: χ² = 5.7, df = 3, p = 0.1; δ¹⁸O: χ² = 1.7, df = 3, p=0.6). There were also no differences in isotopic variance among these groups (Levene's test p > 0.05 for both isotopes). Nevertheless, there were apparent trends. Specimens identified as unaltered by both Ca/P_surface and CO₃/PO₄ had apparently higher, more variable δ¹³C values, and similar but apparently less variable δ¹⁸O values, than specimens identified as altered by both proxies. Specimens with altered Ca/P_surface but unaltered CO₃/PO₄ had a similar range in δ¹³C and δ¹⁸O values to specimens identified as altered by both proxies. Conversely, specimens identified as unaltered by Ca/P_surface but altered by CO₃/PO₄ had a similar range in δ¹³C and δ¹⁸O values to specimens that were identified as unaltered by both proxies (Figure 6).

To avoid the influence of taxon, we also compared isotope values among alteration groups for just bison. Using this smaller dataset (N = 12), isotope values among alteration groups remained statistically indistinguishable (δ¹³C: χ² = 1.5, df = 3, p = 0.7; δ¹⁸O: χ² = 3.6, df = 3, p = 0.3; Levene's test p > 0.05 for both isotopes). Apparent patterns in δ¹³C and δ¹⁸O values among the alteration groups were also somewhat similar to the whole dataset. Specimens identified as unaltered by both Ca/P_surface and CO₃/PO₄ had apparently more variable δ¹³C and less variable δ¹⁸O values than specimens identified as altered by both proxies (Figure 6). Specimens with unaltered Ca/P_surface and altered CO₃/PO₄ had similar trends in δ¹³C and δ¹⁸O values to specimens identified as unaltered by both proxies, but the range in δ¹³C values was higher. The single bison with altered Ca/P_surface but unaltered CO₃/PO₄ had a similar δ¹³C value to most of the specimens identified as altered by both proxies and a low δ¹⁸O value (comparable to one of the bison identified as altered by both proxies).