Introduction

Paleodietary studies are essential to reconstructing human lifeways from the past, with stable isotope analysis as a commonly applied method to study the human diet, mobility and environments1,2,3,4,5,6,7. Stable isotope values of the bio-elements (δ2H, δ13C, δ15N, δ18O, and δ34S) in body tissues are linked to the composition of an individual’s diet. Through δ13C values, we can see the content of C3 versus C4 plants as well as indications of marine versus terrestrial sources in diets, and δ15N values are related to the quantity and origin of the animal protein in the diet (meat, dairy products, freshwater or marine fish), and can also be influenced by the type of plant consumed. The trophic enrichment between the herbivorous fauna and omnivore humans is ca. 1–2‰ for δ13C, and 3–6‰ for δ15N8,9,10. Additionally, δ34S can provide evidence of marine influences (sea products, sea-spray effect, sediments of marine origin). The trophic shift of sulfur isotope ratios between diet and consumers tissues is known to be low11.

Different biological tissues reflecting distinct phases of an individual’s life can be selected for analysis, such as teeth, bones, hair or nails4,12,13,14. During the ontogenesis of permanent teeth, dental enamel and dentin form in childhood15 and remain unchanged throughout an individual´s lifetime16. Secondary dentin forms after root formation and continues to develop throughout life, though at a slower rate than primary dentin, while tertiary dentin is produced in response to injury or decay17.

Consequently, stable isotope data from human teeth can provide information on where individuals spent their childhood and adolescence and whether their diet changed from childhood to adulthood18,19.

Unlike dental tissues, human bone growth is completed in early adulthood, after which bone remodeling occurs. The main component of the organic fraction of dentin and bone is collagen. Its annual remodeling rate in the femoral bone of a 25-year-old adult male averages 2% and a time period of 30–35 years is assumed for its complete remodeling20. Compared to the femur, ribs have a higher proportion of cancellous bone with a higher turnover rate, estimated to be at around 6% per year. Therefore, ribs can be used to determine the average composition of an individual’s diet in a more recent period prior to death21, whereas the femur could be used to look at diet over a longer range of time.

Stable isotope analysis and ancient proteomics

Stable isotope analysis has been extensively used in archaeology to reconstruct ancient and historical diets. However, not all food components can reliably be inferred from the stable isotope data. In particular, the interpretation of the δ15N values regarding the proportion and origin of animal protein in the diet can be challenging9. There is a lot of δ15N data on archaeological bone, but the offset between food and body tissues in human nitrogen isotope values is not clearly defined and the interpretation of the values is therefore not without controversy10. When estimating the intake of animal protein and distinguishing between terrestrial and aquatic food, the known values of the potential food animals are of great importance. However, in the case of rather similar values for terrestrial animals and riverine fish, it is difficult to estimate the quantities of the respective food components.

Marine fish can be recognized as dietary components through elevated δ15N and δ13C values, but freshwater fish can be more difficult to detect. Due to the complexity of the food webs in inland waters, freshwater fish δ15N and δ13C values can be highly variable22,23. Moderate consumption of lower trophic level fish would not be detected by isotope analyses, and dietary models indicate that a terrestrial based diet could include up to 20% marine protein without raising bone collagen δ 13C values24,25,26. In addition, estimating the dietary contribution of plant proteins, especially from protein-rich legumes, is challenging, as they can only be tracked to a very limited extent27.

To fill in this gap, biomolecular analyses such as ancient DNA (aDNA) and proteomic studies have been used to provide a more detailed picture of ancient dietary preferences, as they can often identify the organisms that were consumed to the genus or even species level28,29. Ancient proteomics is a relatively new field of research having an enormous potential to reconstruct ancient human lifeways30,31,32. Due to their chemical structure and abundance in biological organisms, proteins often preserve longer than aDNA in many archaeological contexts31,33 and can be extracted from various archaeological remains (e.g., dental calculus, bones, ceramics).

Within this framework, dental calculus is an important reservoir for both ancient proteins and DNA. Researchers have been able to identify and analyze proteins from various sources within this biological tissue, including proteins of the human host and the numerous bacterial species that comprise the human oral microbiome to dietary proteins ascribed to animal and plant species that were consumed28,34,35,36. In particular, dairy proteins have been most commonly recovered from samples with variable chronological age and geographical provenance, though most studies have been restricted to Holocene individuals28,36,37,38. Dietary findings have been especially compelling as proteins allow us to differentiate both the tissues and taxa of animals that were preferably consumed. These data can be used to investigate several types of archaeological questions about human-animal interactions, especially in areas lacking in faunal remains34,36,37,38. While proteomic analysis of dental calculus has been successful in identifying proteins from a broad range of dietary sources, such as dairy products34,36,37,38, meat/blood28, or fish, plant proteins have been less abundantly reported (e.g. Hendy et al., 2018b). In contrast, plant micro-fossils in dental calculus are regularly observed with microscopy, indicating that some plant fragments can be preserved in this biological tissue39,40,41,42,43.

Although a rich array of analyses is available for studying ancient dietary habits, few studies have employed more than one approach38,44,45,46. In this study, we attempted to reconstruct historical dietary habits by using a two-pronged approach employing both ancient protein and stable isotope analyses. Ancient proteomics was applied to dental calculus samples while isotope analyses were performed on bone and dental collagen as well as dental calculus. Dental calculus has already been tested in the past as a suitable material for stable isotope analyses47, and although in some cases isotope signatures of dental calculus correlated with the values observed in bone or dentin48, they do not seem to be reliable for dietary reconstructions49. Here, we analyzed the stable isotope values in calculus as well as in collagen from dentin and bone, to test whether data obtained from dental calculus are related to those from collagen and thus can reliably indicate individual diet.

Dental calculus formation and composition

Dental calculus or tartar is a solid deposit that forms on tooth surfaces when dental biofilm (i.e., plaque) mineralizes50,51,52. Bacterial plaque is composed mainly of dead bacterial cells, bacterial metabolites, and proteins that are present in the saliva53,54. Plaque naturally forms on tooth surfaces, especially near salivary glands51,55. If not adequately removed through brushing, plaque can harden, forming tartar deposits. The mineralization process of plaque is possible through the deposition of calcium phosphate crystals from the saliva into the bacterial plaque51. The major component of calculus (around 80%) is calcium phosphate (e.g., hydroxyapatite, octacalcium phosphate, whitlockite, brushite), while calcium carbonate makes up around 3%52,56.

In addition to inorganic elements, there are also organic components (around 15–20%) entrapped within these deposits such as oral bacterial cells, human proteins and remnants of food57.

Tartar has proven to be an exceptionally valuable material in bioarchaeology as it can preserve micro-fossils, proteins and aDNA within the oral cavity of ancient individuals28,29,58.

Results

Stable isotopes results

The results of the collagen preparation and stable isotope analyses of the bio-elements carbon (C), nitrogen (N) and sulfur (S) in the dentin, bone, and dental calculus samples of the selected burials from Dalheim are shown in Table 1 and Fig. S1 online. The collagen yields of the dentin and bone samples ranged from 6.8–17.7% and 7.4–15.8% respectively (Supplementary Table S2 online). The high yield of collagen and its molar C/N ratios (3.1–3.2) show good collagen preservation and no evidence of chemical or microbiological degradation of the native protein59. Bone collagen δ13C values range from −20.4 to −19.7‰, with a mean of −20.0 ± 0.2‰, while δ15N values range from 9.3 to 11.0‰, with a mean of 10.0 ± 0.7‰ and δ34S values range from 8.3 to 10.3‰, with a mean of 9.4 ± 0.7‰. Dental collagen δ13C values range from −20.3 to −19.9‰, with a mean of −20.1 ± 0.1‰., while δ15N values range from 9,2 to 10.8‰, with a mean of 10.1 ± 0.5‰. Furthermore, δ34S values range from 6.5 to 9.6‰, with a mean of 8.6 ± 1.0‰ (Table 1). Fig. 1 and Supplementary Figs. S2S3 online show the δ13C and δ15N values per individual in bone and dental collagen and in dental calculus.

Table 1 Stable isotope results. bone collagen (n = 11); dental collagen (n = 8); calculus (n = 8).
Fig. 1
figure 1

Carbon and nitrogen isotope values of bone collagen, dental collagen and dental calculus of the individuals from Dalheim. Error bars represent the analytical precision based on repeated measurements.

The dental calculus samples show C/N ratios between 4.9 and 17.2 (Table 1). These samples differ from the collagenous samples, having mean δ13C, δ15N and δ34S values of −22.4 ± 1.2‰, 10.6 ± 1.2‰, and 6.8 ± 4.9‰ (Table 1).

The δ13 C values of dentin, calculus, and bone were significantly different (H = 16.067, p value < 0.001) (Table 2). There were not significant differences in the nitrogen and carbon values between dentin and bone samples. There were significant differences in the carbon values between bone and calculus samples (W = 80, p value < 0.001, while no significant differences were observed for the nitrogen values.

Table 2 Kruskal Wallis (H) test and Mann–Whitney-U (W) test coefficients. Bold values are significant at a level of 0.001.

Ancient proteomics results

During sample preparation, negative extraction controls (n = 7) were analyzed in parallel with the archaeological samples to check for contaminants in dental calculus. No dietary peptides were observed in any of the negative extraction controls or blanks. Moreover, general protein preservation was assessed using the Oral Signature Screening Database (OSSD) score published in Bleasdale et al. (2021) (Supplementary Dataset online: Table 1; Supplementary Text1B online). From the initial 52 samples, 37 passed the OSSD screening (Baar = 19; Dalheim = 18).

A total of 16 dietary proteins (34 unique peptides; 124 peptide spectrum matches—PSMs) were recovered from 15 individuals (Baar = 4; Dalheim = 11). Of those, 14 proteins are from plant sources and two are of animal origin (Fig. 2; Supplementary Dataset online: Table 2). The most representative plant family identified was Fabaceae (tribe Fabeae) and was only detected in the Dalheim group of individuals. Five proteins, all present in the seeds, were identified: vicilin, legumin A and J, convicilin, and p54 protein (Supplementary Fig. S5 online; Supplementary Text2B online). Seven unique peptides (24 PSMs) were specific to the tribe Fabeae, whereas 15 unique peptides (49 PSMs) were found to be specific to the species Pisum sativum (Fig. 2; Supplementary Dataset online: Table 2).

Fig. 2
figure 2

Histogram of taxonomic specificity of dietary peptide spectral matches per individual. Peptide count refers to the number of peptide spectral matches (PSMs) (Icons are from noun-project.com).

Cereals (i.e., C3 plants) and pseudo-cereals, such as millet (C4 plant), were also identified in both Dalheim and Baar based on three uncharacterized proteins and one plant development protein (NB-ARC domain-containing protein) (5 unique peptides, 21 PSMs). In the individuals from Dalheim, 14 PSMs (three unique peptide sequences) were ascribed to common wheat (Triticum aestivum), millet (Panicum), and rice (Oryza sativa). In the Baar group, 7 PSMs (with two unique peptides) were found to be specific to wheat and barley (Hordeum vulgare).

Five other proteins (two uncharacterized proteins and two proteins associated with plant metabolism) with 18 PSMs (and five unique peptide sequences) were ascribed to other plants (Fig. 2). In this category, there are peptides specific to domesticated plant species (e.g., spinach) but also uncharacterized proteins shared by various species belonging to the same group (i.e., Eudicots). Evidence of Pentapetalae (Eudicots) and Chenopodiaceae (Spinacia oleracea or Beta vulgaris) in the Dalheim population amounts to 16 PSMs. Within this plant group, two PSMs were specific to Spinacia oleracea (spinach).

In only one individual from Baar (Z147), four PSMs specific to Pentapetalae and two PSMs ascribed to the Fabids clade were attained.

The two animal proteins recovered (12 PSMs) included one milk protein and one protein derived from fish. A fish protein ascribed to the European perch (Perca fluviatilis) (one unique peptide, 10 PSMs) was present in four of the Dalheim individuals and in only one individual from Baar. Beta-lactoglobulin (β-Lg), a milk whey protein found in only one individual from Baar (one unique peptide, 2 PSMs), was taxonomically assigned to the family Bovidae (includes cattle and ovicaprines).

Four samples that did not pass the OSSD screening (G48, B89, Z19, Z145) also bear evidence of dietary proteins (31 PSMs) (Supplementary Dataset online: Table 2). The dietary peptides found in the samples with poor preservation of the oral signature (i.e., proteins found within saliva, belonging to both the host and oral bacteria) were 19 PSMs specific to Pisum sativum, 8 PSMs specific to Fabeae and 4 PSMs specific to Triticum aestivum.

Discussion

Biomolecular data can provide comprehensive reconstructions of past dietary habits. Isotope analysis is commonly applied when investigating past human diets, however, there are limitations when it comes to identifying specific food types. Other biomolecular methods, such as ancient proteomics and aDNA, can narrow down the range of the food types regularly consumed by past human populations28,29.

In addition, data collected during anthropological analysis (e.g., tooth wear and dental pathologies) can also provide indirect evidence of dietary habits60.

Stable isotopes

We performed δ13C, δ15N and δ34S analyses on tooth, bone, and dental calculus samples from 8–11 individuals from the Dalheim anthropological collection. Our results are consistent with the δ13C and δ15N data of an earlier isotope study done on bone samples of 19 adults of the same population61 (Supplementary Fig. S4 online). With a mean human bone collagen value of −20.0 ± 0.2‰ for δ13C and 10.1 ± 0.7‰ for δ15N, we could confirm that the Dalheim population mostly relied on C3 plants, possibly with a minor intake of C4 plants such as millet. In addition, this population might have consumed a fair amount of animal protein (terrestrial meats, dairy, or freshwater fish).

The comparatively low standard variations observed for the δ13C values (Table 1) point to a rather homogeneous diet within the group, but the range of δ15N values of 2.7‰ shows a certain difference in the individuals’ supply of animal protein. The relatively high δ34S mean value of 9.4 ± 0.7‰ indicates that the individuals’ diets were influenced by a marine signal62. This δ34S signal can be the result of the consumption of marine products but can also be associated with the consumption of plants and terrestrial animals living under the influence of evaporates. Dalheim is located on the Paderborn plateau, which is of marine origin in terms of its geological structure. Hence, the elevated δ34S values in the samples from Dalheim are not necessarily indicative of the consumption of sea fish. Frequent consumption of marine resources would also be reflected in higher carbon and nitrogen isotope values, which is not seen in these samples. On the other hand, higher nitrogen and lower carbon values would be expected in the case of a high consumption of freshwater fish63. The freshwater fish specimens may overlap in their nitrogen isotope ratios with the terrestrial omnivores, but their carbon isotopic range is much larger. As has been shown for sites in central and northern Europe, carbon and nitrogen isotope values of aquatic animals can be spread over a broad range64,65. Unfortunately, we are not aware of any isotope data for fish from the Dalheim region. Studies on fish from other regions show that C-N-S isotope data are highly variable and depend on the fish species, size, diet, habitat, etc.23,62.

However, a frequent intake of freshwater fish is also not evident from the values at hand. Overall, the stable isotope values obtained suggest that the population of Dalheim relied mainly on proteins from terrestrial animals in addition to plant proteins (mainly C3 plants), as observed by Olsen et al. (2018). If the individuals had consumed a lot of millet (C4 plant), higher δ13C values would be expected. The average δ13C value of C3 plants is −26.7 ± 2.3‰, while C4 plants such as millet have δ13C values that are around 14‰ higher (−12.5 ± 1.1‰)66. The metabolic fractionation factor from the food plant to the collagen of the consumer is approximately + 5 ‰67. The mean isotope values in the tooth and rib samples are very similar, indicating that the individuals’ diets during childhood versus adulthood did not vary substantially (Table 2, Supplementary Fig. S1 online). This might be seen as a contrast to the data from Olsen et al. (2018), in which juveniles (aged 7–12 years) had significantly lower δ15N values than younger adults (aged 20–40 years). However, most of the investigated teeth in our study were first molars, containing the information from an earlier period of life (3–9 years).

In our study, the isotopic values from the tartar samples are not directly comparable with the ones obtained from tooth and bone collagen (Table 2, Supplementary Fig. S1 online), as observed in other studies49. Here, the δ13C values of the calculus samples are distinctly lower compared to those of the dentin and rib samples. The mean δ15N value of the calculus (10.6 ± 1.2‰) is slightly higher compared to that of the collagen samples, but the single calculus values are highly variable. Furthermore, the extremely high differences in values between the single measurements of each calculus sample, especially for the δ34S values, indicate that the calculus material was very heterogeneous (Table 1).

In fact, unlike collagen, dental calculus is a highly heterogeneous material. Its composition reflects a diverse array of carbon sources, such as lipids and carbonates, each contributing distinct isotope values. The nitrogen isotope values are also variable, reflecting inputs from various food residues. Similarly, sulfur isotope values in calculus values can vary depending on the relative presence of sulfate and sulfide compounds.

It may also be, that homogenization of the calculus using mortar and pestle was not successful. As shown in other studies, when C/N values deviate from the elemental composition of collagen (> 12)48, carbon and nitrogen isotope values of the calculus are no longer positively correlated to those in bone collagen (Fig. 1, Table 2). High C/N ratios indicate that the carbon does not only come from the protein component of the dietary debris entrapped, but also from other organic (e.g., lipids, plant remains) as well as inorganic components (i.e. calcium phosphate and calcium carbonate)57,68. In view of the different structure and composition of each calculus sample, it is not surprising that there is no discernible relationship between the isotopic values of the collagen (bone and dentin) and the calculus deposits also at the individual level (Supplementary Figs. S23 online).

Proteomic data

Generally, there was a low presence of recovered dietary proteins within the calculus samples from the two sites investigated. Fifteen samples had to be excluded from our findings as their preservation of the oral signature was not sufficient (Supplementary Dataset online: Table 1). Nonetheless, the proteomic study provided evidence of dietary proteins specific to both plant and animal species. Overall, the recovered dietary proteins point to the presence of food types usually associated with medieval rural conditions69,70. At Dalheim, plant proteins and, to a lower extent, fluvial animal proteins were recovered, but no proteins from terrestrial animal sources (muscle, blood, milk) were identified. However, the consumption of animal products at the site, such as Sus sp., ovicaprines, cattle, and equids could be partly confirmed by previous studies through the analysis of food waste associated with the cemetery71. Stable isotope results (this study and Olsen et al., 2018) further confirm the consumption of terrestrial animal products. While the results of the isotope analyses did not provide clear evidence of fish consumption, proteins assigned to the European perch (Perca fluviatilis) were detected in the calculus of four individuals from Dalheim, suggesting the consumption of freshwater fish. Among the plant proteins belonging to the family Fabaceae, which were only found in two individuals from Dalheim, five of the proteins recovered (with 40 PSMs) were taxonomically assigned to green pea (Pisum sativum) (Fig. 2; Supplementary Text 2B online; Supplementary Fig. S5 online; Supplementary Dataset online: Table 2). Interestingly, all of these are globular proteins whose main function is seed storage. To date, archaeological peptides unique to green pea have only been reported in a few studies28,72 (Supplementary Table S3 online). Most of the archaeological peptides unique to P. sativum have been identified in organic residues scraped from the inner walls of ceramic vessels dated to the early Chalcolithic72 found at the Anatolian site of Çatalhöyük73 (Supplementary Table S3 online).

It is important to point out that in our study, plant peptides (i.e., legumes) were also found in three samples with low preservation of the oral signature, forcing us to be careful when interpreting the significance of such proteins. However, the previous documentation of starch grains belonging to the family Fabaceae within calculus deposits of the same collection confirms the consumption of pulses by the Dalheim population45. The authors also found starch grains ascribed to the tribe Triticaceae, this being consistent with our results45. In the same study, aDNA data allowed for the identification of Sus sp. (pig or boar), Brassica sp., and Triticum aestivum. Proteins ascribed to Triticum aestivum (wheat) and Hordeum vulgare (barley) were also recovered in this study, confirming the use of such cereals as staple foods. Millet (Panicum sp.), a C4 plant, was also detected in one individual from Dalheim. Plant proteins belonging to the clade Pentapetalae (Eudicots) and Chenopodiaceae (specific to both Spinacia oleracea and Beta vulgaris) were also identified. The clade Pentapetalae includes a wide range of flowering plants. Among the edible plants belonging to this clade, are the family Rosaceae, which includes most fruits, such as apple, pear, and berries, the family Fabaceae or legumes, the family Brassicaceae, which includes cruciferous vegetables such as broccoli/cabbage (Brassica oleracea), turnip (Brassica rapa), and radish (Raphanus raphanistrum), the family Apiaceae, which includes carrot (Daucus carota), and celery (Apium graveolens) and the family Asteraceae, which includes lettuce (Lactuca sativa) and chicory (Cichorium intybus). There are many archaeobotanical records of these plants having been cultivated in gardens and used for cooking in medieval times74,75.

In the early medieval population of Baar, the dietary proteins detected within dental calculus deposits were associated with plants, ruminant milk, and fish. The plant taxa identified were the clades Pentapetalae and Fabids. Proteins assigned to animal resources were a whey protein (β-Lg) taxonomically assigned to the family Bovidae (cattle and ovicaprines) and proteins specific to freshwater fish (Perca fluviatilis). Unfortunately, no food waste was associated with the site of Baar, so no zooarchaeological data is available. However, leather and fur textile remains are indirect evidence of the animals present at the site and possibly exploited for their meats or secondary products. In some instances, these fragments of textiles were ascribed to ovicaprines76. Among the amulets recovered at the cemetery, a bear tooth pendant was also found77, but this is most likely not linked to the consumption of bear mea

Medieval diet reconstructed through a dual approach

The use of a dual or multi-proxy approach in the study of bioarchaeological remains has the potential to refine paleodietary reconstructions (Table 3). When taxonomic-specific analyses (i.e., aDNA, proteomics) are coupled with stable isotope analyses, a thorough reconstruction of ancient and historical dietary habits is possible.

Table 3 Summary table of previous studies investigating medieval diets through multi-analytical approaches applied to medieval dental calculus, bone, and ceramic vessels.

Stable isotope analysis, particularly of carbon (C), nitrogen (N), and sulfur (S) isotopes, provides critical insights into the types of plants and animals consumed. Our isotopic data revealed significant consumption of C3 plants (i.e., grains, legumes, most fruits, and vegetables). This is in line with the observation that C3 plant-based diets were prevalent in Central Europe during the Middle Ages, in the Dalheim population (this study and Olsen et al., 2018) as well as in other medieval populations4,43,46,79,80,81 (Table 3). Furthermore, the diet of the populations analyzed contained terrestrial animal proteins, with evidence suggesting intake of dairy products. More specifically, the nitrogen isotopic values indicated that the Dalheim population consumed a considerable amount of animal products (i.e., meat, eggs, dairy).

The proteomic analysis of dental calculus deposits from the medieval sites of Baar and Dalheim also identified several C3 plants, including their adscription to the genus/species level. The taxa identified were common wheat (Triticum aestivum), rice (Oryza sativa), barley (Hordeum vulgare), and plants belonging to the families Fabaceae, Pentapetalae (Eudicots), and Chenopodiaceae (Spinacia oleracea or Beta vulgaris). This aligns well with historical records, which trace the presence of domesticated spinach to the tenth century in Mediterranean Muslim regions and to the thirteenth century in continental Europe84. The C4 plant category was present only with millet (Panicum sp.). Evidence of legumes and cereals (i.e., starch grains and aDNA within calculus) was also reported in a previous study investigating the Dalheim population71 (Table 3).

Medieval recipes often included a combination of meats, vegetables, and spices available at the time, and stews were a popular way to cook meat and vegetables together. Cereals were likely consumed frequently in different forms, such as soups, breads, and porridges, and even within pottages70. Interestingly, in a recent study, Dunne et al. (2019) identified lipids of animal tissues mixed with cruciferous biomolecules in medieval vessels at different sites in England (Table 3). This confirms the use of vessels in the preparation of stews, where a variety of foods (i.e., meats, blood, cereals, and greens) were mixed and cooked for hours on low heat. This food preparation method was particularly efficient and economical, as it minimized the dispersal of nutrients69.

The widespread consumption of nuts, vegetables, and fruits in the Middle Ages has been confirmed by archaeobotanical remains85. Pulses were particularly important for individuals from lower socio-economic statuses, who at times could not have adequate access to protein-rich animal resources70,86,87. For example, charred seeds of lentils and green peas were found in southwest Germany at sites dated to the transition from the late Antiquity to the early Middle Ages88, in medieval Ireland89, France90, northern Italy85, Spain75, and Portugal91. Starches of legumes, along with those belonging to the family Triticaceae, have also been frequently found within dental calculus deposits of medieval individuals43,46,71, confirming the frequent intake of these plants. Legumes (from the Latin legere, which means to gather) or pulses (from the Latin puls, which indicates a thick soup or porridge) were cultivated as far back as the Neolithic92,93. In the Middle Ages, legumes were a staple food, as they were affordable and easy to grow and were used in a variety of dishes69,86,87. The most often consumed species of the tribe Fabeae in the Middle Ages were beans (Vicia sp.), lentils (Lens sp.), chickpeas (Cicerum sp.), and peas (Pisum sp.). Green peas were particularly important in the diet of peasants as they are nutritionally very beneficial (Supplementary Text 2 A online). Differently from most plants, peas contain all nine essential amino acids and are particularly rich in lysine and threonine. However, peas are relatively poor in sulfur-rich amino acids (i.e., methionine and cysteine)94,95. Including peas and other legumes96 in ancient diets would have provided individuals with proper protein intake, even in the absence of animal sources. Unfortunately, legumes as food components are difficult to detect isotopically. In contrast to animal food, they have significantly lower δ15N values because they fix atmospheric nitrogen (δ15N value of air = 0 ‰). This suggests that individuals with low δ15N values may have frequently consumed legumes. As such, the combination of isotopic and proteomic data is very useful for reliably assessing the consumption of legumes.

Aside from plants, the nitrogen isotopic values in this study indicate that the Dalheim population consumed animal products (i.e., meat, eggs, dairy). Unfortunately, no proteins specific to blood, meat, eggs, or dairy, were found within calculus deposits, however these have only been recovered in rare archaeological cases28,97.

The only dairy protein (β-Lg, a whey protein) was found in one individual from Baar, for which population stable isotope data are not available. Meat, especially from game, and fish were commodities most often available only to the wealthier class70,86. Although not confirmed by our study, Olsen et al. (2018) were able to highlight marked differences in animal protein intake between juveniles and adults at Dalheim. This might reflect a general low availability of animal proteins in the form of meats. To supply more animal proteins into their diets, people at Dalheim may have included meats from freshwater fish. According to Christian rules, fish was allowed to be eaten on religious fast days, in contrast to meat from terrestrial animals98. The consumption of European perch is suggested for both Baar and Dalheim, as proteins assigned to this species were identified in both populations. Perch, along with carp and pike, was one of the most common fishes consumed in the Middle Ages in Europe69,99,100. Overall, the proteomic study did not reflect changes or discernible differences in the diets of the two investigated populations.

Data from the previously conducted paleo-odontological analysis can further enrich our understanding of the dietary habits at Baar and Dalheim101,102. Generally low incidences of caries were observed at both sites, suggesting minimal consumption of cariogenic foods such as honey and sticky cereal-based porridges. Elevated levels of tooth wear in certain individuals were attributed to flours and breads containing particles from the millstones used for grinding cereal seeds. Moreover, no disparities in food access based on sex were detected within both populations.

This study, in line with others (Table 3), showed that a combination of biomolecular methods in the study of bioarchaeological remains improves the overall reconstruction of past dietary habits. In particular, the use of complementary methods when sample preservation is not optimal is particularly useful to maximize the extraction of information from bioarchaeological remains. Lastly, since dietary proteins were also found in a few samples with low preservation of the oral signature, it would be recommended in the future to systematically apply strict decontamination protocols to all calculus samples before analysis in order to enhance the reliability of results on ancient specimens43,46,103.

Considerations: Differential preservation of dietary proteins in ancient dental calculus

Dairy proteins are frequently found in proteomic studies investigating dental calculus, so these seem to preserve well under specific conditions28,34,35,36,37,38. Beta-lactoglobulin (β-Lg), the most abundant whey protein, is also the most commonly recovered dairy protein in ancient dental calculus28,104,105. The fact that we did not find dairy proteins (especially from the whey portion of milk) in the dental calculus from the Dalheim individuals and that there is only one attestation in the Baar population does not mean that milk or dairy products were not consumed but could rather suggest that dairy products were not frequently consumed in large quantities. It may also be that the dairy products consumed were those with the whey processed out106, so in this case, fewer β-Lg proteins would have been incorporated in the dental calculus. However, poor dietary protein recovery might also depend on the sediment conditions at both sites, which might have affected the recoverable proteins in the calculus104.

As for plant proteins, it has been challenging to recover them from ancient dental calculus28,35. One of the reasons is that the databases used for protein identification may be biased, and the sequences for many understudied plant proteins are missing from known protein databases28. The varying degree with which dietary proteins are incorporated in the calculus, preserved, and subsequently recovered, may pose serious issues in interpreting and reconstructing the diets of past human populations. An absence of dietary proteins does not mean that these foods were not eaten; rather, that we do not have direct evidence of their consumption. This suggests that protein recovery and human diet reconstruction based on these datasets may also depend on taphonomic or site-formation processes, which can either favor or hinder biomolecule stability and preservation104,107. There is limited knowledge about which conditions are conducive to dietary protein incorporation in dental calculus and protein preservation. The likelihood of preservation might also depend on the initial abundance of proteins in specific foods. For instance, legumes contain more legumins than vicilins on average and should be detected more often. However, initial abundance may not always be the determining factor, as demonstrated by dairy proteins108. In general, whey proteins are more likely to be preserved in the archaeological record than caseins (e.g. Hendy, 2021), even though caseins comprise a much higher percentage of fresh milk proteins with a ratio of caseins to whey proteins in cattle, sheep, and goats of ~ 80:20109.

An additional factor affecting ancient protein preservation could be related to the secondary and tertiary structures of proteins. Secondary structure refers to the local folding patterns of the amino acid chain, typically forming alpha helices or beta sheets through hydrogen bonding between amino acid residues. Alpha helices resemble a coiled spring, while beta sheets are composed of strands that may be either parallel or antiparallel. Tertiary structure refers to the overall 3D arrangement of proteins (Fig. 3), resulting from interactions between different regions of the protein chain, including hydrophobic interactions, hydrogen bonding, electrostatic interactions, and disulfide bonds. Tertiary structure gives proteins unique, functional shapes and determines their ability to interact with other molecules. It also plays a crucial role in protein stability and function110. Some of these structures may prove more favorable for best preservation in sediments108. For instance, the inner part of globular proteins is hydrophobic, while the outer part is hydrophilic to perform their functions in solution110. Hence, the inner parts of these proteins could be better preserved in archaeological contexts, even when water is present.

Fig. 3
figure 3

3D representations of globular proteins found in green pea seeds and in bovine whey (downloaded from https://www.uniprot.org). a) Vicilin; b) Legumin A; c) Convicilin; d) Bovine beta-lactoglobulin (β-Lg).

The leguminous proteins recovered in this study are all globular (Fig. 3: a-c). Legumin (11S globulin) is a hexamer with molecular weight ranging from 320 to 400 kD andit is characterized by a compact quaternary structure with six sub-units assembled via two trimeric intermediates (Fig. 3: b). It has a β-sheet-rich structure, and disulfide bonds link the α-chain and β-chain together95,111. Disulfide bonds are not present in vicilin (7S globulin) due to the lack of cysteine residues. Vicilins are trimeric molecules with molecular mass ranging from 150 to 170 kDa built up by noncovalent interactions (Fig. 3: a)95,111. The subunit convicilin is another globular protein with a molecular mass of 70 kDa (Fig. 3: c). It displays a substantial amino acid sequence homology with vicilin. Unlike vicilin, convicilin has a sulfur-containing amino acid, cysteine, and a highly charged N-terminal extension region95,112.

The most abundant protein in whey, β-Lg, also has a globular structure (Fig. 3: d). While caseins generally have little secondary and tertiary structures, β-Lg has a compact structure, composed mainly by β-sheet (43%), a smaller fraction of α-helix (10–15%), and the remaining portion of unordered structures108,109. Caseins usually form micelles with multiple molecules held together by noncovalent bonds and calcium phosphate. These micelles, being porous, are suggested to be more vulnerable to environmental factors, potentially explaining why caseins deteriorate more than β-lactoglobulins in archaeological samples108. It emerges that the careful evaluation of secondary and tertiary structures of proteins retrieved from ancient dental calculus could provide valuable hints about their selective preservation and aid researchers in accurately interpreting the dietary data obtained from proteomic analysis.

As discussed above, one important consideration is the effect of secondary and tertiary protein structures on the preservation and detectability of proteins in ancient samples. The higher-order structures of proteins can influence their stability and resistance to degradation. In some cases, proteins that retain their secondary and tertiary structures may be more resistant to environmental damage and, thus, more likely to be detected in preserved samples. Conversely, proteins that have lost these structures may degrade more rapidly and be underrepresented in analyses. Understanding the role of these structures can help refine preservation assessments and improve protein recovery techniques.

Based on our results and others (Table 3), we put forward the possibility that plant seed material caught within the dental calculus biofilm may benefit from extended preservation, even as several of the salivary and oral microbiome proteins did not (Supplementary Dataset: Tables 12). Consequently, by excluding samples that do not meet high preservation standards, there is a risk of losing valuable data that may still contribute to the overall understanding of the study, albeit with caution. Marginal samples may still contain valuable information, especially in cases where well-preserved samples are scarce. We think that researchers should carefully weigh the benefits of including marginal samples against the potential risks of introducing noise into the data set.

Conclusions

The combined study of stable isotopes and ancient proteins offers a complementary approach to reconstructing past human diets, providing insights that neither method can achieve alone. Despite challenges related to sample preservation and differential protein recovery, our study demonstrates the value of integrating stable isotope and proteomic data for paleodietary reconstructions. We analyzed dental calculus, dentin, and bone samples from two medieval European populations. Stable isotope data revealed that the population of Dalheim relied primarily on proteins from terrestrial animals, alongside the consumption of plants (mainly C3 plants). While isotope analyses did not provide clear evidence of fish or legume consumption, proteomic data confirmed the presence of both in the diet. Proteomic analysis of dental calculus samples from both sites revealed dietary proteins originating from plant and animal sources. Identified taxa included common wheat, barley, millet, green peas, and European perch. Stable isotope values derived from dental calculus were also examined. However, due to the material’s inherent variability and complex composition, we conclude that it does not yield reliable isotopic signals for dietary reconstruction.

Our results highlight the importance of a dual approach in paleo-dietary studies. Stable isotope analysis provided a broad overview of diet composition, while proteomic analysis enabled identification of specific plant and animal species. However, the study also revealed potential limitations, particularly with OSSD screening for sample preservation. Subjectivity in scoring, variability in within-sample preservation, and exclusion of marginal samples can affect result robustness. We recommend improving and standardizing scoring procedures, considering protein structure’s role in preservation. This study paves the way for further interdisciplinary research, encouraging new analytical techniques and diverse datasets to build a more detailed and accurate picture of ancient and historical populations’ lifeways.

Materials and methods

The studied human remains

Two anthropological collections were included in this research: the Dalheim collection, curated at the Institute of Evolutionary Medicine (IEM) at the University of Zurich (UZH), and the Baar—Früebergstrasse collection, curated by the Museum of Prehistory of Zug (Canton Zug). Although no ethical review board approval was necessary for this study, authorization to study the material was granted by the aforementioned institutions. Moreover, this research was strictly committed to the code of ethics of the IEM-UZH113.

The Baar cemetery (Canton of Zug, Switzerland) dates to the seventh century CE and was excavated in 2000 by the Archaeology Department of Canton Zug (Switzerland)77. The Dalheim monastic cemetery (Lichtenau, North Rhine-Westphalia, Germany) dates to the ninth-twelfth centuries CE (Fig. 4: a) and was excavated from 1989 to 1990 by the by the Westphalian Museum of Archaeology114,115,116. Sex and age-at-death of all individuals were determined in previous studies114,117. Dental calculus of four individuals from Dalheim were investigated in previous studies45,118 (Table 3).

Fig. 4
figure 4

a) Map with the location of the two archaeological sites investigated under study. b) A microscopic graph of a labial calculus deposit on a lower right lateral incisor (#42, Fédération Dentaire International tooth numbering system—119) of a female individual from Dalheim (G48).

In this study, calculus was collected from 24 individuals from the Dalheim cemetery and from 28 individuals of the Baar—Früebergstrasse site. In total, 52 samples of dental calculus were subjected to ancient proteomic analysis (Supplementary Table S1 online; Fig. 4: b). Furthermore, samples from bone (n = 11), dentin (n = 8) and dental calculus (n = 8) were taken from eleven individuals from the Dalheim collection for isotope analysis (Supplementary Table S2 online).

Isotope analysis

Sampling

To complement proteomic analysis, tooth samples from 8 individuals and bone samples from 11 individuals (10 ribs and 1 skull fragment from the Dalheim collection were sampled for isotope analysis at the Evolutionary Pathophysiology and Mummy Studies Group laboratory at IEM (Supplementary Table S2 online). Samples of dental calculus (~ 5–8 mg) from 8 individuals were available for stable isotope analysis after sample preparation for proteomic analysis (Supplementary Table S1 online). It was not possible to invasively inspect the human remains of the Baar collection, which is why no isotopic data is available.

Preparation of calculus and collagen

The tooth, bone and dental calculus samples were prepared for stable isotope analysis at the Institute of Legal Medicine of the Ludwig-Maximilians-University in Munich (Germany). Dental calculus samples were cleaned with distilled water. Afterward, the dry material was crushed in a mortar and pestle. The teeth were cleaned with demineralized/deionized water in an ultrasonic bath and divided into crown and root using a saw. About 1 g per individual was cut off from the rib fragments and cleaned with demineralized/deionized water in an ultrasonic bath. The tooth roots and the rib samples were ground with a tungsten ball mill. Collagen was prepared from ~ 0.5 g bone and dentin material (demineralization with 1 M HCl, gelatinization overnight with 1 mM HCl at 900 C, filtration through a glass-microfiber disc and membrane filter (Whatman AE 98; GE Healthcare, Buckinghamshire, UK) and freeze-drying, of the collagen solution)120,121.

Stable isotope analyses

The resulting collagen and calculus samples were used for stable isotope measurements of carbon, nitrogen, and sulfur using an isotope ratio-mass spectrometer (IR-MS) (Isolab GmbH, Schweitenkirchen, Germany). For multi-elemental stable isotope analyses of carbon, nitrogen, and sulfur, 4 × 3 mg each of the dental and the rib bone collagen and 1–4 × 1 mg from the calculus were weighed into tin capsules. Specific internal standards (casein, porcine collagen, and bovine collagen) were used for calibration. The detailed analytical procedures are documented in Lehn et al. (2022, 2020).

The stable isotope abundant ratios of the above elements (13C/12C, 15N/14N, 34S/32S) are given in delta (δ) notation (δ13C, δ15N, δ34S). The values are related to internationally defined standards (C: V-PDB, N: air, S: V-CDT) and are described by the following formula (using the stable carbon isotopes as an example): δ13C = ((13C/12C) sample/(13C/12C) standard)—1) × 1000 [‰]. The following applies to the analytical uncertainties of the measured values: δ13C: ± 0.1 ‰, δ15N: ± 0.2 ‰, δ34S: ± 0.3 ‰.

Statistics

The Kruskal Wallis (H) test and the Mann–Whitney U test (p < 0.01) were applied for comparing carbon and nitrogen values in the different groups (dentin, bone, and dental calculus). All the analyses were performed using R software (version 4.2.2)122.

Proteomics of dental calculus

Sampling

A total of 52 samples of dental calculus from two archaeological sites, the medieval cemeteries of Früebergstrasse at Baar (Canton of Zug, Switzerland) (n = 28) and of Dalheim (North Rhine—Westphalia, Germany) (n = 24), were included in this study (Fig. 4; Supplementary Table S1 online)61,77. Dental calculus samples were removed from teeth using a dental scaler, which was cleaned with ethanol between individuals to avoid cross-contamination. The samples were stored in sterile 2 mL Eppendorf tubes until protein extraction in the dedicated ancient protein laboratory at the Institute of Evolutionary Medicine (IEM) at the University of Zurich (UZH). A detailed description of the methods is found in Supplementary Text 1 online.

Peptide extraction

Proteins were extracted using a single pot, solid phase enhanced sample preparation (SP3) protocol optimized for dental calculus which is fully and freely available at protocols.io (https://doi.org/https://doi.org/10.17504/protocols.io.bfgrjjv6).

LC–MS/MS analysis

Mass spectrometry analysis was performed in the Functional Genomics Centre of Zürich (FGCZ) on a Q Exactive HF mass spectrometer (Thermo Scientific) equipped with a Digital PicoView source (New Objective) coupled to an M-Class UPLC (Waters). Solvent composition at the two channels was 0.1% formic acid for channel A and 0.1% formic acid, 99.9% acetonitrile for channel B. Column temperature was at 50 °C. For each sample 2 to 5 μL of peptides were loaded on a commercial ACQUITY UPLC M-Class Symmetry C18 Trap Column (100 Å, 5 µm, 180 µm × 20 mm, Waters) followed by ACQUITY UPLC M-Class HSS T3 Column (100 Å, 1.8 µm, 75 µm X 250 mm, Waters). The peptides were eluted at a flow rate of 300 nL/min. For measurement on the Q Exactive HF, after a 3 min initial hold at 5% B, a gradient from 5 to 22% B in 50 min and 22 to 40% B in additional 10 min was applied.

The mass spectrometer (Q Exactive HF) was operated in data-dependent acquisition (DDA) mode with a maximum cycle time of 3 s, using Xcalibur (tune version: 4.4), with spray voltage set to 2.3 kV, funnel RF level at 60%, and heated capillary temperature at 275 °C. Full-scan MS spectra (350 − 1′500 m/z) were acquired at a resolution of 120′000 at 200 m/z after accumulation to an automated gain control (AGC) target value of 100′000 or for a maximum injection time of 100 ms. Precursors with an intensity above 45′000 were selected for MS/MS. Ions were isolated using a quadrupole mass filter with a 1.2 m/z isolation window and fragmented by higher-energy collisional dissociation (HCD) using a normalized collision energy of 28%. MS2 spectra were recorded at a resolution of 30′000 and a maximum injection time of 50 ms. Charge state screening was enabled, and singly, unassigned charge states and charge states higher than seven were excluded. Precursor masses previously selected for MS/MS measurement were excluded from further selection for 30 s, applying a mass tolerance of 10 ppm. The samples were acquired using internal lock mass calibration on m/z 371.1012 and 445.1200.

Data analysis

Data analysis was performed by Mascot (2.7.0.1)123 and included the following settings: carbamidomethylation of cysteine (C) as a fixed modification, deamidation of asparagine (N) and glutamine (Q), and oxidation of methionine (M) as variable modifications. The instrument was set as QExactive, with precursor ions mass tolerance at 10 ppm, with allowances for one isotopic mass shift, and fragment ion mass tolerance at 0.01 Da. Trypsin was selected as the enzyme, and we allowed up to 3 missed cleavages per peptide and included peptides with charges of 2 +, 3 +, and 4 +. Our customized database is compiled of the entirety of Swiss-Prot (edition: 2021_04) combined with a database used in previously published studies32,36. Raw, MGF, and MZID files are freely available as a MassIVE dataset at. https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=92df91ac70354e65be1ad47f00f344f5.

For individuals that passed the OSSD (Supplementary Dataset online: Table 1), the recovered peptides were filtered to achieve a peptide FDR of 1% or less in Mascot (Supplementary Dataset online: Table 2). HTML reports summarizing the search and a.csv file containing confidently identified PSMs were generated through MS-MARGE (https://bitbucket.org/rwhagan/ms-marge/src/master/). False discovery rate (FDR) at the PSM and protein level were estimated using the same script. A minimum of two individual PSMs are required for specific protein identifications and only peptides with an e-value below 0.01 are usually accepted. For data reporting, the free software R (version 4.2.2) with the tidyverse, cowplot, R.utils, tools, datasets, ggplot2, and ggpubr libraries were used122.

Dietary protein identification

All dietary peptides were individually assessed and all PSMs were searched against the translated nucleotide database at NCBI using BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Positive peptide identification was achieved when 100% homology and 100% coverage of the peptide to the desired dietary protein were observed.