Introduction

Despite its importance in the evolution of the human diet, direct plant consumption by humans is underrepresented in the archaeological record because plants often have much less protein and lipids than animals that were consumed by humans1,2. Developing innovative proxies can improve the accuracy and resolution of archaeological reconstructions, reveal previously overlooked aspects of plant use, and expand our knowledge of cultural and environmental changes over time. Ultimately, these advances contribute to a more comprehensive picture of how humans interacted with and depended on plant resources throughout history.

The use of plants in human diets has been primarily studied through archaeobotanical methods that are very limited to plant processing and taphonomy, hindering the preservation of plants3,4. For example, the intensity of plants exposed to fire and oxygen levels during charring strongly influence plant preservation, recovery and species identification5,6. In most past human activities, however, the economically important plants perish after processing or consumption, and their use is simply not recognised. Recent application of molecular methods opened the possibility to recover and identify ancient plant proteins that were consumed in the past via paleo proteomics analysis7,8. However, authors in the past had struggled to establish the presence of taxonomically specific plant proteins or other plant biomarkers.

The importance of the use of broomcorn millet (Panicum miliaceum) as a crop has received immense attention due to the impact it has had on past communities and its potential importance as one of the future superfoods9. Broomcorn millet utilises the C4 photosynthetic pathway and has more enriched δ13C values, as it discriminates less against the carbon isotope13C than C3 plants during carbon capture from the atmosphere; this leads to a substantial difference between the ratios of heavy and light carbon isotopes in C3 and C4 plants10,11. A relative proportion of isotopes from the plants eaten is absorbed into animal tissues and can later be directly measured using isotope ratio mass spectrometry. This allows for detecting millet consumers down to the individual level. However, for millet to be detected isotopically in human tissue, millet needs to make up more than 20% of the overall dietary protein consumed12. Therefore, many millet consumers can be overlooked. Thus, we lack the proxies that will allow the identification of individuals consuming low levels of millet, which would otherwise not be detectable by stable carbon isotope analysis of skeletal remains.

Fortunately, broomcorn millet possesses another unique chemical property that can be measured and tracked through time. Broomcorn millet contains pentacyclic triterpene methyl ether (PTME) called miliacin. It is the principal (c. 99%) PTME in broomcorn millet and is absent in other commonly cultivated species13. The pentacyclic triterpene miliacin is an anti-microbial compound and is resistant to decomposition by bacteria13,14,15. The durability of miliacin results in its survival in high concentrations in archaeological contexts. This specific biomarker has been identified in ancient agricultural fields, sediment cores16,17,18,19 and ceramic matrices of archaeological pottery vessels in Europe, as well as East and Central Asia20,21,22,23,24. Yet, the identification of the miliacin biomarker in human dental calculus has never been previously attempted, as the miliacin extraction method required too much sample material (2–5 g used in previous studies); the average total dental calculus weight per adult human is 70.2 mg25. Over the past years, methodological advancements in analytical technique have significantly reduced the sample size required for detection by mass spectrometry (e.g. less than 1.0 mg of sample material is usually sufficient), indicating that miliacin archived in human dental calculus could be successfully detected and analysed.

Miliacin is highly abundant in millet seeds, and human dental calculus has a high potential to trap these biomolecules during consumption. Therefore, the analysis of miliacin from human dental calculus will provide a molecular source to trace a wide range of millet consumption intensities and will be linked together with the obtained stable isotope values and archaeological data. Biomarkers recovered from human dental calculus hold immense potential for studying past human diet and physical health26,27. This research presents a novel methodology for identifying the miliacin in human dental calculus through a case study on the medieval Ostriv cemetery in Ukraine, as the Ostriv population possessed extensive quantities of human dental calculus on their teeth.

To overcome the sample size issue, we adopt the thermal desorption–gas chromatography–mass spectrometry (TD-GC/MS) as the analytical method. This method has been used not only for archaeological ceramics28 and mummies29, but also for dental calculus from the Mesolithic population30 and further Neanderthals31. Critically, this method requires much less sample mass than conventional methods with lipid extraction and a much shorter time for the preparation of analysis. Moreover, miliacin was detected using this method from an archaeological pottery sherd28, opening the possibility for the detection of this compound in different materials such as dental calculus.

Simultaneously, stable carbon and nitrogen isotope analysis of dentin samples from the same individuals was conducted. By doing this, we aimed to confirm the novelty of TD-GC/MS method to detect minute millet consumption compared to the conventional isotopic methods. Additionally, while dental calculus is formed during adulthood, dentin is formed during childhood. This provides a good opportunity to compare the detection of millet biomarker in the former, and their tendency of dietary intake between C3/C4 in the latter at different stages of life.

Results

Thermal desorption – gas chromatography – mass spectrometry

Miliacin was detected in the eight dental calculus samples (MIL1_49, 52, 54, 57, 59, 65, 66 and 71), confirming that these humans directly consumed broomcorn millet (Table 1). Miliacin was not detected in the simultaneous blank control that was processed and analysed with dental calculus samples in the same setting. In addition to miliacin, free saturated fatty acid methyl esters (FAMEs) such as those of palmitic and stearic acids, cholesterol and its derivatives, and Dimethyl phthalate, which is probably from plasticiser of containers used during the process of excavation, curation and storage32 were detected (Fig. 1). As FAMEs and cholesterol are ubiquitous in a broad range of foods, it is difficult to determine the origins of these fatty acids and cholesterol derivatives. Because of the relatively lower concentration of miliacin, it was not possible to quantify this compound’s abundance in the samples. In addition, an archaeological ceramic powder sample containing miliacin (MJR10, Heron et al. 2016) and the authentic miliacin reference substance (PhytoLab Certificate Report No. 74485154- 99 002) were analysed in the same conditions, confirming the retention time and mass spectrum of miliacin (Fig. 1).

Fig. 1
figure 1

(A) Typical partial TD-GC/MS chromatogram of dental calculus sample (MIL1_65) obtained by scan mode showing FAMEs and cholesterol derivatives, (B) by SIM mode showing miliacin. (C) partial TD-GC/MS chromatogram of archaeological ceramic sample with miliacin (MJR10: Heron et al. 2016) obtained by scan mode, PB: Pentamethylbenzene, HB: Hexamethylbenzene, and (D) by SIM mode. (E) Partial TD-GC/MS chromatogram of an authentic sample of miliacin (PhytoLab Certificate Report-No. 74485154- 99 002) obtained by scan mode. (F) Chemical structure of miliacin.

Full size image
Table 1 Summary of the results of the dental calculus samples analysed in this study. FACx: y: fatty acids with carbon length x and number of unsaturation y, chol: cholesterol and its derivatives, DP: dimethyl phthalate.
Full size table

Stable isotope analysis in dentin

Stable carbon and nitrogen isotope analysis of the M2 dentin sample was conducted on 31 individuals, as they had larger quantities of dental calculus. As the dental calculus forms in adulthood, but dentin represents a childhood diet, stable isotope analysis aimed to understand dietary habits during the early stages of an individual’s life and potentially identify C4 plant consumers and those for whom the C4 diet is not visible. The δ13C in all analysed humans ranges from − 15.0 to -20.8‰, with a mean value of -18.5‰. In δ15N from 10 to 8.5‰, the mean value is 9.4‰ (Table 2). If the δ13C cut-off value between predominantly C3 and mixed C3 and C4 consumers − 18‰33,34, two dietary groups of individuals that consumed C4 plants during childhood and the ones that had no or limited intake of C4 plants could be classified.

Discussion and conclusions

Applying the TD-GC/MS method to human dental calculus confirms millet consumption by the medieval populations in Ukraine, supported by the detected miliacin from eight samples out of 31 individuals. Indeed, the Ostriv population had dense layers of dental calculus covering their teeth, some of which caused severe periodontal conditions (Fig. 2).

Fig. 2
figure 2

The severe periodontal disease in grave 68 of the Ostriv cemetery site, caused by depositions of dental calculus, where a trace signal of miliacin biomarker was identified (photo by A. Kozak).

Full size image

Out of eight individuals with miliacin, six of them were female—one died at the age of 20–25, one at 20–30, one at 25–30, one at 25–35, two at 30–40, and one at 40–50. Male individuals were identified as aged 20–30 and 30–40 (Table 2). Interestingly, more depleted carbon isotope values less than − 18‰ were observed in the adults (20–30, 30–40, and 40–50, Table 2). This might indicate that some children with less C4 intake became millet consumers later on in their adulthood, although currently it is difficult to evaluate the amount or intensity of millet consumption solely based on the miliacin detection in dental calculus.

Some examples of intensive C4 intake in childhood have previously been revealed in the Karatuma site in Kyrgyzstan35 and the Taksai site in Kazakhstan36 based on the bone collagen isotope analysis. The results from our study are beginning to reveal a different pattern whereby millet is more intensely consumed in adulthood. However, further evidence is needed to further confirm this pattern. Ostriv burials constitute two distinct populations of local origin and of potentially Baltic origin. According to multiple lines of evidence that include aDNA and artefact typology, the Baltic migrants moved to the vicinity of Ostriv and were buried there between 10 and 12 c CE37. As millet was less common in northern Europe than in Southeastern Europe at the time, it is likely that the isotopic data of δ13C in dentine, which represent childhood diet, were less positive in Baltic migrants. Millet became a part of the staple food of these individuals only upon their arrival in the present territory of Ukraine.

On the other hand, the absence of miliacin in dental calculus does not necessarily mean that the individual did not consume millet due to potential preservation biases. Therefore, the frequency of millet consumption could be underrepresented, as more research is needed to understand the underlying circumstances and taphonomy of miliacin preservation in human dental calculus. Nevertheless, miliacin was detected in both males and females, as well as different age groups, without showing a correlation with a specific age group (Fig. 3), indicating that millet consumption itself was quite popular among these populations, although there should be differences in the intensity of consumption.

Fig. 3
figure 3

(A) Scatter plots demonstrate the distribution of millet consumers against the dentin’s stable carbon and nitrogen isotope values of the Ostriv humans. The cut-off value of −18‰ in δ13C (dotted line) separates predominantly C3 and mixed C3 and C4 consumers33,34. Square symbols stand for females, while circle symbols represent males. Yellow: sample with miliacin. (B) Histogram of miliacin detection rates by age group. Group a: anthropologically identified as 13–15 years old, 16–18, and 18–25. b: 20–25, 20–30, and 25–30. c: 25–35, and 30–40. d: 35–45, 40–50, and 45–50, e: 45–55, 55–60, and 55–65. Black: total number of individuals, yellow: number of individuals with miliacin.

Full size image

It is also tempting to compare the pathological features of these bones with the presence/absence of miliacin, but due to the sample size, it is difficult to find any statistically meaningful relationship among them. Therefore, reconstructing the foodways of millet consumption in these populations requires further research. Nonetheless, most importantly, this study successfully detected the millet-specific biomarker miliacin in human dental calculus for the first time.

The presence of miliacin in eight individuals provides direct molecular evidence for broomcorn millet consumption in medieval Ukraine. The detection of miliacin in individuals with more depleted δ¹³C values, typically less than − 18‰ in δ13C in dentin, demonstrates that small quantities of millet, undetectable through stable isotope analysis, can now be identified. This shows that conventional isotope analysis which considers the value − 18‰ as a threshold of millet consumption, underrepresents millet consumers, particularly those who consumed it intermittently or in minor amounts. The successful extraction of miliacin from microgram-scale dental calculus samples demonstrates the high potential of this matrix for recovering plant biomarkers. This opens new methodological avenues for studying low-level or occasional plant consumption in past human diets across different chronologies. Moreover, this potentially provides a way to compare the diet between childhood and adulthood, given the different formation period between dentins and dental calculus during their lives.

The use of thermal desorption GC/MS (TD-GC/MS) in this context demonstrates a micro-destructive, efficient, and scalable method for detecting specific plant compounds in archaeological dental calculus. This method can be extended to other chronologies, crops and contexts, contributing to more inclusive reconstructions of ancient plant parts in diets. This methodological breakthrough will aid future research in tracking the spread, intensity, and social context of millet and other underrepresented crops in human history.

Table 2 The isotopic values of dentin, age, and sex of the analysed individuals from the Ostriv cemetery.
Full size table

Materials and methods

Archaeological site Ostriv and material selection

The medieval Ostriv cemetery is located c. 100 km south of Kiev in Ukraine. The burial ground of Ostriv is currently being investigated in the framework of a research grant by the German Science Foundation (DFG) at the Leibniz-Zentrum für Archäologie (LEIZA Schleswig) in cooperation with the Archaeological Institute of the National Academy of Science in Kyiv. It consists of Kievan Rus’ inhumation graves, which are dated from the 10th to the 12th century, with a maximum in the 11th century AD. This burial site became the focus of this study due to a few reasons: a) previous research by Shiroukhov and colleagues37 showed the possible consumption of C4 plants as reflected in isotope values of bone collagen from a few graves at Ostriv site; b) millet is known to be used extensively in Ukraine during this period that provided higher chance in finding miliacin biomolecules and developing methodology; c) the population there contained huge quantity of dental calculus on their teeth that was an ideal material for further biomarker miliacin analysis; d) as this research aimed at developing methodology of miliacin extraction from dental calculus, we targeted more recent population for higher chances in finding miliacin.

The population in Ostriv burial was a part of the Kievan Rus’ state that existed from the late 9th to the mid-13th century and encompassed a variety of polities and peoples38. It constituted a crucial link between Scandinavia, Baltic and Byzantium via the Varangian trade routes, influencing the development of later East Slavic identities. The archaeological complex of Ostriv, discovered in 2017, presents a unique site characterised by burial practices and artefacts common not only to the local Slavic population. An analysis of the grave goods, such as flat ladder brooches, alongside burial orientations, suggests connections with Baltic tribes, including the Prussians, Curonians, Scalvians, and Yotvingians. The multiproxy analysis of the population buried at this site suggests a multi-ethnic site with Baltic migrants and local Slavs37, indicating human mobility during the Kievan Rus’ period.

Ostriv cemetery probably belonged to the settlement, situated on the opposite bank of the river Ros’. The settlement was included to the south fortification line, which separated Rus’ from the nomadic Steppe. The primary economic activities in the southern regions of Rus encompassed hunting, cattle breeding, and, to a lesser extent, agriculture. Paleobotanical studies conducted on materials excavated from several sites in the Porosia regions have demonstrated a predominance of crops such as rye, several types of wheat39,40, peas, lentils, and millet, which was an important and widespread crop in this region41. In the city of Yuriev, which was also part of the Porosian defence line, archaeobotanists identified the presence of wheat and millet among the crops40,42,43.

Thermal desorption – gas chromatography – mass spectrometry

By modifying the established method for archaeological ceramic powder28, 31 dental calculus samples (Table 2) together with a blank control sample were analysed using thermal desorption–gas chromatography–mass spectrometry (TD-GC/MS). Powdered calculus samples (ca. 1–15 mg depending on the availability) were embedded in a sample cup to which 10 µl of methyl derivatisation reagent (Tetramethylammonium hydroxide (TMAH) 25 wt% in methanol) and 10 µl of the internal standard n-hexatriacontane dissolved in n-hexane (1 mg ml− 1), internal standard (n-hexatriacontane) was added. The sample was analysed by Thermally assisted hydrolysis and methylation thermal desorption (THM TD-GC/MS), using a multi-shot pyrolyser (EGA/PY-3030D, Frontier Laboratories Ltd, Koriyama, Japan) connected with GC/MS (GCMS-QP2010Ultra, Shimadzu Corporation, Kyoto, Japan). The sample cup was released in the pyrolyser, which was set to 550 °C for 0.20 min, then introduced to the GC at 300 °C by splitless mode. The Ultra ALLOY+-5 (30 m − 0.25 mm − 0.25 μm, Frontier Laboratories Ltd) column was used at a flow rate of 3.0 mL/min of helium gas and analysed at a split ratio of 1/20. FAAST (Fast Automated Scan/SIM) mode was adopted to detect miliacin efficiently while obtaining other major compound profiles with limited sample mass. For scanning, the temperature was set at 50 °C, then raised by 10 °C/min until it reached 325 °C, where it was held for 5 min. The spectra were obtained between m/z 50 and 600. For SIM (Selected Ion Monitoring), m/z 189, 204, 231, 425, and 440 for the detection of miliacin were monitored between 27.0 and 31.3 min. The total acquisition time was 32.5 min. The Electron Ionisation (EI) energy of the MS was 70 eV. The temperature at the interface was kept at 300 °C, while the ion source was kept at 230 °C. Details of the Py-GC/MS instrumental settings in each section are shown in Figure S1.

Stable isotope analysis of human dentin collagen

The dentin in 31 individuals that also contained dental calculus subjected to thermal desorption were analysed for stable isotopes. The analysis was done on M2 human teeth. The second molar begins to calcify at 2.5–3 years and mineralisation of the crown is complete by 7–8 years of age44. Collagen extraction took place in the Bioarchaeological Research Centre at the Department of Archaeology, Faculty of History, Vilnius University, following protocol as described in the previous study45. Hydrochloric acid of 0.5 M was added to the samples until the inorganic component of the bone dissolved, and any debris was eliminated by multiple rinses with ultrapure water. Collagen was gelatinised in a pH 3 solution at 70 °C for 48 h, frozen overnight in a -30 °C freezer, and then freeze-dried for at least 48 h. The collagen carbon and nitrogen stable isotope ratios were measured using an elemental analyser (Thermo Fisher Scientific, Flash EA 1112), which was connected to Isotope ratio mass spectrometer, EA-IRMS (Thermo Fisher Scientific Finningan Delta Advantage, Bremen, Germany) at the Centre for Physical Sciences and Technology in Vilnius, Lithuania (FTMC). Graphite USGS24, caffeine IAEA-600, ammonium sulfate IAEA-N1 and IAEA-N2 were used as calibration standards. High purity caffeine (99,7%, Alfa Aesar) was used as an internal laboratory standard.

Stable isotope data are presented as delta values and expressed in ‰, relative to the international standards V-PDB (Vienna Pee Dee Belemnite) and atmospheric air for carbon isotope values and nitrogen isotope values, respectively. To eliminate the possibility of contamination of diagenetic processes and ensure that the isotopic values represented the dietary signals, collagen quality indicators, such as collagen yields, the C/N atomic ratio, carbon weight%, and nitrogen weight%, were assessed. Precision (u(Rw)) was determined to be ± 0.16‰ for δ13C and ± 0.13‰ δ15N on the basis of repeated measurements of calibration standards. The total analytical uncertainty was estimated to be ± 0.20‰ for δ13C and ± 0.17 for δ15N.