In SeungkiKwak/Kwak_S_PhD_thesis: Sungki Kwak's PhD Thesis

set_parent("thesis.Rmd")

\chapter{Methods, Research design and analytical procedure of the organic geochemical analysis}

\section{Introduction}

In this chapter, I will discuss the methods, research design and analytical procedure of the organic geochemical analysis. I will outline a brief history of the organic geochemical analysis in the discipline of archaeology and elucidate its principles. I will also list some of the implications related to the analysis. Lastly, the details of the specific laboratory experimental process of this project will be mentioned.

\section{Concept of biomolecular archaeology and organic geochemical analysis}

Biomolecular archaeology is the study of ancient biomolecules that can provide information relating to human activities in the past [@Evershed2008b; @Stear2008: p. 24]. According to Stear [-@Stear2008], the area of biomolecular archaeological researches includes (1) the use of collagen from skeletal remains to determine the ancient dietary information [@Corr2008; @Lee2011a; @Reynard2008; @Richards2003; @Thompson2008]; (2) the analysis of DNA from archaeological materials to explore evolutionary origins and migratory patterns [@Edwards2004; @Ho2008; @Jansen2002; @Malhi2007; @Vila2001]; and (3) the study of lipid biomarkers from a range of archaeological contexts relying on the organic geochemical analysis for the reconstruction of culinary, economic and social practices throughout prehistory and history [@Berstan2004; @Bethell1994; @Buonasera2015; @Copley2001; @Copley2005a; @Craig2011; @Craig2013; @Dudd1998; @Evershed1997; -@Evershed2003; @Hansel2004; @Reber2004a; @Regert2003]. Organic geochemical analysis endeavors to determine the types of food groups that were cooked or stored within a pot by attempting to isolate and identify the specific organic compounds trapped in the fabric of its wall or adhering to its surface in residues [@Eerkens2002; -@Eerkens2005; -@Eerkens2007; @Evershed1990; @Reber2004]. Organic compounds have the advantage that they are often preserved within archaeological ceramics [@Charters1993; @Copley2005a; -@Copley2005b; @Evershed1994; @Heron1993], which is not the case in the other methods of diet reconstruction, such as examination of faunal and floral remains. In this regard, the organic geochemical analysis has become an important method of investigation which archaeologists use to better understand local diets and the function of ceramic artifacts. If we conduct it on pottery, we will be able to understand past subsistence behaviors in relation to pots even in the absence of faunal or floral remains. The direct examination of remains of organic resources in the Korean Peninsula has typically been limited to shell middens, because the high acidity of sediments does not allow long-term preservation of bone or plant remains. Therefore, organic geochemical analysis is a suitable method to investigate organic resources in non-midden sites in Korea.

\section{Organic Residues within archaeological potteries}

Among all the compound classes I have mentioned above, solvent-extractable lipids are the most frequently recovered compounds from archaeological contexts [@Evershed1993; -@Evershed2008; -@Evershed2008b]. Because of their stability against degradation and inherent hydrophobicity, they tend to persist at the original place of deposition more than other biomolecules. Due to these characteristics, lipids are nowadays the most widely studied organic compounds in the discipline of biomolecular archaeology.

Under favorable conditions, lipids are preserved at archaeological sites in association with a wide range of archaeological contexts, e. g. potteries, sediments, human and animal remains [@Evershed1993; @Evershed1999; @Mukherjee2004]. Among them, potsherds are probably the most widely distributed at archaeological sites. Due to this reason, the pottery is one of the most extensively studied material cultures for the organic geochemical analysis.

Organic residues are found in association with archaeological potteries either as (1) charred remains on the inner or outer surface of vessels, or, (2) absorbed within the fabric of their wall [@Evershed2008b; @Evershed1999]. The residues both on their surface and in their fabric can provide invaluable information regarding the use of ancient pottery vessels. However the latter case is more commonly encountered, for the fired clay acts as a 'trap' or 'net', protecting and preserving lipids during burial [@Evershed2001; @Reber2004]. Studies have shown that these compounds are relatively well insulated and preserved within that fabric over millennia [@Heron1991; @Eerkens2001; -@Eerkens2005]. The absorbed residues, unlike the visible ones, cannot be removed from a sample by washing or scraping, and remain within the ceramic matrix of the pot until extracted by solvents [@Reber2004: p. 20].

During the usage of pottery vessels in prehistoric times (e.g. during culinary practices), fats, oils and waxes originated from animals, insects or plant products become entrapped within the vessel wall. The fats and waxes are protected from microbial and chemical degradations as well as groundwater leaching by the ceramic matrix. These organic residues can be extracted from potsherds and analyzed hundreds or even thousands of years after the pottery was discarded by ancient people. For example, in case of Great Britain, absorbed residues are typically detected in 50 to 60 % of all the vessels studied [@Mukherjee2004]; however, the actual proportion is dependent on many factors including burial conditions and age [@Evershed2008a]. Though Fats and waxes can also be preserved in the form of charred or dried deposits adhering to the vessel wall, this class of residue is much less commonly observed.

The preservation of organic compounds in the porous wall of the pottery was first recognized over 30 years ago, when the lipids extracted from archaeological potteries were analyzed by the gas chromatography (hereafter GC) [@Condamin1976; @Stear2008: p. 25]. This approach uses the ratio between the amounts of common fatty acids to determine particular classes of food [cf. @Eerkens2005; -@Eerkens2007; @Patrick1985]. But it has a problem, for different kinds of fatty acids decompose at different rates over time due to oxidation and hydrolysis. Since such ratios are not stable over time, researchers have to rely on those of the fatty acids that decompose at similar rates. For example, Eerkens [-@Eerkens2001; -@Eerkens2005; -@Eerkens2007] set up the criteria for distinguishing different food classes, based on four useful ratios involving eight fatty acids which are relatively common in archaeological residues (C12:0/C14:0, C16:0/C18:0, C16:1/C18:1 and (C15:0 + C17:0)/C18:0). Upon these criteria, he was able to distinguish five different food classes which are: meat of terrestrial mammals, fish, seeds/nuts and berries, roots, and greens (Table \ref{tab:fatty_acid_ratio}). After these studies that attempted to determine the origins of the organic residues based on the proportions between individual compounds, more sophisticated mass-spectrometric instruments were employed and made it possible to identify a wide range of organic commodities within archaeological vessels.

The identification and characterization of lipid residues rely upon the comparison of chemical properties of lipid compounds derived from organisms. Those compounds are presented in both the archaeological ceramics and contemporary plants and animals. Such "biomarkers" can help scientists to reconstruct the dietary life of prehistoric peoples [@Evershed2008; @Evershed2008a; @Heron1993: pp. 267-270]. This is achieved by the high temperature gas chromatography (hereafter HTGC) and gas chromatography - mass spectrometry (hereafter GC-MS) techniques that can acquire detailed molecular compositional information from the extracts. That information can subsequently be compared to that of modern reference materials. Through this method, scholars have identified terrestrial and marine animal fats, plant leaf waxes (e.g. cabbage and leek), beeswax, birch bark tar, and palm fruit (Table \ref{tab:biomarkers}). But the biomarkers only occur in case of good preservation of the organic residues; more often we only have the degraded products. More recently, the use of soft ionization techniques in MS, such as electrospray ionization (ESI), has proven particularly useful in the structural characterization of high molecular weight compounds preserved within the archaeological pottery like triacylglycerols (hereafter TAGs). They are more difficult to examine with the GC-MS technique [@Stear2008: p. 26; @Mirabaud2007].

\begin{table}[ht] \centering \begin{tabular}{p{2cm}p{2cm}p{2cm}p{2cm}p{2cm}p{2cm}p{2cm}} \hline ratio & State & terrestrial mammals & fish & Roots & greens & seeds/nuts and berries \ \hline C16:0/C18:0 & Fresh & \textless 3.5 & 4-6 & 3-12 & 5-12 & 0-9 \ & degraded & \textless 7 & 8-12 & 6-24 & 10-24 & 0-18 \ C12:0/C14:0 & Fresh & \textless 0.15 & \textless 0.15 & \textgreater 0.15 & \textgreater 0.05 & \textgreater 0.15 \ & degraded & \textless 0.15 & \textless 0.15 & \textgreater 0.15 & \textgreater 0.05 & \textgreater 0.15 \ \hline \end{tabular} \caption{Criteria used to distinguish food types, based on fatty acid ratios (Eerkens 2005)} \label{tab:fatty_acid_ratio} \end{table}

Most recently, the application of the compound-specific stable carbon isotope analysis (hereafter CSIA) by the gas chromatography-combustion-isotope ratio mass spectrometry (hereafter GC-C-IRMS) enabled a more specific characterization of the organic compounds within the archaeological pottery. The stable carbon isotope analysis has become a powerful method for tracing diet patterns of animals, for the isotopic composition of animals depends upon the food they eat [@Malainey2010]. In archaeological settings, the method has been widely used on human remains for understanding human subsistence patterns by distinguishing C3 diets (e.g. rice) from C4 diets (e.g. millet) [@Barton2009; @Bentley2007]. In the field of ceramic studies, Hastorf and DeNiro [-@Hastorf1985] conducted the bulk carbon isotope analysis for charred organic residues on the surface of potsherds to understand human diets. With the introduction of GC-C-IRMS, the stable carbon isotope value of individual compounds in a mixture can now be measured with high precision, providing a unique opportunity to conduct the carbon isotopic analysis on the fatty acids that are insulated within the fabric of archaeological ceramics [@Mottram1999]. Scholars have been successfully tracing the presence of C3, C4 plants, animal fats, and aquatic resources (e.g. fish and mammals) on prehistoric potsherds through CSIA [@Craig2011; @Craig2013; @Cramp2011; @Evershed1994; @Evershed1997; @Mottram1999; @Reber2004; @Salque2013].

\begin{table}[h] \begin{tabular}{p{3cm}p{8cm}p{4cm}} \hline Commodities & Lipid biomarkers & References \ \hline Terrestrial animal fats & Characteristic distribution of TAGs, diacylglycerols (hereafter DAGs), monoacylglycerols (hereafter MAGs) and free fatty acids. Particularly high abundance of C16:0 and C18:0 fatty acids. & Evershed et al. 2001 \ Marine animal fats & Isoprenoid fatty acids (4, 8, 12-trimethyltridecanoic acid and phytanic acid). Thermally produced $\omega$-(o-alkylphenyl)alkanoic acids & Hansel et al. 2004, Copley et al. 2004, Craig et al. 2011 \ Plant waxes (e.g. brassica wax) & Long chain alcohols, ketones, n-alkanes, aldehydes and wax esters. Specific biomarkers of brassica wax (cabbage) nonacosane, nonacosan-15-o1, nonacosan-15-one. & Evershed et al. 1991 \ Beeswax & Characteristic distribution of odd numbered n-alkanes (C23-C33), even numbered free fatty acids (C22-C30), and long chain palmitic wax esters (C40-C52) & Evershed et al. 1997, Regert et al. 2003 \ Birch bark tar & Triterpenoids from lupane family, namely betulin, lupeol and lupenone & Charters et al. 1993 \ Palm fruit & High abundance of C12:0 and C14:0 saturated fatty acid & Copley et al. 2001 \ \hline \end{tabular} \caption{Identification of fatty acids by using GC-MS (Stear 2008: p. 26)} \label{tab:biomarkers} \end{table}

\section{Identification of lipids}

Different criteria can be used for the identification of lipid residues. For example, the presence of fatty acids can indicate a plant or animal origin through their relative abundance, while the TAG distribution and structure are also potentially useful indicators [@Mukherjee2004]. However, caution must be exercised when using these criteria, for ratios between fatty acids may change over time and TAGs are often only present in very low abundance or completely absent [@Mukherjee2004: p. 14]. In addition, because of the differential degradation and variable extraction rate of organic compounds, it is hard to tell exactly what types of food were processed in the pot only with the GC-MS analysis [cf. @Reber2004a]. A more reliable method for the elucidation of the lipid origin is to determine the stable carbon isotope (hereafter $\delta$^13^C) value of individual C16:0 and C18:0 fatty acids.

In this thesis, I have conducted the organic geochemical analysis on the absorbed lipids extracted from the potsherds. The analysis involves two different analytic methods: GC-MS and CSIA based on GC-C-IRMS. The former is used for separation and identification of organic compounds within a potsherd, and the latter can be employed for the further isotopic analysis of specific compounds. If fatty acids such as C16:0 and C18:0 are found in a range of different food products, the isotopic analysis can further distinguish between their origins. Most of the recent organic geochemical studies on potsherds successfully detected the presence of different food groups including animal fat, ruminant milk, marine resources (e.g. fish and mammals), fresh water resources, C3, and C4 plants with those two methods combined [@Craig2011; @Cramp2011; @Reber2004].

\newpage

\subsection{GC-MS analysis}

GC-MS enables the identification of even highly degraded commodities. A reliable classification of commodities processed in the archaeological pottery can be made by comparing the chemical structure of individual compounds with that of modern and archaeological references [@Mukherjee2004]. A knowledge of the degradative process occurring during vessel use and burial is essential in order to identify the lipid residues preserved within vessels. These analyses are enhanced by analyzing the results of laboratory and field experiments simulating use and degradation [cf. @Dudd1998; @Dudd1998a; @Evershed2008].

Figure \ref{Partial_HTGC_profile} shows an example of degraded animal fat obtained by HTGC analysis of a Romano-British sherd from Stanwick, Northamptonshire [@Mukherjee2004: p. 14]. A low abundance of intact TAGs are observed at retention times above 30 min; however, the majority of the lipid was hydrolyzed during vessel use or burial, resulting in the formation of DAGs, MAGs, and free fatty acids. The fatty acids present, eluted between 10 and 20 min, comprise mainly C16:0 and C18:0 components. A high abundance of C18:0 is indicative of animal fat.

Distributions of TAGs in ancient fats from pots can provide a reasonable evidence for the presence of animal fats and dairy products [@Mukherjee2004: p. 19]. For the detection of TAG 'biomarkers', GC-MS is used, which can help to make distinction between different kinds of animal fats [@Dudd1998a]. For example, bovine adipose fats possess saturated TAGs of every carbon number between C44 and C54 and pig fats contain a narrow distribution of them (e.g. TAGs range from C46 to C54) [cf. @Mukherjee2004]. On the other hand, milk fats are quite distinctive because of their relatively wide TAG distribution ranging from C40 to C54 [@Dudd1998; @Evershed2003]. Figure \ref{TAGdistribution} shows TAG distributions of both fresh/degraded lipid residues gathered from the modern reference fats. Most importantly, however, it should be addressed that distributions of TAGs alone are not sufficient enough for the proper identification of lipid origin [@Mukherjee2004]. Moreover, TAGs frequently do not survive in archaeological residues. Due to this vulnerable characteristic, sometimes TAGs may be misinterpreted. Figure \ref{TAGdistribution}d and e indicate fresh and degraded ruminant milk fat. Since the degradation process during vessel use or burial makes the ruminant milk TAG distribution (\ref{TAGdistribution}e) similar to those of adipose fats (\ref{TAGdistribution}a; b), a more prudent decision has to be made based on a more robust stable isotopic criterion [@Copley2003; @Dudd1998; @Mukherjee2004: p. 20].

In this study, GC-MS was applied to identify the compounds which are only found in certain food groups (cf. Table 4.2). The biomarkers which these compounds constitute are present in different types of fats; for example, short chain fatty acids in dairy fat, unsaturated fatty acids in plant oil, cholesterol in animal fats and plant sterols (e.g. b-sitosterol) in plant oil. Especially, Phytanic acid (3,7,11,15-tetramethylhexadecanoic acid) and 4,8,12-TMTD (4,8,12-trimethyltridecanoic acid) are isoprenoid compounds which are mostly found in particularly high concentrations in marine animals [@Evershed2008b]. Along with thermally produced long-chain $\omega$-(o-alkylphenyl)alkanoic acids, these compounds are indicators of aquatic/marine resources [@Craig2011; @Evershed2008b]. But, as I already indicated, they only occur in case of good preservation of food residues. One way to deal with this preservation issue is to use GC-MS in the selection monitoring (SIM) mode, where the analysis focuses on specific biomarkers, in order to try to get a better signal from the compounds which may be present in very low quantities, or which may be masked by more abundant compounds such as C16:0 and C18:0 fatty acids.

\subsection{Compound specific isotope analysis}

In most cases a pot is reused over time, and may be used to cook different kinds of food from one cooking episode to another. Researches with amino acids show that the first use of a pot essentially saturates it with them, and seals it off further amino acid contributions, that is, the amino acid residues trapped within a pot record only its first use [@Fankhauser1997]. On the other hand, fatty acids and other compounds tend to accumulate in the fabric of the pot wall. Therefore, the result of the analysis is, in this case, more likely to reflect the entire usages of the pot. Generally, the result is assumed to represent the type of food group that was most frequently processed in it. However, this does not mean we can just disregard the complication caused by its multiple usages. Besides, due to the differential degradation and variable extraction rate of the organic compounds, it is not easy to tell exactly what types of food were processed in the pot only with GC-MS analysis [cf. @Reber2004a]. On top of that, animal fats and plant oils offer a great challenges, because their major components, unsaturated fatty acids in particular, rarely if ever, survive, leaving mainly rather undiagnostic n-alkanoic acids such as C16:0 and C18:0 fatty acids [derived mainly through the hydrolysis of triacylglycerols, Figure \ref{degradation}, @Evershed2008a].

Luckily, we do have the last approach that can help us further clarify the origin of the organic compounds in a pot: compound specific stable carbon isotope analysis. Early works of stable isotope study in the archaeological field involved the bulk isotopic analysis [@Hastorf1985; @Morton1988]. However, the application of CSIA via GC-C-IRMS allows us to achieve a greater specificity, for the structure of diagnostic compounds in complex mixtures can be directly linked to their stable isotope value [@Evershed1994]. Thus, the compound specific stable isotope analysis avoids ambiguities arising from contamination by, e.g. plasticizers originating from plastic bags in which sherds are often stored. These ambiguities cannot be resolved in the bulk isotope analysis [@Mukherjee2004]. Most importantly, you do not need to have solid materials (e.g. bone) for the analysis.

Generally, different food groups tend to have different major fatty acids having different ranges of $\delta$^13^C values (e.g. C16:0 and C18:0). For example, $\delta$^13^C values of ruminant (goat, sheep and cow/buffalo), chicken, equine, pig fat, ruminant milk, C3 plant, C4 plant, and aquatic resources (e.g. fish and mammals), have each their own range. Therefore, $\delta$^13^C values of fatty acids provide the basis for distinguishing those food classes. Though these values were obtained from the modern fauna and flora, they have been employed as references for many archaeological studies [@Craig2011; @Cramp2011; @Fraser2012; @Reber2004; -@Reber2004a]. In proceeding in this fashion, these studies assume that the $\delta$^13^C values of modern samples are comparable to those of ancient members of the same species. Scholars were able to detect the presence of the above classes of food by measuring $\delta$^13^C values of the two most common fatty acids in archaeological pots: palmitic acid (C16:0) and stearic acid (C18:0), with GC-C-IRMS, which provides a means to address some key questions concerning human subsistence in prehistory [@Craig2013; @Evershed1994; @Evershed1997; @Mottram1999; @Salque2013].

In nature, carbon exists as three isotopes: ^12^C and ^13^C, which are both stable, and ^14^C, which is radioactive. Occurring as CO~2~ (carbon dioxide), they are organizing respectively 98.89 %, 1.11 %, and 1 x 10^-10^ % of the global carbon pool [@Stear2008]. Being inorganic, carbon dioxide is incorporated into living organisms through the process of photosynthesis. Green plants transform carbon dioxide and water into oxygen and organic sugars. When incorporated into the plant tissue through photosynthesis, the isotopic fractionation occurs and the ratio between ^13^C to ^12^C changes significantly, because plants use the carbon dioxide containing the lighter isotope, ^12^CO~2~, more readily than that of the heavier isotope, ^13^CO~2~. Plants are consumed by herbivores, and herbivores are consumed by carnivores. If one can measure the ratio between ^13^C to ^12^C in the remains of those organisms and compare it with known reference isotope ratios, then it will be possible to trace their diet.

The stable carbon isotope ratio is measured by comparing the relative differences of ^13^C to ^12^C between the sample and the international standard, Pee Dee belemnite (PDB), a limestone from South Carolina [@Malainey2010]:

It is expressed using the delta ($\delta$) notation:

$$\delta^{13}C=\left(\frac{\text{R sample - R standard}}{\text{R standard}}\right) \times 1000$$

Where:
R sample = molar ^13^C/^12^C ratio of the sample,
R standard = molar ^13^C/^12^C ratio of the standard

The $\delta$^13^C value is the difference between the ^13^C content of the sample and that of the standard, and is expressed relatively to the international standard. Differences between samples are very small, so values are counted per mil ($\permil$), rather than percent (%). The standard contains less ^12^C and more ^13^C than most natural materials, so $\delta$^13^C values of the samples are usually negative, ranging between -37 and -8$\permil$. The error range for compound specific $\delta$^13^C values of fatty acids is $\pm$ 0.3$\permil$.

\subsubsection{Modern reference animal fats and plant oils}

Naturally, plants and animals of today cannot be directly compared to those of prehistoric times, due to the various environmental changes that have occurred over the last few hundred years [@Mukherjee2004]. There are several factors of these changes including: (1) consuming fossil fuel since the industrial revolution which has caused changes in the isotopic composition of CO~2~ in the air [@Friedli1986]; (2) commercial farming due to which animals have been fed with supplements to enhance their diets and to improve the nutritional quality of their meat and milk [cf. @Chilliard2001; @Lowe2002]; and (3) selective breeding that has introduced changes in the composition of the fat and milk of domestic animals [@Mukherjee2004: p. 17]. There are also regional level factors. For example in Great Britain, since C4 plants (e.g. millet) have been introduced and incorporated into animals' diet not long ago, it is hard to directly compare $\delta$^13^C values of modern and prehistoric animals [@Stear2008].

The identification of plant oils through the isotope analysis is possible, for the range of $\delta$^13^C values is different in each group of plants that share the photosynthetic pathway. Terrestrial plants use three different photosynthetic pathways, namely C3, C4 and CAM. The C3 plants (e.g. wheat, rye, barley, legumes) are the most abundant, and are found mainly in moderate areas. They fix the atmospheric CO~2~ using the Calvin and Benson cycle [@Calvin1948]. ^13^CO~2~ is discriminated by Ribulose-1,5-bisphosphate carboxylase/oxygenase (hereafter RuBisCO), resulting in relatively low $\delta$^13^C values ranging from -32 to -20 $\permil$ [@Boutton1991]. C4 plants (e.g. millet, maize, sugarcane, sorghum) fix CO~2~ through the Hatch-Slack pathway [@Hatch1966], and the carbon fixation occurs near the surface of the leaf in mesophyll cells with phosphoenolpyruvate (hereafter PEP). The latter pathway gives relatively high $\delta$^13^C values in the range of -17 to -12.5 $\permil$ [@Malainey2010]. Crassulacean acid metabolism (hereafter CAM) plants (e.g. pineapple, aloe vera, jade plant) can either assimilate CO~2~ at night only or night and day. The carbon fixation occurs at night through PEP carboxylase as in C4 plants. On the other hand, during the day time, CAM plants can switch their photosynthetic pathway and use RuBisCO to fix CO~2~. As a result, the range of ^13^C values for some CAM plants is quite broad [cf. @Malainey2010].

For the identification of animal fats originated from the archaeological pottery, they were compared with the carefully assembled data of modern fats [@Craig2013; @Copley2003; @Dudd1998; @Evershed2003]. The treatment of modern fats to create the reference database is slightly different from case to case. In Britain, only the animals that are being reared on known diets were sampled in order to form the database (e.g. C3 plant diet in order to mimic the prehistoric condition, absence of C4 plant), which includes adipose fats from cattle, sheep and pigs, and milk fat from cattle and sheep [@Copley2003; @Dudd1998; @Evershed2003]. The $\delta$^13^C values from these animals reflect their different diets and variations in their metabolism as well as physiology [@Stear2008; @Evershed1999]. The ellipses shown in Figure \ref{reference_example}a indicate the $\delta$^13^C values obtained from the C16:0 and C18:0 fatty acids from each of the reference animal fats; sheep and cattle data are grouped together as ruminant fats. Dairy and adipose fats from ruminant animals can be distinguished, for the C18:0 fatty acid in dairy fat is significantly more depleted in $\delta$^13^C value [average 2.1 $\permil$, @Copley2003]. In Japan, to avoid the effects of commercial farming and selective breeding, modern reference samples were collected from authentic wild animals (Figure \ref{reference_example}b). To facilitate comparison with archaeological data, the $\delta$^13^C values obtained from all modern reference animals were adjusted by the addition of 1.2% considering post-Industrial Revolution effects of fossil fuel burning [@Friedli1986].

![Reference database created based on modern fats for CSIA. (a): Only the animals having been reared on known diets were sampled (e.g.C3 plant diet in order to mimic the prehistoric condition, absence of C4 plants) [@Copley2003; @Dudd1998]. (b): The modern reference samples were collected from authentic wild animals to avoid the effects of commercial farming and selective breeding [@Craig2013]. The d^13^C values obtained from all modern reference animals were adjusted by the addition of 1.2 permil, considering post-Industrial Revolution effects of fossil fuel burning [@Friedli1986]. \label{reference_example}](figures/reference_example.jpg)

\newpage

\subsubsection{Interpretation of CSIA}

For the interpretation of CSIA, the $\delta$^13^C values acquired from the C16:0 and C18:0 fatty acids in archaeological potsherds are plotted in the figure of the reference animal fat ellipses (Figure \ref{mixingcurve}a). When the $\delta$^13^C values of fatty acids plotted within an ellipse, like the case of the pork (porcine) fat in Figure \ref{mixingcurve}a, then the fat in question can be identified as pork fat. When the ^13^C values are plotted just outside the ellipse, then the fat can be identified 'predominantly' as pork fat. However, in most cases the $\delta$^13^C values are located between the ellipses of the reference fats, which indicates the mixing of different classes of food stuffs within the vessel either at a moment or during all the time of its use [@Mukherjee2004].

To account for the mixing of different animal fats in varying proportions within a single vessel, a theoretical mixing model is used to calculate theoretical $\delta$^13^C values [@Bull1999; @Mukherjee2004: p. 22]:

$$\delta^{13}C_{mix}=\delta^{13}C_{(A)}\left ( \frac{(X\times A)}{\left (X\times A \right )+\left (Y\times B \right )} \right )+\delta^{13}C_{(B)}\left ( \frac{(Y\times B)}{\left (X\times A \right )+\left (Y\times B \right )} \right )$$

Where:
$\delta$^13^C~mix~ = predicted $\delta$^13^C value of the fatty acid with contributions from fats A and B
$\delta$^13^C\textless sub>(A)\textless /sub> = $\delta$^13^C value of the individual fatty acid in fat A
$\delta$^13^C\textless sub>(B)\textless /sub> = $\delta$^13^C value of the individual fatty acid in fat B
X = percentage of fat A present (%)
Y = percentage of fat B present (%)
A = percentage of the individual fatty acid in fat A (%)
B = percentage of the individual fatty acid in fat B (%)

Theoretical mixing curves between the porcine adipose fat, ruminant adipose fat and ruminant dairy fat are shown in Figure \ref{mixingcurve}b. The ellipses which represent different food classes (ruminant adipose fat, ruminant dairy fat and porcine adipose fat) are connected by a theoretical mixing curve (Figure \ref{mixingcurve}b).

When utilizing this theoretical mixing model for the interpretation of the contributions of different foodstuffs within a mixture, we need to consider several important points. First of all, it is nearly impossible to quantify exactly how much mixing was occurred during each vessel use, and how often each vessel was subsequently re-used [@Mukherjee2004]. It is also difficult to estimate the exact relative amount of different food classes cooked in a vessel over its lifetime usage, for the concentration of the fatty acids from different food classes varies significantly [@Enser1991].

![Interpretation of the results of CSIA (a): $\delta$^13^C values acquired from the C16:0 and C18:0 fatty acids in archaeological potsherds are plotted along with the reference animal fat ellipses [@Evershed2007] (b): The theoretical mixing curves between the porcine adipose fat, ruminant adipose fat and ruminant dairy fat are shown [@Evershed2008a] \label{mixingcurve}](figures/mixingcurve.jpg)

\subsubsection{Possibilities of variation in $\texorpdfstring{\delta^13^C}{TEXT}$ values of the fatty acids from the archaeological lipid}

According to the Mukherjee [-@Mukherjee2004: pp. 23-30], there are several possible sources that can affect $\delta$^13^C values of fatty acids from archaeological lipid.

C3, C4 and marine plant contributions

Plants are consumed by herbivores and herbivores are consumed by carnivores. Since $\delta$^13^C values in living organisms are influenced by their food, a careful consideration is demanded, when the researcher tries to trace their identity based on $\delta$^13^C values. For example, discriminating the contribution of C4 plants to the diet of animals is not an easy task. This task might not be a problem in the areas where there are no native C4 plants (e.g. the northern part of Europe or Britain), but must be taken into account when analyzing the lipids from more arid regions, where C4 plants are quite ubiquitous. Table \ref{tab:bulk_isotope_reference} shows the ranges of bulk $\delta$^13^C values of the major ecosystem; it provides a guideline to the trends that might be observed in the archaeological lipids [@Mukherjee2004].

\begin{table}[h] \centering \begin{tabular}{@{}ll@{}} \toprule Material & Bulk $\delta$ \textsuperscript{13}C value (\permil) \ \midrule C3 plant & -32 to -20 \ C4 plant & -17 to -9 \ CAM plant & -20 to -10 \ Groundwater & -25 to -10 \ Atmospheric CO$_{2}$ & -8 \ Sea grasses & -15 to -3 \ Marine vertebrates & -17 \ Marine carbonates & 0 \ \bottomrule \end{tabular} \caption{The ranges of bulk $\delta$ \textsuperscript{13} C values of natural materials (modified from Mukherjee 2004)} \label{tab:bulk_isotope_reference} \end{table}

In case of the archaeological lipids present in potsherds, the contribution of C4 plants to an animal's diet would have cause more enriched $\delta$^13^C values of the C16:0 and C18:0 fatty acids. So, if the reference animals used to compile the database are reared on C3 diets, the $\delta$^13^C values from the archaeological lipids will show a deviation from those given by the database, when identifying animal fats with a possible C4 diet contribution. This means, for example, it is quite possible that the pure ruminant adipose fat can be misinterpreted as a mixture of ruminant and porcine adipose fats, or even as pure pig fat (cf. Figure \ref{mixingcurve}).

This can be overcome by comparing the difference between the $\delta$^13^C values of the C16:0 and C18:0 fatty acids in the reference fats and that in the archaeological fats ($\Delta$^13^C). The comparison will be expressed by following formula:

$\Delta$^13^C = $\delta$^13^C18:0 - $\delta$^13^C16:0

This can separate fats based on physiological differences between the animals (e.g. ruminant adipose, ruminant dairy, and non-ruminant adipose) regardless of differences of their diets or surrounding ecosystem [Figure \ref{Bigdelta}, @Copley2003; @Evershed2008b].

At coastal sites such as shell middens, the contribution of marine plants needs to be considered. For example, the diet of sheep in North Ronaldsay, Great Britain is dominated by seaweed, only a small quantity of terrestrial grass being grazed by them seasonally. As a result, the bulk $\delta$^13^C values acquired from their bone collagen measured around -13$\permil$, a range of $\delta$^13^C values which overlaps that of pure marine consumers [@Mukherjee2004].

Most marine plants cannot absorb carbon dioxide directly from the atmosphere, but from dissolved gasses in the surrounding water. Though the $\delta$^13^C value of marine CO~2~ is variable, and mainly depends on depth with other localized factors, it is usually in the region of 0 $\permil$. Despite the difference in photosynthetic mechanism between marine and terrestrial plants, marine plants fractionate carbon approximately to the same extent as terrestrial C3 plants; and they have $\delta$^13^C values in the range of -11 to -19 $\permil$ [@Chisholm1982]. Foreshore plants that are not permanently submerged underwater may be more complicated, but still show a marine signature. Their $\delta$^13^C values are distinguishable from - 25 $\permil$ of C3 plants [@Mukherjee2004]. Therefore, animals eating a large amount of marine and foreshore plants (e.g. seaweed), should be distinguished from those which eat predominantly terrestrial diets. However, if archaeological samples were collected from where both C4 plants and marine/foreshore ones are present, researchers need to carefully consider whether relatively enriched $\delta$^13^C values in animal fats are from C4 plants or marine/foreshore ones.

Forest density and depletion of ^13^C

In the areas covered with dense forest we see a significant deviation of ^13^C distribution from the global average causing plants to be depleted of ^13^C. In these regions, a positive correlation between the forest density and the degree of depletion of ^13^C is observed. In addition, there is a gradual variation of $\delta$^13^C values of tree leaves from the ground to the top of the tree; and it indicates that the most negative values occur near the ground [@Medina1980; @Vogel1978]. This is what we call the 'canopy effect' [@Medina1980]. The average bulk $\delta$^13^C value of C3 plants in open air areas is about -26 $\permil$. However, for the leaves in a subtropical monsoon forest, a $\delta$^13^C value of -35 $\permil$ was recorded, and a value as low as -37 $\permil$ was observed in the Amazon forest [@Ehleringer1987; @Medina1980a].

This phenomenon in dense forest areas will influence the $\delta$^13^C values of fatty acids extracted from the local ruminant animals and pigs dwelling in forest [@VanDerMerwe1989]. Therefore, if the reference animals used for the study were not raised within the forest environment, they may have more enriched $\delta$^13^C values of fatty acid, compared with their ancient counterparts which dwelled in forest. That is, fatty acids from archaeological fats might indicate more negative $\delta$^13^C values than those of their modern counterparts; and this must be carefully considered.

![$\delta$^13^C values of the cellulose from the oak tree-ring sequence (a): 11-year running mean from ancient Irish oaks [data obtained from @McCormac1994; @Mukherjee2004] (b): Yearly measurement from 1970 to 1995 of modern oaks in east England [@Robertson1997] \label{tree-ring}](figures/tree-ring.jpg)

Variations in $\delta$^13^C values of CO~2~

Things change over time. Any variation in the atmospheric CO~2~ which occurred over time as a result of a climate change or environmental fluctuation, may have caused a deviation of $\delta$^13^C values of archaeological animal fats from the reference values. The variation in $\delta$^13^C value of the atmospheric CO~2~ from the multiplied tree-ring record obtained from oaks suggests it can vary up to 1.5 $\permil$ [Figure \ref{tree-ring}a), @McCormac1994]. Even within a relatively short term, the $\delta$^13^C value of the atmospheric CO~2~ can vary quite dynamically [Figure \ref{tree-ring}b, @Robertson1997]. It is likely that other terrestrial plants will also show variations in a similar way, but their scale might differ between the species [@Mukherjee2004]. The differences in $\delta$^13^C values between modern and ancient fats resulting from such a temporal variation of the atmospheric CO~2~ can be overcome by comparing $\delta$^13^C values of modern reference and archaeological fats.

Variation related to human activities

As mentioned above, the theoretical mixing curve was calculated to consider mixing of different food products within a single vessel during its lifetime usage. However, in some cases, mixture with other uneatable natural products is often observed. For example, the beeswax contained in lipid extracts from potsherds appears often as a mixture with degraded fats from foodstuffs [@Mukherjee2004]. Beeswax is characterized by a distribution of linear hydrocarbons of odd-numbered carbon (C21 - C23), free fatty acids of even-numbered carbon (C22 - C30), and/or long-chain wax esters with the carbon number range from C40 to C52 [@Kolattukudy1976; cf. @Mukherjee2004]. Though the exact reasons for the presence of beeswax in archaeological pottery vessels are yet unknown, it may have been used as a 'slip' due to its hydrophobic characteristic, or it might be a byproduct of the use of honey in cooking/flavoring. The abundant C16:0 fatty acid present in modern beeswax exhibits a $\delta$^13^C value of around -26.4 $\permil$, while the C18:0 fatty acid is present in low abundance [@Mukherjee2004]. In this situation it is important to assess whether there is a significant isotopic contribution from natural products like beeswax and how it may influence our interpretation of isotopic analyses.

When conducting CSIA, the best way to establish the reference database is to collect modern samples of fauna and flora from the same region where archaeological materials were collected. However, as I mentioned above, the modern day's commercial farming with supplements makes it impossible for us to directly compare the $\delta$^13^C values from archaeological materials with those from modern samples. To overcome this issue, scholars have been collecting samples from wild fauna and flora for creating the reference database. Unfortunately, in case of Korea, since wild terrestrial mammals are extremely rare, it is beyond the scope of this study.

In this thesis, as for the CSIA, the archaeological samples from the central part of the Korean Peninsula were sent to the Stable Isotope facility at the University of California-Davis, and analyzed by Varian CP3800 GC coupled onto a Saturn 2200 ion trap MS/MS. Based on the results, the stable carbon isotope values of C16:0 and C18:0 fatty acids from the archaeological samples will be compared with the available modern references that were obtained from the modern fauna and flora that exist in either Japan, Northern Europe or North America [@Copley2003; @Craig2011; @Craig2013; @Cramp2011; @Dudd1998; @Dudd1999; @Evershed1994; -@Evershed1997; @Mottram1999; @Reber2004; @Salque2013; @Steele2010] to detect the presence of the potentially cooked resources in the prehistoric Korean Peninsula. Since the overall ecosystem of Japan, Northern Europe, and North America is similar to that of Korea and almost all the fauna and flora having produced the data for reference exist also in the Korean Peninsula, this approach assumes that the $\delta$^13^C values of available modern samples are comparable to archaeological ones from the Korean Peninsula.

\section{Analytical procedures}

Lipids are medium-sized molecules which possess predominantly linear, branched or cyclic hydrocarbon skeletons making them soluble in organic solvents (Correa-Ascencio and Evershed 2014). For this reason, the most well-known way of the extraction of organic compounds is using a solvent mixture (e.g. chloroform-methanol 2 : 1 v/v) and the ultra-sonication of powdered potsherds. The main purpose of this approach is to extract free fatty acids and other organic compounds which are absorbed and trapped in the voids of clay matrixes. This way of extraction of lipids from archaeological ceramics by a solvent mixture has proven its effectiveness in different parts of the world. However, Craig and his colleagues [-@Craig2004] showed that the lipid recovery can be incomplete when extracting with a solvent mixture, and some portions of residues do remain non-extractable without the use of a stronger extractant (e.g. methanolic sodium hydroxide). As a response to that, Correa-Ascencio and Evershed [-@Correa-Ascencio2014] recently developed a new extraction protocol which uses acidified methanol (2% sulfuric acid-methanol v/v). According to Correa-Ascencio and Evershed, this new "methanolic acid extraction" has several advantages over the method of conventional solvent extraction:

(1) The new method can recover both free and bound lipids from the ceramic matrix and therefore, is especially effective in increasing the recovery rate of lipid residues from the archaeological pottery containing those of low concentration (Figure \ref{SEvsAE}). In this regard, the application of this new method has the potential to expand the limits of the analysis of archaeological lipid residues when lipid preservation is limited.

(2) The simultaneous extraction and derivatization of lipid residues, for the further isotopic analyses, shorten significantly the examination time to one day of overall laboratory time instead of four to five days required when the chloroform : methanol extraction method is applied; and they also shorten the require of materials.

(3) The major disadvantage of the new method is the compositional information loss due to the hydrolysis of complex lipids (e.g. acylglycerols and wax esters) during the extraction process. However, the loss of these lipids is not problematic, as they are the components that occur rarely, or in very low abundance, in most archaeological assemblages.

In this thesis, both methods were employed to test their suitability for the Korean Peninsula. Figure \ref{protocol} shows the differences between the solvent and acid extractions.

\subsection{Glassware, solvents and reagents}

All the solvents used for this research were HPLC (High-performance liquid chromatography) grade. The reusable glassware were washed with Decon 90 (Decon laboratories), rinsed with acetone, dried in the oven at first and heated in the furnace (450 $^{\circ}$C; 24 hours). In order to prevent contamination, combusted foil and tweezers were used to manipulate the samples. Analytical blanks were prepared with each batch of samples during each procedure of lipid extraction and derivatization to monitor any possible source of contamination. Analytical grade reagents (typically $\geq$ 98% purity) were used throughout.

\subsection{Solvent extraction of lipids}

The lipids were extracted following an established protocol outlined in Figure \ref{protocol}a. Approximately 5-10 g of each potsherd was sampled and its surface was cleaned using a drill (Dremel 3000) to remove any external contaminants, such as those originating from soil or fingers due to handling during the excavation/curation process. The cleaned sample was ground to fine powder in a glass mortar & pestle and accurately weighed to be put in a glass vial. The lipids were extracted using chloroform : methanol (2:1; 10 mL) and sonicated (20 min. $\times$ 2). The extract was then centrifuged (2500 rpm; 10 minutes.) and only the liquid portion containing the Total Lipid Extraction (hereafter TLE) was removed and transferred to a glass vial. The TLE was filtered through a silica column (1 g) to remove any particulate matter and accidental inclusions of solid materials. About a half portion of the TLE was derivatized to form Trimethylsilyl (hereafter TMS) ethers prior to analysis by GC-MS. The other half was derivatized to fatty acid methyl esters (hereafter FAMEs) and analyzed by GC and GC-C-IRMS.

\subsubsection{Preparation of TMS derivatives}

One half of the TLE was treated with N,o-bis(trimethylsilyl)trifluoroacetamide (hereafter BSTFA) containing 1 % trimethylchlorosilane (40 uL; 70 $^{\circ}$C; 1 hour). Then, BSTFA was removed under gentle nitrogen gas and the derivatized TLE was dissolved in toluene (50 uL) prior to GC-MS.

\subsubsection{Preparation of FAMEs}

The FAME derivatives of the free fatty acids were prepared by heating them with BF3-methanol (14 % w/v; 100 uL; 70 $^{\circ}$C; 1 hour). Nano-purified water was added (1 mL) and the FAME derivatives were extracted with chloroform (3 x 2 mL) and the solvent was removed under nitrogen. The FAMEs were redissolved in hexane prior to the analysis by GC-MS and GC-C-IRMS.

\subsection{Methanolic acid extraction of lipids}

The lipids were extracted following an established protocol outlined in Figure \ref{protocol}b. Approximately 5g of each potsherd was sampled and its surface was cleaned using a drill (Dremel 3000) to remove any external contaminants. The cleaned sample was ground to fine powder in a glass mortar & pestle and accurately weighed. The sample was transferred into a culture tube (I) and 5mL of H~2~SO~4~ (sulfuric acid) : MeOH (methanol) were added to it; and the whole was heated (2% v/v, 70 $^{\circ}$C, 1 hour, vortex-mixing every 5 minutes). It is important to check the pH after extraction to examine whether the sample is still acid, for carbonate- rich ceramic fabrics might neutralize acid. If the pH is $\geq$ 3, then more H2SO4 : MeOH should be added.

The H~2~SO~4~ : MeOH solution containing the extract was transferred to the test tube, and centrifuged for 10 minutes (2500 rpm). The clear solution was transferred to another clean culture tube (II) and 2mL of nano-purified water were added. Then, 4 mL of hexane were dropped in the culture tube (I), and vortex-mixed to recover any lipids which are not fully extracted by the methanol solution. The hexane portion was transferred in the culture tube (II) and vortex-mixed with the H2SO4:MeOH solution to extract the lipids. The washing of the culture tube (I) with hexane and vortex-mixing in the culture tube (II) were repeated twice. Then, the hexane portion was transferred to a clean vial. Following this, 2 mL of hexane were added directly to the H~2~SO~4~ : MeOH solution in the culture tube (II), and vortex-mixed with it to extract the remaining lipid residues. The hexane extracts were gathered in a clean vial, and evaporated under a gentle nitrogen blow, and re-dissolved in 300 uL of hexane for GC-MS and GC-C-IRMS.

\subsection{Analysis with GC-MS and GC-C-IRMS}

\subsubsection{High Temperature GC-MS}

The trimethylsilylated TLEs and FAMEs were analyzed by 6890N Network GC system with a 5979 Mass selective Detector from Agilent Technologies at the Sachs laboratory, Department of Oceanography, University of Washington. The GC was equipped with a fused silica capillary column (J&W; DB5-MS; 60m x 0.32 mm; 0.25 B5m film thickness) and the interface was maintained at 110 $^{\circ}$C. The mass spectrometer was operated in the full scan mode. Helium was the carrier gas and the GC oven was programmed as follows: 2 min isothermal at 50 $^{\circ}$C are followed by an increase to 350 $^{\circ}$C at a rate of 10 $^{\circ}$C min^-1^ and following this, the temperature is held at 350 $^{\circ}$C for 10 min. The peaks are identified based on their mass spectral characteristics and GC retention times, and also by comparison with the NIST mass spectral library.

\subsubsection{GC-C-IRMS} The CSIA Analysis was performed using a Thermo GC/C-IRMS system composed of a Trace GC Ultra gas chromatograph (Thermo Electron Corp., Milan, Italy) coupled onto a Delta V Advantage isotope ratio mass spectrometer through a GC/C-III interface (Thermo Electron Corp., Bremen, Germany). A compound identification support for the CSIA laboratory is provided by a Varian CP3800 gas chromatograph coupled onto a Saturn 2200 ion trap MS/MS (Varian, Inc., Walnut Creek, CA U.S.A.). The FAMEs dissolved in hexane were injected in the splitless mode, and separated on a Varian factor FOUR VF-5ms column (30m $\times$ 0.25mm ID, 0.25 micron film thickness). Once separated, the FAMEs are quantitatively converted to CO~2~ in an oxidation reactor at 950 $^{\circ}$C. Following water removal through a nafion dryer, CO~2~ enters the IRMS. The $\delta$^13^C values were corrected using the working standards composed of several FAMEs calibrated against the NIST standard reference materials. Each sample was analysed ten times.

\begin{figure} \centering \includegraphics{figures/SEvsAE.jpg} \caption{The GC chromatograms of the same archaeological sherd sample (KIM014) showing different recovery rates. (a) chloroform : methanol solvent extraction (b): acidified methanol extraction (IS = Internal Standard). In both extractions, the same amount of internal standard was injected. The acidified extraction method showed a much higher recovery rate (more than 20 times) compared with the prevailing chloroform/methanol solvent extraction protocol.} \label{SEvsAE} \end{figure}

\section{Summary}

In this chapter, I have discussed the methodological background, research design and analytical procedure of the organic geochemical analysis. I have briefly outlined the history of the organic geochemical analysis in the discipline of archaeology and elucidated some of the main principles of the method. I have also listed the implications related to the analysis. Lastly, the details about the laboratory experimental processes were elucidated.

SeungkiKwak/Kwak_S_PhD_thesis documentation built on May 9, 2019, 1:22 p.m.