Stable C & N isotopes in 2100 Year-B.P. human bone collagen indicate rare dietary dominance of C4 plants in NE-Italy

C4 plants (e.g. maize, millet), part of our current diet, are only endemic of reduced areas in South-Europe due to their need of warm climates. Since the first vestiges of agriculture in Europe remains of C4 plants were recorded but their overall proportion in the human diet remains unknown. Therefore, isotopic (δ13C and δ15N) composition of bone collagen from the skeletal remains (human and animals) of a Celtic population, Cenomani Gauls, from Verona (3rd to 1st century BC) in the NE Italy provide a new perspective on this matter. The δ13C collagen values of 90 human skeletal individuals range between −20.2‰ and −9.7‰ (V-PDB) with a mean value of −15.3‰. As present day C4 plants have δ13C values around −11‰, which is equivalent to −9.5‰ for samples of preindustrial age, the less negative δ13C values in these individuals indicate a diet dominated by C4 plants. This palaeodietary study indicates that some European populations predominantly consumed cultivated C4 plants 2100 year B.P. This is supported by the paleobotanical records and ancient Roman sources (e.g. Pliny the Elder), which indicate that millet was a staple food in South-Europe.

Before the onset of agriculture, herbivores collagen isotopic δ 13 C values from S-Italy (32.6 to 13.3 ka B.P.) suggest that C 4 biomass was practically absent (calculated vegetal biomass is around − 25‰ vs V-PDB) 16 . Even during the warmer periods of the Late Miocene, fossil isotopic data indicate the absence of C 4 biomass in central and southern Europe 17,18 . Man has introduced certain species, such as millet, with a planting strategy in the warm season (i.e. summer), possibly as a response to Holocene arid periods 11 .
The stable isotope composition (δ 13 C and δ 15 N) is a very effective tool to study the trophic web in ecosystems [19][20][21][22][23] . This methodology is crucial for studying past ecosystems and paleodiets, because direct observation is impossible [24][25][26][27] . In addition this technique allows quantitative analysis, which even in current ecosystems is difficult to obtain with the classical methods of observation. Isotopic carbon traceability suffers less change in the steps of the food chain than isotopic nitrogen one. Therefore, the footprint of primary producers (e.g. plants) is clearer 28 .
The importance of carbon isotopes as a racer for the presence of C 4 plants in the diet, is that atmospheric CO 2 is fixed into organic matter by the enzyme PEP carboxylase (C 4 plants) which discriminates much less against 13 C than the enzyme RuBisCO (C 3 plants) 29 . In present day C 3 plants biomass show typical δ 13 C values around − 26‰ (V-PDB), while C 4 plants are around − 11‰ (V-PDB) 30,31 , − 24.5‰ and − 9.5‰ (V-PDB), respectively in preindustrial age, because δ 13 C values of CO 2 were less negative by 1.5‰ compared to the present values (− 8‰ V-PDB). This 15‰ difference between C 3 and C 4 plants allows an accurate calculation of the percentage of C 4 plants in the diet. The δ 13 C collagen values of humans and animals are proportional to the isotopic signal from the base of the food chain (plants), but enriched by 0.8 and 1.3‰ for herbivores and carnivores, respectively 32 . The δ 15 N of the collagen is also related with the food chain and can be used to evaluate the trophic system, because in a particular ecosystem there is a significant enrichment in 15 N between 3 and 5‰ for each trophic level [25][26][27] .On the other hand δ 15 N cannot be used to distinguish between C 3 and C 4 plants consumption.
There is little knowledge about the nutritional habits of protohistorical populations such as Pre-Roman-Celtic ones. Therefore, we performed an isotopic study to unveil the dietary habits of the Celtic population Cenomani Gauls, from the necropolis of Seminario Vescovile in Verona (Italy) dated between the 3 rd to 1 st century BC 33 . This necropolis counts with a minimum of 174 skeletons in a good state of preservation, and the majority of them are non-adults (see Supplementary Information for detailed descriptions of the archaeological context). Surprisingly, we recorded relatively high, i.e. less negatives, δ 13 C collagen values in the human samples, and even in some animals found in the necropolis. In our initial considerations, we suspected the presence of protein in the diet from marine sources, or even anomalous fresh water sources (enriched 13 C in DIC-Dissolved Inorganic Carbon, "dead carbon" from marine carbonates), or alternatively an extensive consumption of C 4 plants. The first hypothesis seemed from the beginning improbable, taking into account the geographic location of the necropolis, which was far from the sea (about 120 km). Additionally, the less negative δ 13 C collagen, − 10‰ (V-PDB), is higher than what would be expected for a signal from marine primary production 22 . A second alternative explanation is a freshwater tropic web with an anomalous enriched 13 C source of carbon (DIC) from dissolution or thermic decomposition of carbonates. This geochemical scenario is present in some areas of Italy 34 in which natural water becomes enriched in CO 2 . This archaeological site was near to the Adige River that runs through Verona. Therefore, water samples from the Adige river and its tributaries were collected and analysed. In addition, 14 C dating was done on some human remains to detect the presence of "dead carbon".
The main objective of this study is to demonstrate that these less negative δ 13 C values are related to the consumption of C 4 plants and refute the improbable hypothesis of a significant marine or anomalous fresh water dietary intake (enriched in 13 C). At the same time, we set out to quantify the proportion of C 4 plants in the diet and propose an interpretation about the alimentary habits of Celtic populations of Italy which up to now have been elusive.  Table 1;  Supplementary Table S2). The atomic C/N ratio of the samples falls inside the range of 2.9-3.6, which corresponds to typical values of well-preserved samples, as suggested by DeNiro 35 . Isotopic data from animal bones (Tables 1 and 2) suggests that domesticated herbivores had a diet mainly based on C 3 plants, although one of these (VRAR-1 with − 17.2‰ V-PDB) showed a less negative δ 13 C value that could suggest that this animal partly fed on C 4 plants. Dogs (n = 2) usually eat men's leftovers, showing relatively high δ 13 C values (≈ − 13‰ V-PDB). One of them also has a relatively low δ 15 N value for an omnivore. This implies a diet poor in animal proteins, supporting the hypothesis of an important influence in their diet of human food scabs, rich in (C 4 ) plants.

Isotopic
Isotopic values of DIC of Adige's river and Radiocarbon data. The DIC δ 13 C data (Table 3 and Supplementary Figure S1b) of the Adige river ranges between − 4.5‰ and − 5.0‰ (V-PDB); this, after isotopic fractionation processes of photosynthesis in algae, would result in values ranging from − 23.5‰ and − 24‰ (V-PDB) and consequently it would imply δ 13 C values in fish tissues very close to these values (≈ − 22.5‰ to − 23‰ V-PDB). These data would result in a value for human collagen resulting from an exclusively fish diet (an extreme that would be relatively rare), between − 20‰ to − 22‰ (V-PDB), which is not compatible with the least negative values found in this work. The rest of δ 13 C values of DIC of other tributaries, wells, etc. are even more negative so they also cannot justify the higher δ 13 C values found in collagen of human remains 36 .
On the other hand, the age obtained from 14 C dating of bone collagen from four individuals (those with less negative δ 13 C values) is consistent with that extrapolated (Table 4) from the preliminary archaeological analysis of the grave goods 36 . This indicates the absence of "dead" carbon (e.g. dissolution of marine carbonate rocks), which would lead to apparently much older ages. Such dissolved inorganic carbon (DIC) have δ 13 C values close to + 0‰ or positive values, which would also lead to an aquatic trophic chain with less negative values of δ 13 C, this being the only alternative to the consumption of C 4 plants. The absence of "dead" carbon in the humans collagen rules out this hypothesis, and thus consequently supports the presence of C 4 plants in the diet of these individuals.

Discussion
The isotopic results show the important role that C 4 plants had in the diet of this Celtic population from Verona (Fig. 1). Generally, δ 13 C values of − 20‰ (V-PDB) or less negatives would indicate some C 4 plant contributions. Thus the vast majority of studied individuals (over 90%) seem to include C 4 plants in a direct way (or indirectly through the consumption of herbivores that fed on them) in their diet. Additionally, adult females (mean value of 48.8 ± 15.1%) show significantly (Mann-Whitney U test = 198.0; p = 0.008) higher consumption of C 4 plants than adult males (mean of 37.4 ± 16.8%) (Fig. 2). We also found a significant statistical difference in δ 15 N values between sex in the adult group (t = 3.15 p ≤ 0.01): men display higher δ 15 N values compared to women (Fig. 3). Thus, there is a clear differentiation in the diet according to sex with a higher intake of animal protein (meat and meat products) for men, while the diet of women included more cereals and vegetable proteins. It is important to     Table S2), indicating that domestic herbivorous have also consumed large amount of C 4 plants, although the few herbivorous studied show a consumption of C 3 plants. However, two dogs show δ 13 C values which are very similar to those of humans (Fig. 1). Especially the younger dog (VRSV-94) presents a typical value of the middle-low animal protein intake of omnivores: δ 15 N = + 7.1‰ (AIR). While the δ 13 C value of − 13.7‰ (V-PDB) is very close to those of the humans with a C 4 diet ( Table 2). As mentioned before, a possible explanation is that dogs were fed with the impoverished-protein waste from the meals of their owners. DIC values (Table 3) from local meteoric water show relatively negative δ 13 C values (from − 4.5 to − 13.3‰ vs V-PDB; i.e. an aquatic diet should be approximately between − 23.5 and − 32.3‰ vs V-PDB) and 14 C values of the human samples with less negative δ 13 C collagen values does not indicate the presence of "dead carbon". Consequently, this excludes a freshwater diet to explain the measured less negative isotopic values of this particular Celtic population and thus supports the hypothesis of a large consumption of C 4 plants in their diet. Furthermore, δ 15 N values in humans (+ 9.4‰ AIR) plot in about a one trophic level above terrestrial herbivorous (+ 4.7‰ AIR) 36 , rule out a marine diet, in agreement with the relative large distance from the coast.
Considering that there are not many known cultivated C 4 plant species in this time period (3 rd -1 st BC), the hypothesis of millet consumption seems reasonable based on the archaeological record and ancient written   Table 4. 14 C dating of the less negative δ 13 C human collagen samples. Unfortunately, during the archaeological excavation of Verona, archaeologists have not found seeds (or charcoal) and the ceramic grave goods (vessels and others) are still under study and initially seem not to indicate macroscopic remains of organic waste. Some previous archaeobotanical studies of Neolithic, Bronze and Iron Age sites of the NE of Italy, and particularly of the area of Verona, seem to confirm the presence of Panicum Miliaceum and of Setaria italica among the carpological remains 41,42 . Another comparative paleobotanical study relating to some Iron Age sites located between the Northern and Southern Alps (Eastern Switzerland, Austria and Northern Italy), also describes the presence of millet (Panicum Miliaceum L.) and foxtail millet (Setaria Italica L. Beauv.) remains 43 . This is also confirmed by the results of the study about carpological remains found at the site of Oppeano (Verona), dated to the second Iron Age (about 6 th to 3 rd century BC) and located in the same geographical context of the Seminario Vescovile necropolis. The majority of the determined cereal remains correspond to millet (Panicum miliaceum L.) and foxtail millet (Echinochloa crus-galli L. Beauv. and Setaria italica L. Beauv) 44 . Thus, the Iron Age seems to be characterized by a greater crops specialization that fits with the assumption of the pursuit of a cereal type that best fits different climates (biomass increase with more efficiency in water use). In this case of the sandy Pianura Padana (Verona), the crops of millet were preferred, because it is a plant that adapts very well to poor substrates characterized by water scarcity during summer 44 .
Finally, other isotopic results from bone collagen analysis of some individuals proceeding from the Bronze Age necropolis of Olmo di Nogara, Verona (1600-1200 cal. BC) and of Arano di Cellore, Verona (2040-1890 cal. BC) support the hypothesis of C 4 plants consumption, and in particular of millet in some areas of the Verona province 45,46 (Fig. 4). In conclusion, to these archaeological data we add the testimonies of some ancient authors such as Pliny the Elder (Naturalis Historiae XVIII, 83-84) and Columella (De Re Rustica, 2,9,[14][15][16] who report the use of millet flour for the production of bread and a sort of porridge cooked in water and salt and often accompanied with vegetables and cheese and very rarely with meat. Pliny (XVIII.XXIV) exactly noted: "millet is used to prepare a very white puls (i.e. similar to present-day polenta). Panic, when ground and freed from bran, and millet as well, makes a porridge which, especially with milk, is not to be despised even in time of plenty". Columella (2.9.19), agreed with Pliny and wrote: "bread is made of millet, and it may be eaten without distaste before it cools".
The isotopic data obtained in this study reveal that the proportion of C 4 plants was substantial (above ≈ 40%) in the daily diet of this Celtic population. Traces of C 4 plant consumption are reported in other studies 45,46 from more ancient necropolis of the zone. While Olmo de Nogara (ODN) dated to Middle Bronze Age share very similar isotopic values with Verona (Fig. 4), confirming a preponderant C 4 plants based diet, Arano di Cellore (AC), dated to Early Bronze Age, shows a diet based on the consumption of C 3 cereal-type plants. Hence, the data from AC and ODN firmly place the shift in C 4 crop use in the region at a transition period between the late phases of the Early Bronze Age and the beginning of the Middle Bronze Age 46 , consequently our data show that successively, in pre-Roman times, the diet of most habitants of the zone was almost exclusively based on C 4 plants. This, indirectly indicates that this Celtic population had mastered agriculture techniques to such an extent that they could live off the harvest of their cultivated crops the year round.

Methods
A sample of 90 human ribs, consisting of both sexes and different ages (Supplementary Information Table S1), and 7 animal bones, was selected for analysis of bone collagen. The animal bone sample is composed of 5 herbivorous species (2 horses, 1 goat/sheep and 2 cows) and 2 omnivorous species (2 dogs). The extraction of collagen is performed using the protocol described by Bocherens et al. 47,48 and following the routine procedures of the Stable Isotope Biogeochemistry Laboratory of the Andalusian Institute of Earth Sciences (CSIC, Granada, Spain). We also determined the current DIC values of the Adige's river (and of some of its tributaries) and dated with 14 C the collagen samples of the less negative human values (a source of "dead" inorganic carbon would show abnormally high ages).
The δ 13 C mean values of present day C 3 and C 4 plants are respectively − 26‰ and − 11‰ (V-PDB) 49,50 , but before the industrial revolution the isotopic composition of atmospheric CO 2 was 1.5‰ less negative 51 . Hence, to make a more precise estimate we considered that these two carbon sources would respectively correspond to values of −24.5‰ and −9.5‰ (V-PDB) (see Supplementary Information for detailed descriptions about the different methodologies applied).  45 and Arano di Cellore (AC) 46 . We have also plotted the faunal ranges from the three necropolises. While Verona and Olmo di Nogara (Middle Bronze Age) have very similar isotopic values, confirming a largely C 4 plants based diet, Arano di Cellore, dated to Early Bronze Age shows a diet based on C 3 cereal-type plants.