Long-Term Resilience of Late Holocene Coastal Subsistence System in Southeastern South America

Isotopic and molecular analysis on human, fauna and pottery remains can provide valuable new insights into the diets and subsistence practices of prehistoric populations. These are crucial to elucidate the resilience of social-ecological systems to cultural and environmental change. Bulk collagen carbon and nitrogen isotopic analysis of 82 human individuals from mid to late Holocene Brazilian archaeological sites (∼6,700 to ∼1,000 cal BP) reveal an adequate protein incorporation and, on the coast, the continuation in subsistence strategies based on the exploitation of aquatic resources despite the introduction of pottery and domesticated plant foods. These results are supported by carbon isotope analysis of single amino acid extracted from bone collagen. Chemical and isotopic analysis also shows that pottery technology was used to process marine foods and therefore assimilated into the existing subsistence strategy. Our multidisciplinary results demonstrate the resilient character of the coastal economy to cultural change during the late Holocene in southern Brazil.


Introduction
The Brazilian coast encompasses a wide range of tropical and sub-tropical ecosystems that have sustained human populations from the middle Holocene to the present day. The large shell mounds, or sambaquis, are a distinctive feature of this coastline, testament to large-scale exploitation of marine resources, from ,8,000 to ,1,000 calibrated years before present (cal BP). In southern Brazil some sambaquis reached more than 35 m high and contained hundreds of burials, post holes and faunal remains testifying the development of a complex social panorama [1]. The exploitation of aquatic (mostly marine) resources was an important subsistence activity at these sites [2] and must have drawn people to the coast. However indirect evidence reveals that the contribution of plants also appears to be important [3][4][5][6][7][8]. Sambaquis containing freshwater and land snail shells are also found along the courses of rivers and their distribution penetrates some distances inland. These ''Riverine sambaquis'' are the same age or even older than their coastal analogues (,10,000 to ,1,000 cal BP) [9] and occasionally finds of marine fauna at these riverine sites suggest some connection to the coast [10], [11], [12].
A dramatic change is seen in the archaeological record at ,1,500 cal BP with the abrupt cessation of large shell mound formation [13]. At this time it is thought that new populations from the southern highlands (known as the Taquara/Itararé tradition) [14] expanded to the coastal lowlands [15], likely driven by rapid population growth, increasing of social interaction and intensification in food production, involving maize and exploitation of pine forest (Araucaria angustifolia) [16]. The appearance of Taquara/Itararé pottery along the southern coast of Brazil therefore may mark a key turning point in exploitation of rich coastal ecotones, as prehistoric groups gained the knowledge and technology to develop new economic practices. However the extent to which the transmission (or imposition) of this new subsistence system transformed the indigenous coastal economy, and its capacity to adjust, persist and maintain its fundamental properties, is still a matter of debate [17].
While there is some evidence for consumption of new cultigens like maize by Taquara/Itararé groups [7] and increased consumption of terrestrial resources [18], marine fauna continue to be found at high abundance [19]. Similarly, while pottery is often assumed to be associated with the processing of new produced and foraged foods, with parallels in the southern Brazilian Highlands [16], there is no direct evidence of what it was used for. Our understanding of the diet and subsistence economy of the prehistoric inhabitants of coastal Brazil has been largely limited to traditional archaeological information, based on faunal, botanical and artefactual remains. In particular, the contribution of marine and terrestrial foods to the diet of both pre-ceramic and ceramic coastal populations of this region still remains largely unknown [11], [20], [21] and only few studies have considered the use of ceramics during this period [22], [23]. As a result, the impact of new economic and technological strategies on coastal adapted hunter-gatherers societies is not yet understood.
Here we report the results of an integrated study into the dietary variability of coastal and inland sambaqui populations. We analysed the stable carbon and nitrogen isotope composition of human bone collagen, a technique widely used to reconstruct paleodiets, and particularly for distinguishing marine versus terrestrial diets [24], [25]. We determined the stable carbon isotope signature of individual amino acids from bone collagen to identify different macronutrient constituents of diet [26][27][28]. We assessed the potential of bone mineral for isotope analysis in order to provide information on whole diet [24]. Finally we considered additional information regarding the diet of 'incoming' ceramic producing groups through the analysis of the organic contents of their pottery [29][30][31]. Hansel et al. [22] and Hansel and Schmitz [23] have already shown the potential for retrieving lipids from pottery in coastal Brazil, but here we report the first compoundspecific isotopic analysis of these artefacts.

Archaeological Setting
The archaeological records include four middle and late Holocene sites located in southeast (São Paulo) and southern (Santa Catarina) regions of Brazil (Fig. 1). These consist of one inland riverine site (Moraes) and two coastal sambaquis (Jabuticabeira II, Piaçaguera), along with a recently excavated ceramic coastal site (Galheta IV), dated to the time of the expansion of the highland groups to the coast. The sambaqui sites dated to the preceramic period show a wide range, but well-established chronology (,6,700 to ,1,700 cal BP) and offer a unique opportunity to elucidate dietary variability between coastal and inland mound builders. They also offer a valuable isotopic baseline for assessing changes in subsistence strategies associated with the spread of pottery technology (,1,500 cal BP). Preliminary stable isotope studies have already been carried out on human remains from some of these sites (Moraes, Jabuticabeira II) [11], [21] as well as other coastal sites in southern Brazil [20], [32]. Here we undertake the analysis of 106 human remains with the aim of greatly expanding the dietary isotope record for this region.
Moraes (MRS) is situated in the Ribera do Iguape Valley, at ,35 km from the São Paulo coast. The site forms a mound with a conspicuous concentration of land snail shells dated between 6,775-6,499 and 5,289-4,887 cal BP [33]. The mound was used mainly for funerary purposes [34] and 55 human individuals were recorded in the deposit [34], [35], along with a consistent amount of faunal remains mostly represented by terrestrial mammals, followed by amphibians, birds, and only few reptiles and fish bones [10], [11].
Piaçaguera (PCG) is located in the south-east coastal plain of São Paulo state, at ,12 km from the present day shoreline. PCG similarly provided a stratigraphic succession containing more than 80 human individuals, artefacts and faunal remains dated to 5,894-5,326 and 5,887-5,314 cal BP [35], [36]. Fish and molluscs dominate the assemblage, followed by some crustaceans, a few terrestrial mammals and carbonized seeds [35][36][37][38].
Jabuticabeira II (Jab-II), at ,7 km from the present day shoreline at Santa Catarina, is one of the best-studied sambaqui sites of Brazil [21], [39]. The stratigraphic sequence is composed of alternating layers of marine shells and fish bones [21]. Numerous human burials and the absence of compelling domestic activities point to a ceremonial function of the site from 3,137-2,794 to 1,860-1,524 cal BP [13], [40].
Finally, Galheta IV (G-IV) is a funerary site overlooking the Atlantic Sea, at ,20 km north of Jab-II, in Santa Catarina [41]. The site yielded 8 adult human burials, some of them partially cremated, associated with abundant marine faunal remains (fish, sea mammals, seabirds) and artefacts, including potsherds of Taquara/Itararé tradition. G-IV was 14 C dated between 1,304-1,140 and 913-739 cal BP, and it dates to the expansion of inland ceramic producers (southern Jê speakers) to the coast.

Human and Faunal Remains: Sampling Procedure and Ethical Statement
A total of 106 human individuals (different age and sex) from MRS, PCG, Jab-II and G-IV were sampled for isotopic analysis. In order to build a faunal isotopic reference for the region, 36 animal bone remains (terrestrial mammals, sea mammals, birds and fish) from these four sites were also selected for isotopic analysis (Tab. 1). Human samples were obtained almost entirely from the ribs, whereas fauna samples were from a range of different skeletal elements. All necessary permits were obtained for the described study Isotopic analyses were also conducted on modern fish (n = 10) caught using traditional fishing techniques and acquired at the central market of Florianópolis (Santa Catarina, S. Brazil). The use of modern fish specimens was carried out in strict accordance with the recommendations of the Brazilian Institute of Environment and Renewable Natural Resources -IBAMA. No In Vivo experiment was developed, thus ARRIVE guidelines (Animal Research: Reporting In Vivo Experiments) for this study is not applicable. The transport of modern specimens to the University of York was approved under the Convention on International Trade in Endangered Species of Wild Fauna and Flora -CITES-IBAMA (protocol nu 113508). Modern fish collagen d 13 C were corrected (+1.14%) for the global decrease of atmospheric d 13 C values [42]. We also incorporated d 13 C and d 15 N values from Plens [11] and De Masi [20], who report collagen isotopic composition of archaeological faunal remains from MRS and other coastal sites of southern Brazil.

Assessing Bone Collagen and Mineral Preservation
Bulk collagen d 13 C and d 15 N derive primarily from dietary protein, although macronutrients (carbohydrates, lipids) may variably contribute to collagen carbon, particularly for the nonessential amino acids in collagen [43]- [45]. The d 13 C of bone apatite instead reflects the total dietary pool of carbon ingested [46]. Therefore the combination of collagen and apatite d 13 C has been shown to give information on main energy and protein consumed several years prior to death [47], [48]. In archaeological contexts, however, the burial environment may impact the physical and chemical composition of bones in different ways [49], particularly through loss of collagen and alteration of bioapatite and the stable isotope signature of this fraction [50].
Assessment of collagen preservation in both human and faunal remains was carried out following the criteria proposed by van Klinken [51]. In addition, Raman spectroscopy studies were performed on randomly selected human bone samples from MRS (n = 9), Jab-II (n = 9), G-IV (n = 4) and PCG (n = 10) to assess diagenetic change to the mineral fraction, with modern lamb bone being used as a control [52]. To optimise the quality of the Raman spectra, the samples of bone were flattened and smoothed by gentle rubbing with fine-grade diamond paper. The Raman spectra were collected from the samples using an HORIBA XploRa instrument at 532 nm laser wavelength and under x100 magnification in confocal mode (NA = 0.9, with 2400 g mm-1 grating). Five spectra were collected from each bone specimen using 1s laser exposure at ,3.5 mW power at the sample, with each measurement averaged over 40 spectral acquisitions. The spectra were collected over 4 spectral windows to achieve a total spectral range of 200-3200 cm 21 . The software package IGOR Pro 6.32 was used to average, baseline correct, and analyse the Raman spectra using Gaussian peak-fitting procedures, with the n 1 carbonate peak (at ,1070 cm 21 ) de-convoluted according to published protocols [53].

Collagen Extraction and Isotope Analysis
Collagen preparation follows the protocol described in Craig et al. [54]. Before isotopic analysis, lipids were removed from modern fish bones with dichlormethane:methanol (2:1, x3).
Between 300 and 500 mg of cleaned human and animal bones were used for collagen extraction. Samples were agitated in 8 ml of 0.6 M hydrochloric acid at 4uC to demineralize. Once demineralization had occurred the samples were removed from the acid and washed with ultrapure water three times. The samples were gelatinised in pH 3 hydrochloric acid and maintained for forty-eight hours at 80uC. The gelatinised samples were then ultrafiltered and a .30 kDa fraction was lyophilised. Duplicates (1 mg) were measured using a continuous flow isotope ratio mass spectrometry Thermo Finnigan Delta Plus XL in the Department of Archaeological Sciences of the University of Bradford (UK), to determine the d 13 C and d 15 N values. The results are reported using the delta scale in % relative to internationally accepted standards, V-PDB and AIR respectively. Analytical error, calculated from repeated measurements of each sample and measurements of the bovine control from multiple extracts, was ,0.2% (1s).

Analysis of Individual Amino Acids in Collagen
Stable carbon isotope analyses were performed on collagen amino acids isolated from randomly selected coastal individuals from Jab-II (n = 10) and G-IV (n = 7). Approximately 1 mg of collagen was hydrolysed under vacuum in amino acid free 6 M hydrochloric acid (1 ml) at 110uC for 24 hours. After hydrolysis the samples were dried in a rotary vacuum concentrator and stored at -20uC until analysis. Prior to isotopic analysis, the samples were redissolved under sonication in MilliQ water with the addition of an internal standard (2-amino-isobutyric acid).
Instrumental analysis was carried out using Thermo Scientific Liquid chromatography isotope ratio mass spectrometry (LC-   This study IRMS) system consisting of an Accela 600 pump connected to a Thermo Scientific LC Isolink and a Delta V Plus Isotope Ratio Mass Spectrometer housed at the La Trobe Institute for Molecular Sciences (LIMS, La Trobe University, Melbourne, Australia). LC-IRMS analysis of single amino acid fractions of collagen hydrolysates were carried out using a three phase LC/IMRS method similar to that described in Smith et al. [55], with the exception that a Primesep A (SCIELC) column (2.16250 mm, 100 Å , 5 mm) was used. This is a narrower column to that used by Smith et al. [55] and thus lower LC flow rates were used. Conditioning runs were made at 110 mLmin 21 using phases ratios of 85B:15C-95B:5C, analytical runs were made using flow rates of 60-80 mLmin 21 and oxidation reagent flow rates were set at 35 mLmin 21 each. Approximately 10 mg of amino acid hydrolysate (on column) was used for each run. Amino acid peak d 13 C values were measured against CO 2 gas pulses throughout the run (d 13 C VPDB = -2.8%) calibrated against international standard USGS-40 L-Glutamic Acid (d 13 C VPDB-LSVEC = -26.460.04%). In house standard runs were made during the sample runs to monitor measurement quality.

Organic Residues Analysis
Molecular and isotopic analysis of organic residues absorbed into porous vessels or preserved in surface deposits offer valuable information concerning pottery use [29], [30], [54], [56][57][58]. Lipids were extracted (and methylated in one-step) from 14 potsherds according to protocols reported in Craig et al. [31], [59]. Briefly, after cleaning the surface, methanol (4 ml) was added to powdered ceramic samples (70 to 240 mg) and the mixture was sonicated for 15 min and then acidified with concentrated sulphuric acid (800 ml). The acidified suspension was heated in sealed tubes for 4 h at 70uC and then cooled, and lipids were extracted with n-hexane (2 ml63). The extract was dried under a gentle flux of nitrogen and internal standard (n-hexatriacontane) was added before the direct analysis by gas chromatography/mass spectrometry (GCMS) at the University of York (UK).
Stable isotopic analysis of n-hexadecanoic (C 16:0 ) and noctadecanoic (C 18:0 ) acids from 11 extracted lipid samples were performed using a gas chromatograph (GC) coupled to a combustion isotope ratio mass spectrometry (GC-C-IRMS) at the University of Liverpool (UK) following the protocol reported in Craig et al. [59]. Instrument precision on repeated measurements was 0.2% (s.e.m.).
Charred residues of food were preserved in the internal part of 6 ceramic potsherds. Samples (3-7 mg) were removed and subsamples (1 mg) selected for carbon and nitrogen isotopic analysis at the University of Bradford by using the same IRMS procedure as for bone collagen [56].

Statistical Analysis
The proportional contribution of different food sources to human diet (based on stable isotope values) was estimated using a Bayesian mixing model in SIAR V4 (Stable Isotope Analysis in R) [60]. The bulk d 13 C and d 15 N values of faunal remains (Tab. 1) was used in the model calculation after adding trophic enrichments of 1% and 4% for carbon and nitrogen respectively [61]. Bulk d 13 C and d 15 N were also analysed using the parametric One-Way Anova test in the software PAST 2.13 [62], after checking for normal distribution (Shapiro-Wilk) and using a statistical significance probability threshold of a = 0.05. Available radiocarbon ages were calibrated with OxCal 4.2, using the Southern Hemisphere curve SHCal04 [63], [64].

Bone Apatite and Collagen Preservation
Raman analysis revealed substantial alteration to the mineral phase of all of the archaeological samples that were analysed (Tab. S1, Fig. 2). Substitution of non-biogenic carbonate for biogenic phosphate, or reduction in biogenic phosphate, is indicated by the increased carbonate n 1 (C) to phosphate n 1 (P) intensity ratios in archaeological bones compared to the modern control [65], [66]. An increase in crystallinity due to mineral alteration in the archaeological specimens is expected [49], [67] and confirmed in our Raman data by a decrease in the full width at half maximum (FWHM) of the n 1 P band (at ,957 cm 21 ) relative to the modern bone sample [65], [66], [68]. Increasing crystallinity in diagenetically altered bone has been linked to the loss of collagen [49], which is further corroborated in our data by reduced CH/P ratios compared to the modern bone sample. Here, CH refers to the collagen band (i.e. CH stretching) at ,2933 cm 21 . The CH/P trend is also in agreement with the Raman spectroscopy results of Edwards et al. [69], which demonstrated reduced intensity in the collagen modes in human bones collected at JAB-II. Finally, the large fluorescent background in the archaeological samples (Fig. 2) is attributed to spectral emission from luminescent ions incorporated into the bone lattice due to diagenetic alteration [65].
Despite widespread alteration to the bone mineral phase, most of human burials (82 out of 106) matched the criteria for adequate collagen preservation [51]. Collagen yield and C:N ratios individuals from MRS (n = 15), PCG (n = 11), Jab-II (n = 47) and G-IV (n = 6) range from 0.9 to 7.7 wt% and from 3.2 to 3.5 respectively (Tab. SI2). Three individuals from PCG and one from G-IV contain ,1% collagen but still had acceptable C:N ratios and their d 13 C and d 15 N values are coherent with the other humans. Similarly, most of the animal samples from the study sites (25 out of 36) had adequate (.1%) collagen yields (Tab. 1). Of these, one peccary (Tayassu sp.) and one brocket (Mazama sp.) from MRS show collagen yield of 0.5 and 0.6 wt% respectively, but again the C:N matches the criteria for unaltered collagen. In summary, whilst the bone mineral fraction is unlikely to preserve a vital isotopic signal, the collagen yield and C:N composition  suggests acceptable collagen preservation for the majority of the samples [51], [70]. Human d 13 C and d 15 N values show strong positive linear correlations (r = 0.95; R 2 = 0.92; p,0.001; Tab. S2) and fall between the end-points derived from correcting the observed marine and C3 terrestrial fauna for isotopic fractionation (Fig. 3). Therefore the human isotope values can be largely explained by direct routing of both carbon and nitrogen from dietary protein to collagen, which implies that the diets contained sufficiently high protein [71]. The d 13 C and d 15 N values differ significantly among sites (p,0.001; Tab. 2). Lower d 13 C and d 15 N values were observed in inland individuals from MRS, as opposed to higher  values of coastal pre-ceramic and ceramic individuals from Jab-II and G-IV, in agreement with preliminary isotopic studies [11], [21]. Individuals from PCG, instead, exhibit d 13 C and d 15 N values consistent with a mixed marine/C3 terrestrial diet. The proportional contribution of different marine and C3 terrestrial animal resources to each individual's diet can be crudely estimated through linear interpolation between the marine and terrestrial end-members. We also used a Bayesian model which generates possible dietary solutions from multiple dietary source categories. This model predicts that inland peoples acquired . 90% of their protein from C3 terrestrial resources, whereas people on the coast were assimilating protein mainly from fish, up to 80%, along with some contribution from seabirds and sea mammals (Fig. 4). A large isotopic variability, however, was detected in coastal groups, in particular at Jab-II (Fig. 5) but it seems not to be related to sex and age (Tab. S3).

Bulk Collagen Stable Isotope Analysis
It is important to note that these estimations refer to the protein contribution to total dietary protein (by dry weight) and not the contribution to total diet. In this case, the bulk isotopic analyses are highly insensitive to the other dietary components, such as carbohydrate and lipid, which must have been consumed to some extent in order to avoid protein poisoning [72].

Carbon Stable Isotope Analysis of Single Amino Acids
Stable carbon isotope values were obtained from 15 amino acids corresponding to 97.5% of the carbon atoms in collagen (Tab. S4). Mass balance calculations were used to estimate the d 13 C of whole collagen from the measured individual amino acid d 13 C values. These estimated values were strongly correlated with the observed bulk d 13 C values (R 2 = 0.98) and the offset between the estimated and observed measurements was ,1% in all cases. The d 13 C of individual amino acids were strongly and positively correlated with both the d 13 C (R 2 = .0.8) and d 15 N (R 2 = .0.7) values of whole collagen for both marine and C3 terrestrial consumers, confirming that both dispensable and non-dispensable amino acids were largely derived from a dietary protein source.  Honch et al. [28] devised a method of interpreting dietary intake of individuals using a plot of d 13 C values of phenylalanine (Phe) and valine (Val) of bone collagen hydrolysates. Following this method, the d 13 C values again confirm the two dietary groups; those with mixed marine/C3 terrestrial resource diets (Jab-II, burials 17C, 24A, 102 and G-IV, burial 7) and all other samples as high marine protein (HMP) consumers (Fig. 6).
There are some interesting observations concerning the amino acid d 13 C values of burial 7 at G-IV. In general the values for both dispensable and non-dispensable amino acids (Tab. S4) for this individual are intermediate between the mixed marine/C3 terrestrial consumers and the HMP consumers, indicating a mixed diet. Unexpectedly however, this individual has a lower d 13 C value for alanine (Ala) compared to the others (Fig. 7). In the context of the other amino acid values, the low alanine value is difficult to explain. If the alanine were derived from a protein source with low d 13 C values (e.g. C3 terrestrial or freshwater fish) we would expect other amino acids to be 13 C depleted (especially Val and Phe; Fig. 6), which is not the case. Ala is a dispensable amino acid and can be directly routed from diet but is also readily synthesized from precursors originating in the first steps of glycolysis [73]. The latter are predominantly derived from dietary carbohydrates. Recently Choy et al. [73] noted that alanine values (in red blood cells and hair keratin of modern individuals) were strongly related to carbohydrate intake and not other dietary sources (i.e. meat, fish, marine mammals and corn products) whilst the other dispensable amino acid d 13 C values (Pro, Glx, Asx, Ser and Gly) were not related to carbohydrate intake. We can interpret the d 13 C alanine data from the individual in burial 7 at G-IV as an individual who had a long term diet that contained a higher amount of 13 C depleted carbohydrate, e.g. from C3 plants, compared to the others. Apart from this exception, the carbon isotope analysis of individual amino acids confirms the bulk collagen analysis and suggests that the majority of individuals consumed adequate protein to supply the nearly all the carbon in collagen. Whether alanine d 13 C values can be used as a

Organic Residue Analysis from Pottery at G-IV
Molecular and isotopic compositions of adsorbed organic residues indicate that pottery vessels were used for the processing of marine products, along with plants and other animal resources. Overall the lipid preservation was poor, and the majority of samples contained only low levels (,0.5 mg mg 21 ) of palmitic and stearic acids. Nevertheless one sample (G18E) has a lipid profile consisting of medium-and long-chain saturated (C 14 -C 24 ) and monounsaturated (C 18:1 -C 22:1 ) fatty acids, isoprenoid fatty acids (4,8,12-trimethyltridecanoic acid and phytanic acid) and longchain (C 18 -C 22 ) v-(o-alkylphenyl) fatty acids (Fig. 8, Tab. S5). Such a profile is characteristic of degraded aquatic oils, as established on contemporaneous pottery from this region [22]. A similar lipid distribution was found in G16P but with lesser preservation of v-(o-alkylphenyl) fatty acids, preventing a clear confirmation of aquatic oils. This sample and G24P contain a series of triterpenes (m/z 189, 218) revealing the presence of plant resins which were also found in Taquara/Itararé pottery assemblages and were interpreted as a waterproofing coating [23]. Finally, one sample (G22P) had low quantities of very long chain fatty acids (up to C 28:0 ), traces of long chain dicarboxylic fatty acids (C 22 -C 24 ) and isomers of the C 18 v-(o-alkylphenyl) fatty acid only. The latter compound is formed by heat alteration of polyunsaturated C 18 fatty acids, which is consistent with a plant contribution in the profile. It also contains an unusually high concentration of C 12:0 which in this context could be derived from palm kernel oil [74], [75], although the difference in the fatty acid distribution as the various biomarkers present suggest a complex mixture of lipids from different origins.
The d 13 C values (both C 16:0 and C 18:0 $-25%) of medium chain-length n-hexadecanoic (C 16:0 ) and n-octadecanoic (C 18:0 ) acids from 7 out of 11 pottery samples are within the range of marine oils reported in previous studies (Tab. S5). Lower d 13 C values, in contrast, match those observed in modern pottery vessels used to process freshwater and non-ruminant animal fats and oils [30], [31]. There was no evidence for the processing of ruminants in pottery [59] despite the presence of cervids (Mazama sp., Ozotoceros bezoarticus) in the faunal assemblages [41]. Biomarkers associated with maize processing, e.g. n-dotriacontanol, were absent in all the vessels studied although there are doubts whether these would accumulate or preserve in sufficient quantities to allow identification [76].
Charred surface residues show d 13 C and d 15 N values ranging from 222.6% to 225.8% and from +6.7% to +12.7% respectively (Tab. S6). Samples enriched in 13 C and depleted in 15 N may tentatively indicate some contribution of C4 plants [77]. However the correlation between C4 plants (e.g. maize) and d 13 C and d 15 N values of charred deposit are not straightforward [78], [79], and our d 13 C and d 15 N results are also consistent with those observed in coastal areas of Northern Europe resulting from the processing of aquatic resources [30]. In spite of the complexity underlying food crust isotopic signatures, compound specific isotopic data from the same potsherds reinforces the interpretation that marine foods principally contributed to the isotopic signal of these charred deposits.

Discussion
Marine and C3 terrestrial animals were the main sources of protein for coastal and inland sambaqui builders of S. Brazil between ,6,700 and ,1,700 cal BP. The isotopic gradient from the inland to the coast suggests the existence of confined catchment areas and/or selective targeting of specific resources [80][81][82], which is a common feature amongst sambaqui builders [11], [13], [20], [21] and other coastal populations in South America [81]. There is no isotopic evidence for the contribution of freshwater resources, which is consistent with the very low frequency of these remains in both mainland and coastal sites [11], [21]. Interestingly, some individuals at Jab-II have an unusually high intake of C3 terrestrial proteins, denoting some degree of population variability on the coast. Although the lack of significant isotopic differences between sexes and age is consistent with pervasive food sharing among these populations, the intrapopulation d 13 C and d 15 N variability at Jab-II may point to the presence of non-local individuals, as observed in other preceramic coastal populations [18], perhaps assimilated into the group through post-marital residential practices [83]; however isotopic variability may also be associated to food restrictions among members of the community [84].
The carbon isotope signature of individual amino acids indicate a minor contribution of plants to the diet of some individuals, including those that post-date the adoption of pottery on the coast. The occurrence of mortars and plant macro-remains at several, but not all, sites [1], [4], [5], [98], along with variable degree of caries, starch grains in dental calculus [5], [6] and dental wear [99] also indicate that plants made a contribution to the diets of some coastal groups at this time [1], [3], [100]. Furthermore preliminary studies have also successfully extracted phytoliths and starch grains from charred deposits of pottery from G-IV, and one individual provided amino acid (alanine) d 13 C values suggestive of a larger intake of C3 carbohydrates. However, the bulk collagen isotope analysis indicates that plants were unlikely to be major dietary staples for these coastal groups, rather the diet was protein-rich and oriented toward marine resources.
The continuity in coastal exploitation is further supported by molecular and isotopic results from organic residues preserved in pottery. These data attest to the use of pottery for the processing of animal products, including marine organisms. Hansen and Schmitz [23] achieved similar results from coeval sites in southern Brazil, revealing that pottery was commonly used for the manipulation of marine resources. Combined results therefore indicate that the adoption of pottery in coastal areas is not directly connected with the imposition of food production and did not affect the proportional contribution of marine resources to the diet of coastal populations. Small ceramic vessels had most likely a ritual and symbolic utility [16], [41]. Therefore, the molecular and isotopic analyses provide new direct evidence of the importance of marine resources in symbolic spheres. This is an interesting finding as it may have been assumed that novel or exotic cultivated plants would have had a more ritual and symbolic role in cuisine and therefore would have been more visible in the pottery contents.

Conclusion
Maritime adaptations sustained South American pre-Columbian populations since the late Pleistocene [101] and stable isotope studies reveal the crucial role of aquatic resources to several Holocene coastal groups (e.g. [86][87][88][89][90][91][92]), even during the intensification of food production (e.g. [82], [102][103][104][105][106][107][108]). However these direct lines of evidence are strongly biased towards archaeological records along the western and south-eastern coast of South America. Here we have extended the information to the subtropical Atlantic rainforest coast of Brazil. The isotope results show that it is highly unlikely that these coastal populations relied on plant carbohydrate as a major dietary source. Instead, we demonstrate the strong dependence of marine animal resources, despite the decline of monumental shell mound building and the arrival of a new subsistence strategy at ,1,500 cal BP, involving domesticated plants and pottery technology, from inland areas. Therefore our results imply that the productive maritime economy was highly resilient to social and cultural change. It remains to be assessed if ceramic producing populations on the coast were directly descended from indigenous coastal foragers or immigrants from the highlands who, having reached the coast, oriented their economy toward aquatic resources. However the resilient character of this subsistence system is further expressed by its flexibility. Rather than transforming the coastal economy, as observed in the Atlantic coasts of Europe [109], [110], the adoption of pottery was incorporated into marine focused subsistence strategies [30]. These results emphasize how the archaeological record offers a unique and exceptional opportunity to illuminate the longstanding trajectory of New World maritime adaptations, which still today play a pivotal role to coastal populations in Latin America [111].

Supporting Information
Table S1 Carbonate n 1 (C) to phosphate n 1 (P) intensity ratios, full width at half maximum (FWHM) of the phosphate n 1 (P) band and organic (C-H stretch) to phosphate n 1 (P) intensity ratios determined for each averaged spectrum as a function of the sample type. Samples are ordered from youngest to oldest with data acquired using the same Raman confocal settings across all samples. Modern lamb bone is justified as an appropriate control due to the similarities in sheep and human bone as per the RS study of Rehman et al. [52]. (DOCX)