Regional long-term analysis of dietary isotopes in Neolithic southeastern Italy: new patterns and research directions

Isotopic analyses of prehistoric diet have only recently reached the threshold of going beyond site-focused reports to provide regional syntheses showing larger trends. In this work we present the first regional analysis for Neolithic southeastern Italy as a whole, including both substantial original data and a review of the available published data. The results show that dietary isotopes can shed new light on a number of traditional and important questions about Neolithic foodways. First, we observe regional variations in the distribution of stable isotope values across the area, suggesting variability in the Neolithic diet. Secondly, we show that, although the plant food calorific intake was primary for these communities, animal products were also important, representing on average 40% of the total calories. Third, we note that marine fish was only minorly consumed, but that this could be an underestimation, and we observe some variability in the regions considered, suggesting differences in local human–environment interactions. People in different regions of southeastern Italy may have consumed different versions of a common Neolithic diet. Regional synthesis also allows us to take stock of gaps and new directions in the field, suggesting an agenda for Neolithic isotopic research for the 2020s.


Results
We collected human and animal bone samples from a variety of Neolithic sites from southeastern Italy (Fig. 1). Archaeological information for each site studied here is provided in the Supplementary Information S1.
This section presents the new isotopic data, which will be contextualised using existing data in the following discussion section. New data comprises 94 human samples and 16 animal samples of various wild and domesticated species (namely, Bos sp., Sus sp., Ovis vel Capra, Cervus sp. and Capreolus sp.) (Table S1). In the following analysis, only samples with C:N ratio between 2.9 and 3.6 are included 46 . Out of the total, this limits our results to 88 human samples and 15 animal samples, with mean δ 13 C values of − 19.6 ± 0.5‰ and δ 15 N values of 9.0 ± 1.1‰ for the humans and mean δ 13 C values of − 20.6 ± 0.7‰ and δ 15 N values of 6.4 ± 1.0‰ for the animals. Table 1 reports descriptive statistics of the data that passed the quality check presented here and of those previously published used for comparative purposes. Previously published data are reported in Table S2.
We explored the new data in relation to previously published domestic and wild fauna 8,45 , and present them below distinguishing between Early, Middle and Late Neolithic phases (Fig. 2).

Discussion
Cucina tipica in the Neolithic? Even when people in a region share a common tradition for producing foods, and a common repertory of ingredients and techniques for preparing them, there can be strong local differences in how they eat. For example, recent historic cuisine in peninsular Italy is based on a limited, common range of resources (grains, vegetables, animal products such as cheese, pork and beef) with some regional variations (notably, the availability of fish, and the use of different grains suitable for different local Table 1. Descriptive statistics of the human remains analysed in this study and those from previously published work. Site n mean δ 13 C median δ 13 C 1SD δ 13 C min δ 13 C max δ 13 C mean δ 15 N median δ 15  www.nature.com/scientificreports/ climates). In spite of this, there are strong regional traditions of cucina tipica, or local cuisine, which combine and emphasise these in distinctive ways. Roman writers suggest that there were regional variations in foodways in Classical times, and as food is an important part of the habitus and the reproduction of communities, there is no reason to suppose that it was otherwise in prehistoric periods. Much of regional foodways may be invisible archaeologically, though we might perhaps see local patterns of material culture (e.g. pottery forms), animal remains (e.g. butchery patterns), or food consumption. The only area of Italy where the isotopic data for the Neolithic are dense enough to examine is southeastern Italy, including Puglia and the adjacent regions of Basilicata and Abruzzo. To evaluate regional differences, we explored the variability of δ 13 C (‰) and δ 15 N (‰) of humans throughout the southeastern area of the Peninsula, coupling data reported here with those of published works. The data were elaborated using the Bayesian model AverageR (see Materials and Methods for a description 47,48 , and the outputs presented in Fig. 3a,b. Terrestrial animals δ 13 C (‰) and δ 15 N (‰) values were also explored as these most likely reflect the local environment and therefore can help in the discussion of the human data. Models for the animals are reported in Fig. 3c,d. Both δ 15 N (‰) and δ 13 C (‰) values in animals do not differ according to study area (Fig. 3c,d). This could suggest that the difference in climatic and environmental conditions from one area to another is not so marked to be reflected in the isotopic values of animals, also according to modern climatic data 49 . However, the animal assemblage is limited considering the area of investigation, therefore the evidence is not strong enough to exclude climatic or environmental variations at this stage. On the contrary, the human δ 15 N (‰) and δ 13 C (‰) values appear to vary slightly but significantly from one region to another. The differences visible from the maps were also explored using descriptive statistics (Table S3), plots with mean values and associated 95% CIs (Fig. S1) and pairwise comparisons using Wilcoxon rank sum test (Table S4) 50 . Inland Murge and coastal Tavoliere differ by their mean values, as well as Salento from the coastal Murge and from the coastal Tavoliere (Fig. S1, Tables S3 and S4, p-values < 0.05). Higher δ 15 N (‰) values in the inland Murge and in the inland Tavoliere could suggest a slightly higher consumption of animal products in these regions; in fact, a stronger reliance on animals in the uplands of the Murge has been already discussed 51 . The δ 15 N (‰) values from Salento are higher than those from the coastal areas of the Murge and the Tavoliere. This could suggest higher consumption of marine products compared to other sites, again close to the sea, but from a different area. As for the distribution of δ 13 C (‰) values, differences are found between the coastal Murge and the coastal Tavoliere, as well as between the coastal Tavoliere and the inland Murge and between the inland Murge and the Salento area (Fig. S1, Tables S3 and S4, p-values < 0.05). The coastal Tavoliere and the Salento areas have higher mean δ 13 C (‰) values, which is not surprisingly observed for the coastal Murge (Fig. 3b). Again, this suggests that although these Neolithic communities had equal access to marine resources, this did not translate into similar dietary habits. www.nature.com/scientificreports/ As expected, more negative values are found inland, supporting the lack of marine fish consumption. An explanation for the differences in δ 13 C (‰) values in the three different coastal areas, namely the Tavoliere, the Murgian coast and Salento could be found in the natural environment. The area of the Tavoliere and of Salento are characterised by extensive wetlands and coastal lagoons [52][53][54] . As already mentioned, wetlands are more productive than the open sea 41 and they are also more easily accessible, therefore representing an important potential source of subsistence for those living in these landscapes. On the contrary, the Bari coast (coastal Murge) comprised a more limited marsh area 52 , which probably resulted in the open sea being the only viable option for the procurement of aquatic resources. Curiously, the coastal area of the Tavoliere which reports among the highest δ 13 C (‰) values among the areas explored (Fig. 3b), where extensive marshes are present, is also one of those that show the lowest δ 15 N (‰) values (Fig. 3a). This seems to suggest that the fish consumed by these local communities is most likely composed of low trophic level organisms, such as benthic bivalves, which are abundant in estuaries and lagoons [55][56][57] .
The isotopic evidence suggests small variation of a common Neolithic diet. One possible reason might include proximity to the sea, estuaries and lagoons. Another might include the social role of food in different kinds of communities, for instance among the large groups of the ditched villages of the Tavoliere and of the inland Murge that may have held large social gatherings involving meat consumption (e.g. feasts) more often than more dispersed, smaller communities did. This is a preliminary hypothesis; we would suggest validating it by further investigation, including new isotope data from animals that could confirm that the differences we are observing are not related to environmental and climatic factors, and by integration with archaeological evidence of other kinds (e.g. pottery repertories; animal bones including fish and shellfish; botanical remains; food cooking facilities on sites; lipid analysis).
The role of animal products in the diet. One of the main issues in the discussion of the Neolithic diet is the role of animals in the food economy. Neolithic people in Italy, and throughout much of Europe, had a suite of domesticated animals, including cattle, pigs, sheep, goats, and dogs (though dogs seem to have been eaten rarely). They also had access to wild game, including deer, wild pig, and small animals. But how much did these contribute to their diet? Did they eat large quantities of animal products, or were they mostly subsisting on grains and vegetables, with meat reserved mostly for social occasions? Moreover, recent studies are suggesting that secondary animal products were important in the Neolithic Mediterranean basin 58,59 .
To estimate in calorific terms the contribution of animal products to the diet of Neolithic southeastern Italian groups, we applied a Bayesian mixing model to the isotopic data (see Materials and Methods for the description of the model). It was necessary to exclude marine fish contribution from the estimations on inland sites since there is no evidence of fish consumption from these sites (i.e., no access to marine resources), therefore results are here divided in two groups, reported in Fig. 4a,b for inland and coastal  Animal δ 13 C collagen d) Figure 3. Maps with the distribution of stable nitrogen (a,c) and carbon (b,d) isotope data from Neolithic southeastern Italy, of humans (a,b) and animals (c,d), created using the app IsoMemo, tool AverageR (https:// isome moapp. com/ app/ iso-memo-app). Data comes from this study and from Lelli et al. 8 (Table S5 and Fig. 4). The results suggest that animal products were not only available but also easily accessible during the Neolithic in southeastern Italy. Therefore, cereals, although contributing significantly, were not the only source of energy for these early farmers, as our models show. Our estimates do not differ from the conclusions of other studies that explore the Neolithic diet in other parts of Europe. For example, using stable isotope analysis of human bone collagen, it was observed that the majority of the diet of Neolithic communities in England and Scotland have an animal origin [60][61][62] and considerable consumption of animal proteins was suggested in Neolithic Switzerland 63 , Germany 64-66 , Hungary 67 , north and central France [68][69][70] , Spain 71,72 , Greece 73 and Croatia 74 . Recently, animal products have been estimated to contribute more than 50% of the diet in Neolithic Malta, probably due to a high reliance on dairy 75 . Unfortunately stable isotopes of bone collagen cannot be used to discriminate between different animal products (in this case, meat and dairy). To answer this question other types of analyses and other kinds of material need to be studied, for example the carbon isotopes of lipids from ceramics, and age at death profiles of animal remains at each site. Differences can be outlined between sites in some cases. For example, the diet of a female from Lama dei Peligni, one of the earliest Neolithic individuals recovered in Italy, is composed of ~ 55% of calories of animal origin, probably still influenced by the earlier Mesolithic tradition (Fig. 4a). This signifies that the typical 'Neolithic diet' was not spread uniformly in the area and that cultural and/or social variables might have influenced these communities, as also suggested by the regional analysis. We should acknowledge however the large uncertainties associated with these estimates since the two sources used in the model, C 3 cereals and C 3 -feeding animals, show very similar isotopic values (Table S5 and Fig. 4). Analyses at a higher resolution, such as Compound Specific Stable Isotope Analysis of Amino Acids (CSIA-AA) are needed in order to confirm and better define this scenario 76 . We should also remember that by analysing δ 13 C (‰) and δ 15 N (‰) values we are mainly exploring the protein component of diet 77 . This limits our understanding of plant consumption (notably cereals and legumes) where proteins are low, but they represent the staple for Neolithic people. To explore plant consumption, different types of analysis are needed, for example carbon isotope analysis of apatite 78-80 . Use of marine resources? A major puzzle for prehistoric food use is whether Neolithic people were using marine sources. An increasing number of isotopic studies from around Europe show that, even when they lived coastally in rich marine environments, Neolithic people lived principally on terrestrial resources 62,74,[81][82][83][84][85] . This was a major change from Mesolithic lifeways: overall, isotopic studies have demonstrated that a gradual decline in the exploitation of marine resources at the transition to the Neolithic is observed for the Mediterranean 82,86-88 , with fishing becoming more sporadic and mostly opportunistic 89 . It cannot be ignored that marine protein consumption is believed to be limited in the Mediterranean already in the Mesolithic compared to their Atlantic and Baltic counterparts 9,10,86,90 . However, a recent study using a higher resolution approach (CSIA-AA) has shown that aquatic resources were indeed exploited 91 . This gradual decline at the transition to the Neolithic may be ascribed to cultural practices or organisational factors 83,85,92 . Food procurement from an open sea environment clearly entailed greater risk and might have resulted in limited exploitation. As discussed by Galili et al. 93 , it might be that "fishing was a low preference mode of production, to which Neolithic communities turned only once the quantity and/or quality of terrestrial resources were reduced or impaired". To explore the question in Neolithic southeastern Italy, we plotted all the data according to proximity to the coast, since both δ 13 C (‰) and δ 15 N (‰) would increase values according to the accessibility of marine resources (Fig. 5). What we observe is no substantial differences when we explore data in a chronological perspective (Fig. 5a) and a lack of correlation between both δ 13 C (‰) and δ 15 N (‰) and distance to the coast (km) (Fig. 5b,c, R 2 = 0.16 and R 2 = 0.089, respectively) indicating that marine environments were not exploited substantially and equally  Table S5). We should consider however that the marine fish consumed in those sites characterised by a lagoon environment could present more 13 C-depleted δ 13 C (‰) values compared to the marine fish specimens used in the model (which were the only available for the area of investigation and with a different chronology), and therefore the marine contribution is possibly underestimated. In general, what emerges is a certain degree of variability in the consumption of marine resources at coastal sites from Neolithic southeastern Italy.

Conclusion
This paper has shown the distinctive value of the analysis of a large isotopic dataset, combining new and previously published results, to evaluate regional patterns of food consumption. First, we observed regional variations throughout southeastern Neolithic Italy that are most likely not directly caused by climatic or environmental conditions but rather by how the humans adapted according to the local environment and climate. This suggests the possibility of regional variations in the Neolithic diet, even within a relatively small region with a homogeneous cultural background 94 . Second, Bayesian statistics was used to evaluate the contribution of animal products in the diet of Neolithic people from the area. The results suggest a certain degree of variability across the assemblage and that animal products were certainly not of minor importance in the diet (40% on average of the total calories have an animal origin), implying that cereals and legumes were not the only source of energy for Neolithic communities from southeastern Italy. Finally, we explored marine fish consumption in the area. Our dataset suggests that marine resources were only minorly exploited but that this is likely an underestimation. Some variability in marine fish consumption was observed, and this might be related to differences in the environment (the presence of lagoons makes it easier to access fish) or cultural preferences. These results are preliminary and need to be reinforced by new material (plant and fish remains in particular, but also more terrestrial animals) and by higher resolution analyses, such as CSIA-AA, which has the potential to better define the differences observed, discriminate and quantify the contribution of terrestrial animals, plants and marine fish in the diet of Neolithic communities 76 . Nonetheless, a general picture of 'local' traditions within a larger model appears to be confirmed 8,45 . People adopted diversified cuisines, according to a shared individuals also differentiating between coastal (orange) and inland (light blue) sites; terrestrial animals from this study and the literature 8,45 are also plotted (crosses) as well as marine fish from the Adriatic 8 (a). Correlation between δ 13 C (‰) and δ 15 N (‰) and distance to the coast (km) is explored in (b) and (c), respectively. The grey bands represent the 95% confidence interval bands. Data plotted comes from this study and from 1,8,44,45  www.nature.com/scientificreports/ habitus, that might have been specific to a network of sites in small areas (e.g., the Murge region or the Tavoliere and Salento). The heterogeneity of dietary habits between communities living in similar environments suggests that socio-cultural factors might have played a key role in food choices and/or farming and herding practices, regardless of local availability.

Materials and methods
Collagen extraction. For the majority of the samples, collagen extraction for isotopic analysis followed the Longin's 95 method. In brief, cortical bone (ca. 0.5 g) was cleaned by sand abrasion and demineralized in 0.5 M aq. HCl at 4 °C for several days. The samples were then rinsed to neutral pH. Due to the poor preservation of the human remains from Titolo-Palese, ethylenediaminetetraacetic acid (EDTA) was used instead of HCl following Tuross 96 . These samples were decalcified in 50 mL 0.5 M EDTA at pH 7.4 for one to four weeks, changing the solution every 4 days. Once decalcified, samples were washed with deionised water 14 times, and left in water overnight. All samples were gelatinized in pH 3 water at 75 °C for 48 h. The collagen solution was filtered off with 5-8 μm Ezee filters, then frozen, and freeze dried. Stable carbon and nitrogen isotope ratios were measured using an automated elemental analyser coupled in continuous-flow mode to an isotoperatio-monitoring mass-spectrometer. The majority of the analyses were carried out at the Godwin Laboratory, University of Cambridge. Instrument used was a Costech Elemental Analyzer coupled to a Thermo Finnigan MAT253 Mass Spectrometer. Triplicate reproducibility is less than 0.2‰ for both isotopes. The isotopic standards used are: International Atomic Energy Agency (IAEA) standards of caffeine and glutamic acid for carbon and nitrogen; in-house laboratory standards of nylon, alanine, and bovine liver standard for carbon, nitrogen, and atomic C:N ratios. The samples from Titolo-Palese were analysed at SUERC. Here, stable carbon and nitrogen isotopic compositions were determined on a Delta V Advantage continuous-flow isotope ratio mass spectrometer coupled via a ConfloIV to an IsoLink elemental analyser (Thermo Scientific, Bremen) following Sayle et al. 97 Table S1. Previous isotopic measurements on Neolithic humans and animals are reported in Table S2.
Chronological and geographical assessment. Humans and animals were classified as 'Early' 'Middle' or 'Late' Neolithic following radiocarbon dates when available, and material culture when they were not (Table S1) 98 . The limited availability of radiocarbon dates for many prehistoric Italian sites is in part a result of well-developed material culture typologies; combined, they provide a good estimation of phasing and allow us to explore dietary variability throughout time. We also provide a new radiocarbon date of one sample of human bone (OCH2) from Diga di Occhito, dated to 5060 ± 40 BP (Beta-288147: 3962-3715 cal BC, 95.4%), placing the site within the Late Neolithic 99 (Fig. S3).
Coordinates of the sites were provided by the archaeologists or derived from previous publications. Where these were not available we used the coordinates of the closest town/village or landmark from the site.
Plotting and statistical analysis. Scatter plots, descriptive statistics and statistical tests were carried out using RStudio, R version 4.0.3 using the packages dplyr, tidyr, Hmisc, and ggplot2.

Regional analysis and mixing models.
To answer the archaeological questions, we used two different tools from the open access application IsoMemo (https:// isome moapp. com/ app/ iso-memo-app). AverageR was used to explore regional variability of δ 13 C (‰) and δ 15 N (‰) values as previously described 47,48 . We used the default options with the exception of the smooth type (planar), extrapolation behaviour (constant) and we used the Bayesian model with the number of MCMC iterations set to 5000 for the analysis of the data. ReSources was used instead to quantitatively evaluate the diet of Neolithic individuals using a routed and concentrationdependent model 100,101 . We used terrestrial animal products, C 3 cereals and marine fish as possible food sources for those individuals from archaeological sites located in proximity to the coast (km from the coast ≤ 15) while we only relied on terrestrial animal products and C 3 cereals for those considered inland sites (km from the coast > 15). δ 13 C (‰) and δ 15 N (‰) values of the animals from this and previous studies were used to represent the "animal products" category after having checked that the values were uniform across southeastern Italy using regional analysis. As for the 'C 3 cereals' source, since these are dramatically lacking from the area of investigation, we relied on published wheat and barley δ 13 C (‰) and δ 15 N (‰) values from a Greek Neolithic context 102 . The analysis of carbon and nitrogen stable isotopes of local plant remains is essential 103 , and here we urge for this. δ 13 C (‰) and δ 15 N (‰) values of local marine fish specimens from Lelli et al. 8 were used to represent the 'marine fish' category. Unfortunately, the marine fish used belongs to an older chronology but it represents the only opportunity to use a local baseline for this food category. The sources δ 13 C (‰) and δ 15 N (‰) values were corrected to represent the actual values in the tissues consumed using the recently updated offsets from Soncin et al. 76 . Uncertainties associated with the source values are 1SD. The offset Δ 15 N collagen-diet was set at + 5.5 ± 0.5‰, with 100% contribution from protein, and the offset Δ 13 C collagen-diet was set to + 4.8 ± 0.5‰, with www.nature.com/scientificreports/ 74 ± 4% contribution from protein and 26 ± 4% from lipids and carbohydrates 104 . Concentrations expressed as dry weight (%) are the same of those reported in Soncin et al. 76 using the USDA National Nutrient Database for Standard Reference (available at https:// fdc. nal. usda. gov/) with uncertainties being the standard error. As for model parameters, we used the default options with the exception of the following: model type: Individual targets (no shared info); Source contribution distribution: deselected "Optimal objective prior"; Covariates model: fixed intercept (cat. vars), fixed slope (num. vars).