Anal. Chem. 1996, 68, 4402-4408
δ13C Analysis of Cholesterol Preserved in Archaeological Bones and Teeth Andrew W. Stott and Richard P. Evershed*
Organic Geochemistry Unit, School of Chemistry, Cantock’s Close, University of Bristol, Bristol BS8 1TS, U.K.
Cholesterol preserved in archaeological bones and teeth constitutes an important new source of palaeodietary information. A method is described here for the isotopic (δ13C) determination of cholesterol employing a semiautomated sample preparation procedure and the technique of isotope ratio monitoring/gas chromatography/ mass spectrometry (irm/GC/MS). High-temperature gas chromatography (HT-GC) and high-temperature gas chromatography/mass spectrometry (HT-GC/MS) were used to identify the lipids and quantify the cholesterol present in the total lipid extracts. δ13C values are then readily obtained from nanogram amounts (∼50 ng) of cholesterol resolved and determined directly by high-resolution capillary irm/GC/MS of trimethylsilylated total lipid extracts. The protocol developed allows effective processing of the large numbers of samples essential for palaeodietary determinations. Analytical precision and reproducibility have been assessed through multiple sampling of the same skeleton (femur, 9th century). Comparable δ13C values have been obtained from different skeletal members from the same individual. The utility of the approach is demonstrated through a study of the δ13C values of cholesterol isolated from sections of femoral bones of individuals excavated from cemeteries (dated Saxon to 18th century) at a coastal site in the U.K. The mean δ13C value (-22.2 ( 0.3‰, σ ) 0.9) determined for cholesterol in 50 different individuals indicates a strong preference for marine foods by the members of the community extending back over the last ∼1500 years. A minority of individuals exhibited δ13C values as low as -26‰, indicating preferences for terrestrial rather than marine foodstuffs.
carried out following pretreatment of the bone or shell with oxidants, e.g., sodium hypochlorite1 or hydrazine,2 to remove organic matter. Diagenetic carbonate is then eliminated by treatment with acetic acid3 or triammonium citrate.2 The bioapatite is then reacted with concentrated phosphoric acid in evacuated sealed tubes to generate CO32- as CO2 gas, which is then cryogenically trapped prior to analysis using off-line isotope ratio mass spectrometry (IRMS). Isolation of bone collagen, for the purposes of bulk isotopic analyses, proceeds with dissolution of the biomineral phase and extraction of the noncollagenous proteins using either dilute HCl or ethylenediaminetetraacetic acid (EDTA).4 After extensive washing, the collagen is freeze-dried, checked for purity and chemical integrity, and then combusted (1-5 mg of collagen is usually required) in an evacuated quartz tube with a mixture of cupric oxide or silver/copper metal at elevated temperatures. The carbon dioxide and nitrogen gases produced are cryogenically separated and analyzed using IRMS. A refinement of this latter approach involves the isotopic analysis of individual amino acids separated from collagen using cation exchange resin columns.5 This has proven to be a useful criterion for testing the indigeneity of diagenetically altered fossil organic matter such as collagen.4-6 The opportunity now exists for increasing the specificity of stable isotope ratio studies in the palaeodietary field by the use of on-line isotope ratio monitoring/gas chromatography/mass spectrometry.7 This technique enables GC separation and δ13C analysis of a wide range of individual compounds, including hydrocarbons,8 amino acids,9 carbohydrates,10 fatty acids,11,12 etc. Such compounds are derivatized where necessary and separated by capillary GC and then oxidized to carbon dioxide in a cupric oxide packed combustion interface at 850-1300 °C to produce
Palaeodietary reconstruction of ancient populations, individuals, and foodwebs was revolutionized in the 1970s with the introduction of stable carbon and nitrogen isotopic analyses. Since then, studies have focused on both the inorganic, e.g., biomineral constituents of bone, teeth, and shell, and organic components, e.g., collagenous and noncollagenous bone proteins, amino acids, and shell proteins, preserved in archaeological remains. Bone bioapatite (δ13C) and collagen (δ13C and δ15N) bulk isotopic analyses are routinely utilized in palaeodietary studies, providing important information regarding dietary behavior in the past, including (i) protein versus carbohydrate consumption, (ii) consumption of marine versus terrestrial foods, and (iii) consumption of C3 versus C4 photosynthetically derived foods. Up to now, isotopic analyses of skeletal remains have relied on the use of off-line techniques. For example, bioapatite isotopic analyses are
(1) Termine, J. D.; Eanes, E. D.; Greenfield, D. J.; Nylen, M. J. Calcified Tissue Res. 1973, 12, 73. (2) Lee-Thorp, J. A.; Van Der Merwe, N. J. J. Archaeol. Sci. 1991, 18, 343. (3) Hassan, A. A.; Termine, J. D.; Haynes, C. V., Jr. Radiocarbon 1977, 19, 364. (4) Tuross, N.; Fogel, M. L.; Hare, P. E. Geochim. Cosmochim. Acta 1988, 52, 929. (5) Hare, P. E.; Fogel, M. L.; Stafford, T. W.; Mitchell, A. D.; Hoering, T. C. J. Archaeol. Sci. 1991, 18, 277. (6) Serban, A.; Engel, M. H.; Macko, S. A. Org. Geochem. 1988, 13, 1123. (7) Matthews, D. E.; Hayes, J. M. Anal. Chem. 1978, 50, 1465. (8) Freeman, K. H.; Hayes, J. M.; Trendel, J. M.; Albrecht, P. Nature 1990, 343, 254. (9) Macko, S. A.; Fogel-Estep, M. L.; Engel, M. H.; Hare, P. E. Geochim. Cosmochim. Acta 1986, 50, 2143. (10) Macko, S. A.; Helleur, R.; Hartley, G. Org. Geochem. 1990, 16, 1129. (11) Abrajano, T. A.; Fang, J.; Comet, P. A.; Brooks, J. Abstracts of Papers, 203rd National Meeting of the American Chemical Society, San Francisco, CA, Spring 1992; American Chemical Society: Washington, DC, 1992; Abstract 104. (12) Woodbury, S. E.; Evershed, R. P.; Rossell, J. B.; Griffith, R. E.; Farnell, P. Anal. Chem. 1995, 67, 2685.
4402 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
S0003-2700(96)00199-0 CCC: $12.00
© 1996 American Chemical Society
pulses of carbon dioxide for the individual compounds. The isotopic content (δ13C value) of the CO2 is determined on-line using a conventional IRMS. Using this arrangement, δ13C values can be determined on nanogram amounts of individual analytes with precisions of better than 0.5‰.7 Surprisingly few studies have been carried out on the application of irm/GC/MS to the analysis of biomolecules, e.g., lipids, preserved in archaeological materials, with the only two published reports to date relating to the sourcing of lipids in archaeological pottery.13,14 The isotopic analysis of lipids from ancient skeletal remains for the purposes of palaeodietary or palaeoecological reconstruction has up to now been completely neglected. Evidence that lipids (in particular, steroidal compounds) do survive in human remains comes from the few analyses that have been performed on bog bodies15,16 and mummies.17,18 Until recently, detailed studies of the lipids recovered from ancient bones had not been carried out. Evershed et al.19 conducted preliminary lipid studies on a wide variety of bones (animal and human) from differing geographical locations and archaeological contexts using gas chromatography (GC) and gas chromatography/mass spectrometry (GC/MS). The discovery of cholesterol in each bone analyzed and in appreciable concentrations (2-50 µg/g dry weight of bone) led to the proposal of using the δ13C content of cholesterol as an additional source (to be used in conjunction with collagen and bioapatite data) of palaeodietary information. We present here the results of a systematic study aimed at developing a protocol using semiautomated sample preparation procedures and irm/GC/MS for the isotopic determination of cholesterol preserved in a suite of archaeological bones dated Saxon to 18th century. Cholesterol homogeneity was tested by multiple δ13C analysis along a single femur and via measurement of representative skeletal members from different individuals. Larger-scale sampling of the inhumations, utilizing the developed protocol, is shown to provide a means of obtaining valid and reliable δ13C values of cholesterol, essential to resolve the palaeodietary history of this population. EXPERIMENTAL SECTION Archaeological Specimens. Femurs from 50 different individuals were sampled from a large collection of inhumations (1803 skeletons) recovered during excavations (1978-1984) at St. Peters Church, Barton-on-Humber, North Lincolnshire, U.K. Following excavation, bones were stored at 4 °C with adhering burial soil until required for analysis. No unusual palaeopathologies were associated with the samples taken. Sample Preparation. Sections of bone (∼3 cm × 4 cm) were cut from each femur, and adhering soil and exogenous lipids were removed from the bone surfaces by abrasion. After cleaning, each bone sample was dried, crushed under liquid nitrogen in a mortar, and stored at 4 °C. Automated Extraction. To avoid contamination of the bone cholesterol with that from exogenous sources, analyses were (13) Evershed, R. P.; Arnot, K. I.; Collister, J.; Eglinton, G.; Charters, S. Analyst 1994, 119, 909. (14) Evershed, R. P.; Stott, A. W.; Raven, A.; Dudd, S. N.; Charters, S.; Leyden, A. Tetrahedron Lett. 1995, 36, 8875. (15) Evershed, R. P.; Connolly, R. C. J. Archaeol. Sci. 1994, 21, 577. (16) Evershed, R. P. Archaeometry 1992, 34, 253. (17) Kuksis, A.; Child, P.; Myher, J. J.; Marai, L.; Yousef, I. M.; Lewin, P. K. Can. J. Biochem. 1978, 1141. (18) Gulac¸ ar, F. O.; Susinii, A.; Koln, M. J. Archaeol. Sci. 1990, 17, 691. (19) Evershed, R. P.; Turner-Walker, G.; Hedges, R. E. M.; Tuross, N. L.; Leyden, A. J. Archaeol. Sci. 1995, 22, 277.
carried out under ultraclean conditions. Lipid analyses were performed using a robotic workstation (Benchmate, Zymark Corp., Hopkinton, MA). Samples were analyzed using the following analytical procedure: The workstation tares the test tubes (disposable borosilicate glass culture tubes, 16 mm × 100 mm, VWR Scientific Inc.) and then remains in stand-by mode until further instructed. The operator manually adds approximately 2 cm3 of powdered bone to each tube (equivalent of ∼2 g of sample per tube). Analytical blanks were assigned at the start (1st tube), during (every 10th tube), and at the end of the analyses (final tube). Following sample addition, the workstation reweighs the tubes and records the weight in ASCII code on disk. The reagent lines are purged with the appropriate solvent(s) (HPLC grade, Rathburn Chemicals) used in the proceeding extraction stages. A known amount of internal standard [50 µL of a 1 mg mL-1 solution of 5R(H)-cholestane or n-tetratriacontane (Sigma Chemical Co.)] is then added to each tube, followed by a mixture of chloroform/methanol (2:1 v/v; 9 mL). The cannula syringe is rinsed thoroughly after each process. The contents are vortexed at a predetermined speed for 20 s, after which the sample rack is removed from the workstation and placed in an ultrasonic bath, and samples are extracted for 4 × 15 min. Samples are left to stand for 12 h, sonicated briefly, and then centrifuged at 2000 rpm to remove suspended particulates. Samples were returned to the workstation, and the supernatant was filtered to collect 6 mL of extract in a second set of tubes. Filters (Acrodisc, Gelman Sciences, 0.25 µm) were prewetted with chloroform/methanol (5 mL) before filtration. Total extracts were transferred directly to an evaporation unit (TurboVap LV Zymark Corp.). Sample racks were placed in a thermostatic water bath (50 °C) and concentrated to dryness under a stream of nitrogen. Extracts were then dissolved in minimum solvent, transferred to preweighed sample vials, reduced to dryness under a gentle stream of nitrogen, and stored at 4 °C. Derivatization of Total Lipid Extracts. The lipid extracts are reconstituted in solvent (chloroform), and approximately one quarter of the total extract is removed for further analysis. N,OBis(trimethylsilyl)trifluoroacetamide (BSTFA) containing 1% (v/v) trimethylchlorosilane (Sigma Chemical Co.) (30 µL) was added to each extract, heated at 70 °C (∼1 h), and evaporated under a stream of nitrogen. The extract was diluted with an appropriate volume of solvent prior to analysis by gas chromatography (GC), gas chromatography/mass spectrometry (GC/MS), and isotope ratio monitoring/gas chromatography/ mass spectrometry (irm/GC/MS). Instrumental Analysis. Fully automated GC analyses were carried out on a Hewlett Packard 5890 Series II gas chromatograph fitted with a fused silica capillary column (30 m × 0.32 mm i.d.) coated with a 5% phenyl methyl silicone stationary phase (HP5, 0.25 µm film thickness). Samples were diluted in dichloromethane and introduced by on-column injection using a Hewlett Packard 7673 automatic sampler. Following an isothermal hold at 50 °C (2 min), the oven temperature was increased to 250 °C at 10 °C min-1 and then to 325 °C (10 min) at 4 °C min-1. Hydrogen was used as carrier gas, and flame ionization detection (FID) was used to monitor the column effluent. HT-GC was carried out as above, using a fused silica capillary column (15 m × 0.32 mm i.d.) coated with a dimethylpolysiloxane stationary phase (DB-1, 0.1 µm film thickness). Following an isothermal hold at 50 °C (2 min), the temperature was increased to 350 °C Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
4403
(20 min) at 10 °C min-1. Data were analyzed on an Opus V PC using HP Chemstation software. GC/MS analyses were carried out using a Finnigan 4500 quadrupole mass spectrometer (electron voltage, 35 eV; filament current, 0.35 mA; electron multiplier, 2 kV; source temperature, 280 °C) interfaced to a Carlo Erba HRGC 5160 Mega series gas chromatograph. The GC was fitted with a fused silica capillary column (50 m × 0.32 mm i.d.) coated with a dimethylpolysiloxane stationary phase (CP Sil-5-CB, 0.25 µm film thickness). Hydrogen was used as carrier gas. Data were acquired and processed using an INCOS data system. Peak assignments were made by coinjection of authentic standards, followed by comparison of mass spectra and retention times. Automated irm/GC/MS analyses were performed using a Varian 3500 gas chromatograph coupled to a Finnigan MAT DELTA-S isotope ratio monitoring mass spectrometer via a Finnigan MAT combustion interface. The GC column used was as described above for the GC/MS analyses. Helium was used as carrier gas, and the mass spectrometer source pressure was 9 × 10-5 Pa. The temperature program was as follows: after an isothermal hold at 50 °C (2 min), the temperature was increased from 50 to 250 °C at 10 °C min-1, then to 300 °C at 4 °C min-1. The temperature was held at 300 °C (25 min). RESULTS AND DISCUSSION The aim of this investigation was to develop a protocol using semiautomated sample preparation procedures and irm/GC/MS to determine the δ13C values of cholesterol preserved in archaeological bones and teeth for the purposes of palaeodietary and palaeoecological reconstruction. The development of the protocol proceeded with (i) assessment of the isotopic homogeneity of bone cholesterol through multiple sampling of the same skeletal member, (ii) comparison of the δ13C values of cholesterol recovered from representative skeletal members of different individuals recovered from the same archaeological excavation, and (iii) demonstration of the utility of the procedure by largescale sample preparation and measurement of the δ13C values of cholesterol recovered from femoral sections sampled from an English population dated Saxon to 18th century. GC and GC/MS. Analyses of the 50 samples proceeded with the identification and quantification of the bone lipids using GC and GC/MS. Chromatographic separation of cholesterol was achieved using high-resolution capillary GC columns. Two internal standards were used for the purposes of quantification: 5R(H)-cholestane and n-tetratriacontane (n-C34). The latter was chosen for use in those instances where 5R(H)-cholestane coeluted with other compounds. A partial GC profile typical of a lipid extract of a femoral section is shown in Figure 1. The extract is dominated by cholesterol (I) (∼18.5 min retention time), identified unambiguously by GC/MS (M•+ 458, [M - TMSOH]+, m/z 368). A number of cholesterol degradation products (II to IV) were present (Chart 1) representing two dominant degradation pathways, distinguishable by the type of diagenetic congeners present in the lipid extracts. The most dominant degradation product is 3β-hydroxycholest-5-en-7-one (II) (retention time 20.5 min, indicating oxidizing conditions pre- or postburial. The presence of 5R-cholestanol (III, eluting as a shoulder on the cholesterol in Figure 1) and 5β-cholestanol (IV) in somewhat lower abundance than 3β-hydroxycholest-5-en-7-one is probably indicative of anaerobic microbial reduction under waterlogged conditions.19 Cholesteryl fatty acyl esters (V) were confirmed unambiguously (m/z 368, [M - RCO2H]+), eluting at longer retention 4404
Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
Figure 1. Partial HT-GC chromatogram of trimethylsilylated lipid extract recovered from femoral bone of skeleton SK554 BH 80 (9th10th century; typical of all 50 lipid extracts from Barton-on-Humber specimens). Peak assignments: (1) 5β(H)-cholestanol, (2) cholesterol, (3) 5R(H)-cholestanol, (4) n-octacosanol, (5) hopan-22-ene (diploptene), (6) 3β-hydroxycholest-5-en-7-one, (7) n-tetratriacontane (internal standard), (8-11) unknown compounds, dominant m/z 311 base peak, (12-14) cholesteryl fatty acyl esters.
Chart 1
times (∼25-28 min) than cholesterol, by HT-GC/MS analysis. The nature of the fatty acyl moieties in cholesteryl esters cannot be determined from their EI spectra due to the lack of diagnostic fragment ions. Their more detailed structural investigation would require the use of chemical ionization techniques20-23 which were beyond the scope of this study. Evershed et al.19 in a previous (20) Evershed, R. P.; Goad, L. J. Biomed. Environ. Mass Spectrom. 1987, 14, 131. (21) Evershed, R. P.; Prescott, M. C.; Spooner, N.; Goad, L. J. Steroids 1989, 53, 285. (22) Evershed, R. P.; Goad, L. J. In Studies of Natural Products Chemistry; Attaur-Rahman, Ed.; Elsevier: Amsterdam, 1991; p 447. (23) Evershed, R. P. In Developments in the Analysis of Lipids; Tyman, J. H. P., Gordon, M. H., Eds.; Royal Society of Chemistry: Cambridge, 1994; p 123.
study also noted the presence of cholesteryl fatty acyl esters in the lipid extract of a section of a C9th-C10th human midshaft femur (Chimney farm, Bampton, Oxfordshire, U.K.) and inferred that these were probably endogenous to the bone at the time of death of the individual arising from blood-borne cholesterol. Diploptene (hopan-22-ene; M•+ 410, base peak m/z 189; VI), a biological marker characteristic of bacteria, was identified at ∼19.5 min retention time and most likely derives from soil microorganisms contributing to the decay of the tissue during burial. No triacylglycerols (the most abundant components of the lipid extracts of fresh mammalian bone) were observed in the total lipid extracts analyzed, indicating extensive degradation of the fatty components of skeletal tissue (i.e., bone marrow) during burial. Saturated (C12:0, C14:0, C16:0, and C18:0) and unsaturated (C18:1) free fatty acids were detected in low concentrations at shorter retention times than shown in Figure 1, together with lower concentations of n-alkanols (C14 and C18) and n-alkanes (C22-C23). The latter probably arise from the burial soil or decay microorganisms. Previous work19 has shown that the cholesterol and related diagenetic products are endogenous components of the bones in which they occur and do not arise from the burial environment, i.e., the soil. Failure to detect cholesterol in any of the laboratory blanks that were analyzed in this work confirms the cholesterol is not the result of laboratory or procedural contamination. irm/GC/MS: Correction for the Derivatizing Carbon. Functionalized compounds containing hydroxyl moieties, e.g., cholesterol, are usually analyzed as trimethylsilyl (TMS) ethers to improve their GC behavior.24,25 A cholesterol standard with a firmly established δ13C value was derivatized (BSTFA) such that a correction for the three carbon atoms incorporated on derivatization could be made (cholesterol, BDH Chemicals). The derivatization method ensured that the carbon (from the trimethylsilyl group) incorporated into the cholesterol derived from one source, namely the BSTFA. The contribution of the derivatizing carbon to the overall δ13C value of the cholesterol in a sample was accessible by irm/GC/MS analysis of the authentic standard of known δ13C value using the method of Jones et al.:24
27δOH + 3X ) 30δOTMS
where δOH is the δ13C value for the underivatized cholesterol (27 carbon atoms), δOTMS is the δ13C value for the derivatized cholesterol (30 carbon atoms), and X is the δ13C value of the derivatizing TMS group (three carbon atoms). Ten separate aliquots of the authentic cholesterol standard of well-established δ13C value were derivatized to calculate the δ13C value of the three derivatizing carbon atoms. The δ13C value of the TMS group containing three carbons was calculated to be -125.2‰, i.e., 3X. Thus, the δ13C value of X is equal to -41.8‰. The standard is run in triplicate during each irm/GC/MS analysis to test for batch variance. Isotopic Homogeneity of Cholesterol. To assess the isotopic homogeneity of cholesterol in bone, it was necessary to perform multiple sampling of the same skeletal member (SK554 BH80; 9th century) to ensure that consistent δ13C values could be measured throughout a single sample. Bone sections (3 cm (24) Jones, D. M.; Carter, J. F.; Eglinton, G. Biol. Mass Spectrom. 1991, 20, 641. (25) Evershed, R. P. In Handbook of Derivatives for Chromatography; Blau, K., Halkett, M., Eds.; John Wiley and Sons Ltd.: New York, 1993; p 51.
Figure 2. Plot of the absolute concentrations (µg g-1 dry weight of bone) and δ13C values of cholesterol as a function of distance along an archaeological human femur (skeleton code SK554 BH 80, 9th century). Sample distances: FA, 3; FB, 7; FC, 12; FD, 17; and FE, 22 cm. The range of the δ13C scale represents approximately the maximum range that can exist in nature.26,27
× 4 cm) were taken at specific intervals, 3 cm apart, along the length of an intact right femur dated at 9th-10th century. Figure 2a shows a plot of the absolute concentrations of cholesterol (µg g-1 dry weight of bone) measured along the length of the femur (SK554 BH80). Cholesterol was observed in each femoral section studied; more importantly, it was found in quantities sufficient to be analyzed directly by high-resolution capillary irm/GC/MS (>50 ng). A unimodal distribution of values was obtained along the bone, ranging from 10.7 µg g-1 near the femoral head to a maximum of 29.7 µg g-1 along the midshaft (12 cm). Variance in the observed concentrations of cholesterol is presumed to relate to localized differences in the relative extent of bone decay rather than having any physiological or metabolic significance.19 Triplicate irm/GC/MS analyses were carried out on each of the five femoral lipid extracts to determine the δ13C isotopic signature of the bone cholesterol along the length of the femur (Figure 2b). δ13C values measured on cholesterol from along the femoral section are shown in Table 1. Statistical analysis of the scatter between triplicate analyses across the femur (from FA to FE, n ) 15) showed that the precision of the technique was better than (0.3‰. Comparison of the highest (FE ) -22‰) and lowest (FA ) -22.6‰) means showed that there were no significant differences (Student’s t test, t ) 5.43; P < 0.001) in the δ13C values of cholesterol across the femur, exemplifying the use of irm/GC/ MS as a highly reproducible analytical tool for the accurate determination of the δ13C values of individual lipids preserved in archaeological bones. Most importantly, variations in the δ13C values of cholesterol are marginal, indicating the homogeneity of the cholesterol isotopic signal within the femur. Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
4405
Table 1. Comparison of the Cholesterol δ13C Values Measured along a Single Femur femoral sectionsa
mean (n ) 3) standard deviation (σ) 99% confidence limits pooled standard deviation (A vs E) critical t ) 5.43 (four degrees of freedom) a
FA
FB
FC
FD
FE
-22.6 0.06 (0.1
-22.4 0.05 (0.2
-22.4 0.16 (0.3
22.5 0.08 (0.1
-22.0 0.19 (0.3
0.14 P < 0.001
See Figure 2 for femoral profile.
Table 2. Comparison of the Cholesterol δ13C Values Measured in Representative Skeletal Members from Four Individuals skeletona (age)
mean δ13C of skeleton (n ) 7) variance standard deviation (σ) 95% confidence limits a
SK1158 (C10-11th)
SK1674 (C12-13th)
SK1773 (C14-15th)
SK1409 (C14-17th)
-21.5 0.51 0.71 (0.5
-22.5 0.24 0.49 (0.4
-20.8 0.22 0.47 (0.4
-21.5 0.47 0.68 (0.5
See Figure 3.
δ13C Analysis of Cholesterol in Different Skeletal Members. Seven representative skeletal members, including rib, vertebral arch, radius, femur, pelvis, cranium and teeth, were sampled from four complete adult skeletons (SK1158, 10th-11th century; SK1674, 12th-13th century; SK1773, 14th-15th century; and SK1409, 14th-17th century). Lipid extraction and quantification were carried out using the methods described above. δ13C analyses (Table 2) were carried out on all samples and corrected for the contribution of the derivatizing carbon atoms. Figure 3 shows a comparative plot of the absolute concentrations of cholesterol and the corresponding corrected δ13C values recovered from the representative bones of the four individuals. Maximum concentrations of cholesterol were recovered from the teeth of all four individuals at 37, 30, 38, and 114 µg g-1 (SK1158, SK1674, SK1773, and SK1409, respectively). Femoral cholesterol was measured at 19, 7, 22, and 18 µg g-1, respectively. Corrected mean δ13C values for the cholesterol plot as follows: SK1158 (-21.5 ( 0.5‰, σ ) 0.71), SK1674 (-22.5 ( 0.4‰, σ ) 0.49), SK1773 (-20.8 ( 0.3‰, σ ) 0.47), and SK1409 (-21.5 ( 0.5‰, σ ) 0.68). Based on these observations, femurs and teeth are our preferred sources of cholesterol δ13C information since (i) teeth were found to contain the greatest concentrations of cholesterol on a dry weight basis, (ii) femurs and teeth are the most often recovered skeletal remains found at sites of archaeological excavations, and (iii) the femur is the largest bone in the human skeleton, providing sufficient opportunities for sampling for isotopic analyses without causing interference to other types of skeletal investigations. The δ13C isotopic data presented above are vital to the success of future research in this field, since they confirm that the δ13C values from different skeletal members are constant for a given individual. Based on the results obtained so far, it is apparent that the clear and consistent differences in the δ13C values of cholesterol that exist between skeletons undoubtedly reflect differences in their dietary preferences. Measurements using irm/GC/MS of the δ13C content of cholesterol recovered from representative skeletal 4406 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
Figure 3. (a, top) Plot of the absolute absolute concentrations (µg g-1 dry weight of bone) and (b, bottom) δ13C values of cholesterol from representative skeletal members sampled from four different inhumations (skeleton code SK1158, SK1674, SK1773, and SK1409). C, cranium; F, femur; Ri, rib; T, tooth; P, pelvis; Ra, radius; and V, vertebra. See caption to Figure 2 for explanation of range of δ13C axis.
members (six bones and a tooth) of the different individuals serve to demonstrate the simplicity and reliability of accessing palaeo-
Figure 4. Partial irm/GC/MS profile of individual lipids extracted from skeleton SK554 BH 80, 9th-10th century. The upper chromatogram represents the instantaneous ratio of the m/z 45/44 ions, while the lower chromatogram represents the m/z 44 ion current.
dietary information from archaeological bones (and teeth) using a single biomolecule such as cholesterol. Utility of the Protocol for Large-Scale Palaeodietary Studies. The protocol developed herein uses semiautomated methods and irm/GC/MS, which allows effective processing of the large numbers of samples essential for palaeodietary determinations. Femoral sections (1.8-2.6 g) were sampled from 50 different inhumations. Absolute concentrations of cholesterol were measured relative to an internal standard in each total lipid extract, followed by the determination of δ13C values by irm/GC/MS. Figure 4 shows a partial irm/GC/MS chromatogram of a typical bone total lipid extract recovered from SK554 BH80 from the coastal population at Barton-on-Humber. The top chromatogram (Figure 4a) represents the instantaneous ratio of the m/z 45 relative to the m/z 44 ion currents, while the lower trace shows the chromatogram of the m/z 44 ion as a function of time (Figure 4b). Chromatographic baseline resolution of cholesterol is essential for valid isotopic measurements to be made and was achieved from the crude total lipid extracts (trimethylsilylated) using a high-resolution capillary GC column (50 m CP-Sil 5CB; see Experimental Section for further details). Absolute concentrations of cholesterol recovered from the 50 femoral sections ranged from 9 µg g-1 dry weight of bone to 396 µg g-1. Triplicate isotopic measurements carried out on a selection of these samples showed that the analytical reproducibility was (0.3‰. Figure 5 shows the δ13C values measured from the 50 different individuals. It is apparent from these isotopic data that significant differences in the δ13C values for cholesterol exist between some individuals, which must reflect dietary variation within the population. The majority of the individuals have δ13C cholesterol values of about -23‰; however, several individuals plot to the extreme left (about -26‰) and right (about -20‰) of these values (Figure 5). In a study of pigs raised isotopically on pure diets,26,27 the δ13C value for the bone cholesterol biosynthesized by a pig raised on a pure C3 cereal-based diet was about -28‰. All cholesterol δ13C values measured for the 50 femoral sections of the archaeological humans were enriched in 13C compared with the C3 modern pig cholesterol. Since pigs are physiologically similar to humans and (26) Stott, A. W., Evershed, R. P., Tuross, N. Abstracts of American Chemical Society, Division of the History of Chemistry, 209th ACS National Meeting, Anaheim, CA, April 2-6, 1995. (27) Stott, A. W., Davies, E. J., Tuross, N.; Evershed, R. P. Naturwissenshaften, in press.
Figure 5. Plot of the δ13C values of cholesterol measured for 50 Barton-on-Humber skeletons, showing evidence for diets reflecting differing preferences for marine (13C-enriched) versus terrestrial C3 (13C-depleted) foods in the population (isotopic values plotted in rank order). N.B.: Modern C3 pig reference cholesterol δ13C value ≈ -28‰.
foodstuffs utilizing the C4 photosynthetic pathway were not available for consumption over the chronological burial period at Barton-on-Humber, the δ13C values recorded from the 50 individuals indicate that their dietary subsistence was strongly influenced by the consumption of marine foodstuffs. Some individuals exhibited δ13C values closer to the reference pig cholesterol data, indicating a greater proportion of C3-based foods in their diets. The immediate aim of this large-scale study is to correlate cholesterol δ13C values with collagen δ13C and δ15N isotopic data to gain further insights into the subsistence behavior of this coastal population over a relatively long chronological time scale. In the long term, the δ13C values of cholesterol measured from the bones and teeth of individuals recovered from archaeological excavations of coastal and inland sites in the U.K. are being used to compare the importance of marine foods in their respective diets. CONCLUSIONS An analytical protocol using semiautomated sample preparation techniques and irm/GC/MS has been shown to be a valuable tool for the processing of large numbers of samples essential for palaeodietary studies. The utility of the developed protocol is demonstrated by the analysis of the isotopic signal of cholesterol present in human femurs from inhumations excavated at St. Peters Church, Barton-on-Humber, North Lincolnshire, U.K. These procedures serve to demonstrate the relative ease and analytical accuracy with which δ13C information can be derived from a single molecule such as cholesterol on a relatively short laboratory time scale compared with conventional bone collagen and bioapatite analyses. The rationale behind using the δ13C value of cholesterol is that it more accurately reflects the isotopic content of the bulk diet than the collagen, the latter being biased toward dietary protein. Hence, cholesterol can be used in preference to bone and tooth apatite as a source of 13C data, reflecting the carbohydrate constituents of an individuals diet. Recovering valid δ13C information from bioapatite and collagen may be hindered by two main factors: diagenetic alteration and contamination from humic constituents. Alteration of the bioapatite or collagen by the introduction of exogenous solutions or organic matter by these two processes changes their overall δ13C isotopic signature. Cholesterol recovered from archaeological bones and teeth Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
4407
provides a single compound, whose structure can be determined unambigously by GC/MS, and allows valid isotopic measurement to be recorded by irm/GC/MS. Moreover, the measurement of δ13C values of individual lipid moieties preserved in mineralized tissues provides a method of investigating carbon sources and cycling on a potentially shorter time scale biologically28,29 when compared to the slower turnover rates of carbon measured in bone collagen. Consequently, δ13C information derived from cholesterol should provide reliable information concerning the migration or seasonal variation in diet of populations due to the shorter turnover rate of the biomolecule. Hence, the δ13C content of cholesterol, isotopically labeled with the 13C signature of the foodstuffs consumed prior to death, should, therefore, reflect the most recent diet of the individuals, information which is not obtainable from isotopic measurements of collagen or bioapatite or, indeed, any other source.
ACKNOWLEDGMENT NERC is thanked for financial support for mass spectrometry facilities (GR3/2951, GR3/3758, and FG6/36/01) and a research grant (GR3/9543). Mr. Jim Carter and Mr. Andrew Gledhill are thanked for technical assistance. We thank Dr. Noreen Tuross for the encouragement and support she has given to us during this work. Dr. Juliet Rogers (Rheumotology Unit, Bristol University) is thanked for provision of Barton-on-Humber samples, and Miss Susan Jim is thanked for assistance in the preparation of the femurs for lipid extraction.
Received for review September 6, 1996.X
February
29,
1996.
Accepted
AC960199R (28) Sabine, J. R. Cholesterol; Dekker: New York, 1977. (29) Guo, Z. K.; Luke, A. H.; Lee, W. P. Schoeller, D. Anal. Chem. 1993, 65, 1954.
4408
Analytical Chemistry, Vol. 68, No. 24, December 15, 1996
X
Abstract published in Advance ACS Abstracts, October 15, 1996.
