Compound specific isotope analysis of lipid residues provides the earliest direct evidence of dairy product processing in South Asia

The early evidence of domesticated animals and human–animal interaction in South Asia can be traced back to the seventh millennium BCE; however, our understanding of their use is incomplete and limited to the analysis of animal bones from archaeological sites. By the third millennium BCE with the emergence of the Indus Civilization, cattle and water-buffalo became the primary domesticates and outnumbered any other animals at the majority of the Indus settlements. Based on the analysis of skeletal remains and ethnographic data, a number of studies have suggested that cattle and water-buffalo were utilized for their meat, dairy, hides, and other labor-oriented jobs. While some of these claims are backed by empirical data, others are primarily discussed as hypotheses, for example, the exploitation of dairy. In this paper, by analyzing the absorbed lipid residues from fifty-nine ceramic sherds recovered from an agro-pastoral settlement that was occupied during the peak of the Indus period around mid- to late third millennium BCE, we provide the earliest direct evidence of dairy product processing, particularly from cattle and possibly from some water-buffalo. By providing direct evidence of animal product processing, we identify the use of primary domesticated animals and other resources in the diet during the Indus Civilization.

. Highlighted map of the Indus Civilization showing the location of major settlements and settlements that are mentioned in this paper. The black dashed lines indicate the region under study and the red dot indicates the site under study. The map was created using QGIS Desktop version 2.18.14 (www.qgis.org), and raster and vector map data was acquired from Natural Earth (https ://www.natur alear thdat a.com/), which is available in public domain. The design idea with permission was adapted (Fig. 1, page no: 3) 17 .
Scientific RepoRtS | (2020) 10:16095 | https://doi.org/10.1038/s41598-020-72963-y www.nature.com/scientificreports/ originating from animals that predominantly consumed C 3 type vegetation vs. animals that predominantly consumed C 4 type vegetation 41,42 . In recent years, the analysis of lipid residue from archaeological vessels has been successfully used to determine the processing of other products as well, such as aquatic product processing 43 , cheese-making 44 , processing of plants to produce alcoholic beverages 45 , and the type of oil used in lamps 46 . Lipid analysis has also helped in the identification of resinous materials that have been used as adhesive and waterproofing layers on ceramic vessels [47][48][49][50] . This tool has been successfully used to identify the sources of organic materials in wall paintings 51,52 and in ashy deposits from archaeological settlements 53 .
Archaeological data from the settlement under study. For the present study, the settlement of Kotada Bhadli (23°20′N; 69°25′E), which is located in the Nakhatrana Taluka of District Kachchh, Gujarat was selected, as previous research by the primary author suggests that a sedentary to semi-sedentary form of animal husbandry was the primary occupation at this settlement 54,55 (Fig. 1). This site measures around 3.11 ha and is surrounded by a settlement wall. The excavations from 2010 to 2013 have unearthed a central residential complex with ten interconnected rooms, and the recent AMS dates suggest that it was occupied between 2300 and 1950 BCE 55,56 (Fig. 2). The preliminary zoo-archaeological report indicates that cattle, possibly some waterbuffalo, goats and sheep were the primary domesticated animals consumed at this settlement, along with Sus species, although the identification of the Sus as domesticated is ambiguous 57 . A similar distribution of domesticated animals can be observed at other settlements in this region during the Indus period 13,17,[58][59][60] ; however, water-buffalo were possibly not a major domesticated animal in Kachchh 17,61 (except perhaps at Dholavira 61,62 ). The strontium isotope values from the tooth enamel of the primary domesticated animals at the site of Kotada Bhadli indicate that cattle, water-buffalo, goat, and sheep were possibly raised locally, and the carbon isotope values indicate that human-induced foddering played a major role in the rearing of these domesticates 54 .
Animal rearing at Kotada Bhadli and its possible effect on lipid residue data. A complete zooarchaeological study of animal remains at the settlement of Kotada Bhadli is underway at the time of writing. A preliminary study of twenty individual tooth samples suggests that most of the cattle and water-buffalo died at an older age, suggesting production of secondary products, whereas the majority of goat/sheep died when they were young, indicating their use primarily for meat; the sole sample identified as possibly sheep also died at an older age 54 . This preliminary study is based on a limited number of samples and may not be representative of the entire assemblage from the settlement; however, such a kill-off pattern is similar to those found at other contemporary settlements in the same region, where the majority of cattle/water-buffalo lived into adulthood and the majority of sheep/goats were slaughtered when young, with some kept alive to maintain herd numbers [16][17][18]22 . Such a pattern would not result from the exploitation of cattle/water-buffalo exclusively for dairy products. If this  www.nature.com/scientificreports/ were the case, one would expect to see the majority of young male cattle and water-buffalo slaughtered at a young age, in order to maximize milk production by maintaining a herd consisting primarily of adult females 17,61,63,64 . Instead, it has been argued that these culling patterns observed in Gujarat indicate that a large number of male cattle (bulls and bullocks) and possibly some water-buffalo were kept alive for traction and for labor-oriented jobs as well as for selective exchange of animals between settlements during the Indus period 17,61 . Analysis of stable carbon isotopic ratios (reported as δ 13 C (enamel) ) of biogenic remains (tooth enamel) of herbivorous animals can be used to distinguish between diets based on plants that follow a C 4 photosynthetic pathway or a C 3 photosynthetic pathway 65,66 . Prior to the Industrial Revolution, δ 13 C (enamel) values between − 11.5‰ and − 6.5 would indicate a C 3 dominated diet, whereas, for animals consuming a C 4 dominated diet, the δ 13 C (enamel) will be around 2.5‰ [66][67][68] . At the settlement of Kotada Bhadli, the δ 13 C (enamel) values from twenty individuals suggest that cattle/water-buffalo and the only possible sheep varies from − 2.9‰ to 1.5‰ with a median value of 0‰, whereas for goat/sheep, it ranges from − 10.7‰ to − 2.3‰ with a median value of − 5.5‰ 54 . The results for biogenic isotope data of tooth enamel of cattle, water-buffalo, goat, and the one animal idenfied as likely a sheep at the site of Kotada Bhadli indicate that primarily cattle/water-buffalo and the possible sheep consumed a varied degree of agricultural fodder that followed a C 4 photosynthetic pathway, along with occasional intake of C 3 vegetation; goats/sheep primarily consumed vegetation that followed a C 3 photosynthetic pathway 54 . Similar foddering practices for cattle/water-buffalo and goat/sheep has been observed at the neighbouring settlements of Bagasra, Shikarpur and Jaidak 22,90 that were also occupied during Indus period. Cattle/water-buffalo and goat/ sheep are often conflated in zooarchaeological studies, and it is difficult to individually identify them in general based on fragmentary skeletal remains. While the detailed analysis of archaeozoological material of the site is not yet complete, the preliminary zoo-archaeological study at Kotada Bhadli and at other nearby settlements also grouped these animals as cattle/water-buffalo and goat/sheep 13,17,22,57,59 .
Similarly to what is observed for tooth enamel, the δ 13 C values of C 16:0 and C 18:0 fatty acids from lipids would reflect the δ 13 C composition of the primary food sources of the animal, as the δ 13 C values of fatty acids in herbivorous animals reflects the δ 13 C values of both carbohydrates and fatty acids of plants that they consume. Many studies have used observed ranges of the δ 13 C values of C 16:0 and C 18:0 fatty acids to differentiate food sources based on C 3 and C 4 directly 37,[39][40][41][42] . Due to such differences in the carbon isotope values between C 4 and C 3 plants 69 , animals primarily consuming C 4 type vegetation would produce enriched δ 13 C values of C 16:0 and C 18:0 fatty acids for both dairy and adipose fats, compared to animals predominantly consuming C 3 type vegetation 42 . Thus, measuring the δ 13 C values of C 16:0 and C 18:0 fatty acids can indicate the relative consumption of C 3 and C 4 vegetation. If cattle and/or water-buffalo were used for dairy exploitation, and if dairy was a major dietary constituent during the Indus period, we should be able to identify it based on the compound specific isotope analysis of the most abundant fatty acids.

Material and method
Samples. Fifty-nine uncleaned pottery fragments from throughout the settlement of Kotada Bhadli were collected from both the residential area and near the settlement wall during the excavations of 2011-2012 and 2012-2013 (see Supplementary Information table S1 and Fig. 2). The area near the wall was used to dump occupational debris during the site's occupation. The majority of sherds collected were from cooking vessels and various type of bowls; fragments of a perforated jar, pots, ladles, lids and unidentified coarse red-ware vessels were also included in this sample (Fig. 3, Supplementary Information table S1 and S2).
Upon excavation, these pottery fragments were immediately transferred into paper bags without washing and were stored for analysis. Once in the lab, any sediment and/or calcite depositions on the pottery surfaces were removed using an abrader drill bit. No water was used in the process of cleaning; prior to sampling, any dust adhering to the surface of the sherds was removed using pressurized air flow. Approximately 5 gm of cleaned ceramic was powdered using an agate mortar, and stored in aluminum foil that was previously baked at 450 °C. All the tools used for cleaning and powdering were cleaned after each use, using RO water and organic solvents.
Extraction. One-step acidified methanolic extraction 70 was used as the primary method to extract absorbed organic residues from all the 59 samples. Along with this, 5 samples were analyzed using conventional chloroform-methanol extraction and BSTFA derivatization 71 (see Supplementary Information for the detailed methodology) in order to identify the possible presence of compound lipids and any underivatized hydroxyl groups that the rapid and more aggressive acidified method may have failed to identify 70 . These samples were chosen based on the fact that they had extractable fatty acids both higher and lower than the settlement mean, and in the acid extracts of those samples, the distribution of fatty acids and the ratio between C 16:0 and C 18:0 fatty acids indicated the presence of plant oil, animal fat and aquatic fats. Results published elsewhere 55 suggest no preservation of any complex lipids such as glycerids, sterols, waxes and alcohols in the conventional extract; the rapid and efficient one-step acidified methanol extraction 29,70,[72][73][74][75] was therefore chosen over the conventional method because it produced high lipid residue yields from the very old and poorly preserved ceramic sherds of Kotada Bhadli 55 . Briefly, for each sample, 2 gm of ceramic powder was weighed and transferred into a 40 ml glass vial (vial 1), and 5 ml of MeOH-H 2 SO 4 (2% v/v.) was added and left for one hour at 70 °C, while shaken every 10 min to extract fatty acids from ground ceramic powders. The pH of the extraction was monitored to maintain a pH level < 3. After the extraction, the supernatant was transferred into another 10 ml vial (vial 2), and 2 ml of DCM extracted double distilled water was added, then 3 ml of hexane was added to vial 2, vortexed, and the hexane layer with Fatty Acid Methyl Ester (FAME) was transferred to another 10 ml vial (vial 3). Following this, 3 ml of hexane was added to vial 1 to extract any lipids not fully solubilized by the methanol solution. The hexane layer from vial 1 was then poured into vial 2 and whirl mixed, and then was finally transferred to vial 3. This step was repeated thrice, and all the hexane layers were mixed together in vial 3; finally, it was taken to  www.nature.com/scientificreports/ a virtual dryness under a gentle stream of nitrogen. It was re-dissolved and following a standard protocol 73,75,76 , prior to the GC-MS and GC-IRMS analysis 20 μl of n-triacontane (1 mg/ml) was added as an internal standard for quantification purposes, and then the contents of vial 3 were transferred into a 300 μl insert for GC-MS and GC-IRMS analysis. As our aim was to get high yields of fatty acids especially for compound specific isotope analysis, we did not analyze all of our samples using conventional extraction, neither did we derivatize our acid extracts to recover high polar compounds. Due to the limitation of direct transesterification, however, any study wishing to extract high polar compounds, such as alcohols, require that an aliquot must be derivatized with BSTFA prior to GC-MS analysis.
GC-IRMS analysis. The GC-IRMS analysis was also carried out at McMaster University, Canada. Based on the concentration of C 16:0 and C 18:0 fatty acid methyl esters, 21 ceramic sherds were selected for the GC-IRMS analysis. As a precautionary cleaning step prior to GC-IRMS analysis, a secondary liquid chromatography separation of FAMES was carried out using silica gels and a number of organic solvents of increasing polarity 37,77 , and the fractions containing hydroxy fatty acids were selected for the GC-IRMS analysis. Samples were either concentrated or diluted for GC-IRMS analysis. The gas chromatographic analysis was performed on an Agilent 6890 GC coupled with a Thermo-Finnigan DeltaPlus XP isotope ratio mass spectrometer via a Conflo-III interface. To achieve a better separation between unsaturates and saturates, the sample separation was performed on an HP-88, 100 m × 0.25 mm with a 0.20 μm thickness column. Injection was achieved through a splitless injector at 310 °C with an injection volume of 2 μl. The column was connected to a 1 m deactivated pre-column. Helium was the carrier gas. The temperature program was 80 °C held for 1 min, ramped to 175 °C at 10 °C/min and held for 12 min, ramped to 190 °C at 2 °C/min with a 10 min hold, finally ramped to 240 °C at 10 °C/m with a hold for 15 min. Acquisition and data analysis were performed using Isodat 2.03 software. Analytical accuracy was confirmed via isotopically characterized standards run before and after each set of samples. Accuracy and precision on triplicate sample analysis was between 0.1 and 0.7 per mil (2 sigma). The isotope values of the samples have been corrected by the known δ 13 C value of the methanol used for the extraction. The mathematical formula below has been used for the correction: Here C FA is the corrected δ 13 C value of the fatty acids, C n represents the number of carbon atoms in the chain, δ 13 C FAME is the observed carbon isotope value of the fatty acids and δ 13 C MeOH is the known carbon isotope value of the methanol used for the extraction of fatty acids.
This research has been designed to identify the earliest usage of dairy from ruminant animals as well as the utilization of adipose fats from ruminant and monogastric animals in the diet during the Indus period. Therefore, twenty-one samples out of 59 samples were selected for GC-IRMS analysis, as they had indications of the preservation of animal fats. Following the most common practice in archaeology 40,70,[72][73][74] , the samples with higher concentrations of C 16:0 and C 18:0 fatty acids, those indicative of animal fats, were selected. Along with the high abundance of C 16:0 and C 18:0 fatty acids, the presence of odd chain fatty acids (C 13:0 to C 19:0 ) and branched chain fatty acids (C 15:0br to C 18:0br ) were used to tentatively identify ruminant fats 40,70 (see Supplementary Information table S1). A low abundance of C 18:0 fatty acid as compared to C 16:0 fatty acid, which is often associated with the presence of plant oils 40,78 , were observed in six samples ( Supplementary Information table S1). These six samples were not included in the compound specific isotopic analysis as we focus on animal utilization at the settlement of Kotada Bhadli.
One out of the 59 samples, a red-ware untreated bowl, contained some evidence of aquatic fat (this sample was excluded from the isotopic analysis). This tentative identification was based on the presence of isoprenoid acids (phytanic acid, pristanic acid, 4,8,12 TMTD), a low concentration of ω-(o-alkylphenyl) alkanoic acids, and long chain mono-unsaturated fatty acids particularly C 20:1 and C 22 Fig. 1). Interestingly, a wide range in the δ 13 C values of C 16:0 and C 18:0 fatty acids and distinct groupings (Fig. 5) were observed in the Kotada Bhadli samples. The biogenic carbon isotope data of tooth enamel from major domesticated animals 54,55 earlier suggested that due to selective foddering practices at Kotada Bhadli, one group of domesticated animals (cattle, water-buffalo and one likely sheep) ate a considerable portion of C 4 type vegetation, whereas, the other group of domesticated animals (indistinguishable goats/sheep) ate primarily C 3 vegetation. As we hypothesized earlier, this foddering practice has influenced the variation in the observed range of δ 13 C values of C 16:0 and C 18:0 fatty acids. The distinct groupings (Fig. 5) suggest that ruminant animal fats from the nine samples that produced δ 13 C (C 16:0 ) values between − 24‰ and − 30‰ were influenced by a C 3 rich diet, and in case of Kotada Bhadli, were possibly goat/sheep or other ruminant animals that predominantly consumed C 3 type vegetation. Ruminant fats from the remaining 12 samples that produced δ 13 C (C 16:0 ) values between − 22‰ and − 14‰ indicate the influence of C 4 in their diet, which were likely to be coming from cattle, some water-buffalo, and possibly from some sheep (based on the enamel isotope data from a single likely sheep from Kotada Bhadli 54 ), which is also expected because cattle and possibly some water-buffalo were the primary domesticated animals at the settlement of Kotada Bhadli 57 .
The Δ 13 C (C18:0-C16:0) values of the 21 samples selected for isotopic analysis varied between 4.9‰ and − 6.5‰, indicating the presence of both ruminant and non-ruminant adipose fat as well as ruminant dairy fat (Fig. 5). When Δ 13 C values are plotted against the δ 13 C (C 16:0 ) values (Fig. 5), it appears that all the ruminant dairy fats in our assemblage were exploited from cattle, some water-buffalo, and possibly from some sheep that consumed primarily C 4 type vegetation; adipose fats were coming from both sheep/goat and cattle/water-buffalo, as well as from monogastric animals, likely pig. Unfortunately, the modern foddering pattern in this region does not replicate the foddering pattern that we observed at the site of Kotada Bhadli 54,55 , limiting our ability to develop a site-specific modern reference collection. At present, C 3 vegetation and cash crops play a major role in animal foddering as compared to C 4 type vegetation, such as millets.
As suggested above, dairy residue from any ruminant animal that consumed predominantly C 4 type vegetation at this settlement may produce similar enriched δ 13 C(C 16:0 ) values, whether cattle, water-buffalo, or possibly sheep. However, sheep were never a primary domesticated animal at this settlement 57 or at the neighbouring settlements 13,17,22,59,61 , and those found in the eastern Indus regions are proposed as exploited for meat and wool 81 . Cattle and water-buffalo (depending on the region) were the primary domesticated animals, ranging www.nature.com/scientificreports/ from 60-90% of the assemblage during the Indus period, not only in Gujarat, but also in the other parts of the Indus Civilization 5,13,17,22,57,59,61 . It is therefore reasonable to argue that dairy, which may have been an integral part of everyday diet at Kotada Bhadli, was likely to be exploited primarily from cattle, and possibly from some water-buffalo, although occasional exploitation of dairy from sheep and goats is also a possibility. Due to the expansion of the Indus Civilization into vastly different geological and environmental zones and the possibility of local adaptations, our observations from Kotada Bhadli likely represent one of several Indus animal-exploitation strategies. By taking a slightly conservative approach and allowing for the possibility of mixing between dairy and adipose fats, as well as probable occasional consumption of deer meat, as suggested by the presence of uncommon deer skeletal remains at this settlement and from the neighboring sites during the Indus period 13,57,59 , the Δ 13 C adipose fat of ruminant animals can be extended up to − 4.3‰ 82 . Even with this adjustment, it is clear that there are at least five samples that produced negative Δ 13 C below − 4.3‰ and can be securely identified as dairy fat at Kotada Bhadli. It is important, however, to keep in mind that deer have never been considered to be a major constituent of Indus/Harappan diet and were likely to be consumed only occasionally. In addition, all samples indicating the presence of dairy fats have originated from animals that predominantly consumed varying degrees of C 4 type vegetation. In the case of Kotada Bhadli these would primarily be agricultural millets 54,55 , and it is highly unlikely that wild ruminant animals of that region consumed such a high proportion of agricultural vegetation.
Cooking techniques and multiple usage of serving vessels may have resulted in some mixing between animal fats and plant oils; little work has been done on cooking techniques, but cut marks on the bones from this region suggest that both stewing and roasting may have been a standard way of processing meat during the Indus period at least in Gujarat 16 . Mixing between animal fats can be observed in our samples (see Supplementary Information table S2); however, it is difficult to ascertain the mixing between animal fats and plant oils. Oil seeds used in this region all fall under the C 3 category of plants 9,13,15,83 ; they are unlikely to produce δ 13 C values of fatty acids similar to the values influenced by the C 4 photosynthetic pathway, which we have observed for all of our dairy and in some of the adipose fats primarily exploited from cattle/water-buffalo. Also, as previously observed 84,85 , the Δ 13 C (C18:0-C16:0) values of plant oils are generally indistinguishable from ruminant and non-ruminant adipose fats, but they are not similar to the values of dairy fats. On the other hand, cereal grains have low lipid content which can be easily overshadowed by animal fats 73,86,87 . Recent studies have pointed out that mixing between animal fats and plant oils can potentially affect the Δ 13 C values, based on simulated results 88,89 . While mixing of ruminant adipose fats with C 3 plant oils can never produce Δ 13 C values similar to dairy, the simulated mixing of C 3 ruminant fats with C 4 plants was shown to be able to produce dairy Δ 13 C values 89 . This is particularly crucial www.nature.com/scientificreports/ in this present scenario when various millets that are primarily C 4 in nature were possibly consumed by both human and cattle/water-buffalo in Gujarat, whereas goats/sheep were primarily raised on C 3 type vegetation 22,54,90 .
There are several lines of evidence that support that mixing of plant oils and animal fats were not responsible for the observed offsets in the Δ 13 C (C18:0-C16:0) values of dairy samples at the settlement of Kotada Bhadli. First, a thorough search for miliacin, a biomarker of broomcorn millet, did not reveal its presence in any of these samples. Further, the samples analyzed were selected such that the ratio between C 16:0 and C 18:0 and the distribution of fatty acids in samples chosen for isotopic analysis were consistent with animal fats rather than plant sources. But perhaps most importantly, due to the fact that plant oils generally contain high levels of C 16:0 and low levels of C 18:0 fatty acids, addition of plant oils would be expected to result in a shift of C 16:0 values but not in C 18:0 values resulting in a Δ 13 C (C18:0-C16:0) offset consistent with dairy. Noticeably both the C 16:0 and C 18:0 fatty acids of the sample indicating dairy consumption were enriched because of the C 4 dependent selective foddering of cattle/ water-buffalo at this settlement (Supplementary Information table S2 and Fig S1). In addition, because mammary glands in ruminant animals are unable to biosynthesize C 18:0 , whereas, a major portion of it in the adipose fats are due to de nevo biosynthesis from acetate, it is therefore likely that the δ 13 C values of C 18:0 fatty acids in adipose fats will be slightly enriched compared to the dairy fats of animals raised on the same diet 38 . This is what we observed for our samples, where the δ 13 C values of C 18:0 in cattle/water-buffalo adipose fats were enriched compared to cattle/water-buffalo dairy fats ( Supplementary Information table S2 and fig S1).
Based on these lines of evidence we feel that the potential mixing of C 4 plant derived FAMEs with animal derived lipids raised on C 3 type vegetation is unlikely to have occurred in these samples that we have securely identified as potential dairy fats.

conclusions
In this paper we have presented the earliest direct evidence of dairy product processing in South Asia. This evidence was derived from the absorbed lipid residues preserved in the ceramic assemblage recovered from Kotada Bhadli, a sedentary to semi-sedentary agro-pastoral settlement occupied from 2300-1950 BCE 55 , during the Mature/Urban Harappan period, also known as the Integration Era of the Indus Civilization. Our findings suggest that dairy was an integral part of the everyday diet of its residents; this dairy was explicitly acquired from cattle and possibly water-buffalo. Studies based on the analysis of artefacts 91 , bone morphology 23 and mortality profiles 17,61 have previously suggested that during the Indus period both cattle and water-buffalo were used extensively for traction as well as for their meat, milk, and hide. These large-bodied animals were viewed not just as commodities, but may also have played an important role in Indus ideology, as indicated by their depiction in seals, sealing/tokens, pottery, and terracotta figurines 92 . The antiquity of cattle and water-buffalo for dairy exploitation may date back to as early as the sixth millennium BCE, when they were first domesticated in South Asia 4 , necessitating further analysis of older materials.
The presence of dairy fat at the settlement of Kotada Bhadli was identified based on Δ 13 C plotted against δ 13 C (C 16:0 ) values which indicates that they were exploited from animals raised on C 4 type vegetation, and at Kotada Bhadli these were cattle/water-buffalo. Data from three other neighbouring settlements where analysis have been done also indicate that during the Indus period, cattle/water-buffalo were raised on C 4 type vegetation 22,54,90 ; however, the nature of their utilization at these neighbouring settlements remains unidentified. Along with the earliest evidence of dairy use from South Asia, in this paper we have also provided supplementary evidence to support the idea that adipose fats from cattle/water-buffalo, goat/sheep, and monogastric animals were processed for consumption at this site, in line with zoo-archaeological work by Chase at nearby sites 16,17,93 . Additional work in the region is likely to provide significant findings about the versatility of Indus diet and the role of these usually invisible products in regional and international trade networks that developed during the Indus Civilization and played a crucial role in its survival and expansion.