Charting a Future Course for Organic Residue Analysis in Archaeology

Abstract

Working hypotheses, which draw upon as many relevant disciplines as possible to derive the maximum information from a very limited database, are key to the highly interdisciplinary field of organic residue analysis in archaeology, a branch of biomolecular archaeology. Archaeology and chemistry are most important for effectively developing and testing such hypotheses, but botany, zoology, geology, etc. also need to be taken into account. Archaeologically, the goal is to obtain as many relevant samples as possible from the best preserved and dated contexts, which have been subjected to the least degradation and disturbance by later natural processes and human handling, including washing and conservation treatment. Chemically, molecular biomarkers of natural products need to be defined and identified by the best and most appropriate techniques, together with bioinformatics searches and assessment of degradation. With ever-improving techniques and new data, previously analyzed samples need to be retested and hypotheses possibly reformulated. Consideration of three case studies illustrates this holistic approach to inductive hypothesis generation and deductive testing: (1) new chemical findings that attest to grape wine in amphoras on board the 14th c. B.C. Uluburun ship, the earliest recorded Mediterranean wreck; (2) recently published research on beeswax/mead in Chalcolithic Israel and Neolithic China and Poland; and (3) recent articles on milk products from 2nd millennium B.C. Central Asia and Neolithic Poland. Potential pitfalls leading to weak hypotheses and mistaken conclusions are described, and a more productive approach is proposed.

Appendices

Appendix 1: LC-MS-MS Analyses

In reanalyzing the ancient Egyptian sample from Scorpion I’s tomb at Abydos, we followed the general methodology of Guasch-Jané et al. (2004), with modification of the extraction procedure to detect tartaric acid/tartrate in the negative mode (McGovern et al. 2009).

An even more refined extraction procedure was found to enhance the detection of tartaric acid/tartrate by LC-MS-MS for the Uluburun terebinth resin samples. Whole pieces of resin were stirred overnight in 1.5 mL of ammonium hydroxide (2.8 % by vol) and 1.5 mL methylene chloride. The mixture was centrifuged for at least 10 min at 4400 rpm to clarify the layers and cause any remaining materials and emulsions to precipitate. The upper basic aqueous layer was then removed, reduced in volume, filtered through a 0.45-μm membrane, placed into 200 μL vial inserts, and analyzed.

Two MRM transitions (149→87.1 and 149→73), not just one as we had previously done, at a specific retention time (0.80 min) were monitored for tartaric acid, as can be seen on Fig. 5. Together, these transitions and the resulting daughter ions are definitive for the acid’s presence/absence. Monitoring both transitions has not been applied to ancient samples before and provides very strong evidence for tartaric acid/tartrate in KW 181, KW 215, and probably KW 102.

The LC mobile phase conditions were also modified to achieve better peak shapes, as follows: A (water, 0.1 % v/v formic acid) and B (acetonitrile, 0.1 % v/v formic acid) had an initial composition of 98 % phase A and 2 % B, which was changed to 50 % A/50 % B over a 2-min period, and then equilibrated back to 98 % A/2 % B.

Accentuating the detection of a tartrate salt by this new method takes advantage of the much lower solubilities of potassium bitartrate and calcium tartrate, about 4 and 0.3 g/L in cold water, respectively, compared to the acid’s solubility of about 1400 g/L at 20 °C (Singleton 1995: 68). The salts readily form in wine and precipitate out in the lees. Calcium can interchange with potassium in highly calcareous pottery fabrics, which are characteristic of Eastern Mediterranean amphoras.

GC-MS was employed to further test KW 215, which had given a positive tartaric acid/tartrate by LC-MS-MS. The resin sample was completely dissolved in methylene chloride and treated with a small amount of formic acid to acidify any tartrate present to tartaric acid. Silylation was carried out with BSTFA (N,O-bis(trimethyl-silyl)trifluoro acetamide). The sample was injected splitless onto an HP-5MS column (5 % phenyl methyl siloxane) of an Agilent HP-6890 gas chromatograph equipped with a Hewlett-Packard 5973 mass selective detector. The key silylated tartaric acid ion at 219⁺ was tentatively identified by selected ion monitoring which enhances sensitivity.

Both LC-MS-MS and GC-MS yielded an approximately 0.25-ppm concentration of the acid/salt for sample KW 215, which was close to the GC-MS detection limit and only achievable in the selected ion mode.

Appendix 2: FT-IR Analyses

Although potentially many absorptions in the standard FT-IR region of 4000–400 cm⁻¹ can be equivocal, careful observation, combined with rigorous methodology, enables the critical peaks for individual compounds in mixtures to be sorted out. Moreover, fine spectral details can be distinguished by extracting an ancient sample separately with methanol and other organic solvents (e.g., chloroform). While chloroform selectively extracts and accentuates the FT-IR spectra of low-polarity compounds, methanol highlights those of high-polarity compounds.

Figure 6 and Table 2 illustrate how powerful a combined solvent approach can be in distinguishing the carbonyl and acid hydroxyl absorptions of synthetic L-tartaric acid from those of ancient and modern Pistacia sp. resin. The higher-polarity tartaric acid, which was extracted by methanol, has a distinctive doublet in the 1740–1720 cm⁻¹ carbonyl region, with a less intense shoulder at the lower wave number (frequency). Its hydroxyl absorption occurs in the 1450–1430 cm⁻¹ region. By contrast, the carbonyl of the lower polarity terebinth resinous acids, which were extracted by chloroform, have a single intense absorption at 1710–1700 cm⁻¹, and their hydroxyl absorption is in the 1470–1445 cm⁻¹ region.

Fig. 6

The FT-IR analysis of L-tartaric acid is clearly different at key carbonyl stretch and hydroxyl bend absorptions from those of ancient (KW 215) and modern Pistacia sp

Table 2 Comparison of key FT-IR absorption of methanol and chloroform extracts of L-tartaric acid, tartrate, and Pistacia sp. resin

In light of their marked spectral differences, the discussion of the FT-IR spectra of tartaric acid and Pistacia sp. resin by Stern et al. (2008) (pp. 2197–2198) is confusing. They state that their Uluburun resin samples all have absorptions in the 1740–1720 cm⁻¹ region, similar to those of tartaric acid. They then dismiss the FT-IR evidence. Yet, their own Fig. 4 shows that the carbonyl peak (viz., 1710–1700 cm⁻¹) of the Uluburun resin samples is significantly lower and does not overlap with that of tartaric acid.

The FT-IR spectrum for syringic acid also differs from that of tartaric acid and Pistacia sp. resin, as Stern et al. (2008) show in their Fig. 2. Its carbonyl stretch is a single intense absorption at 1690 cm⁻¹, and its OH bend is at 1450 cm⁻¹.

The blanket statement by Stern et al. (2008) (p. 2197) that “these absorption bands are shared by a huge variety of organic substances and are clearly not restricted [added emphasis] to tartaric acid/tartrate or terebinth resin” is far from being clearly established and undermines the value of infrared spectrometry. For the spectrometrist, including Moreira and Santos 2004 which Stern et al. cite in support of their contention, the hydroxyl and carbonyl absorption bands are important discriminators of organic compounds. Certainly, Moreira and Santos would not confuse the spectrum of syringic acid with that of tartaric acid, as implied by Stern et al. (2008) (p. 2195). Stern et al. also assert that interferences from ethanol and acetic acid pose a problem in interpreting the FT-IR spectra of wine, again citing Moreira and Santos 2004. Since such compounds have volatilized and disappeared in ancient wine samples, their point is not relevant.

In summary, it is very important to carefully scrutinize and statistically deconvolute an IR spectrum, and also to run searches of unknown archaeological samples against the large databases which are now available. Stern et al. do not appear to have exercised such care in their analyses, and thus they write off the FT-IR data too quickly.

Three subsidiary points should also be noted:

• 1. Even if it were possible to confuse the IR spectra of L-tartaric acid and syringic acid (Stern et al. 2008: 2195), this is irrelevant to the analysis of an ancient wine sample. Syringic acid is not a “derivative” of malvidin (p. 2189), one of the red pigments in wine, but can only be produced as a breakdown product of the latter by alkaline fusion, which was beyond the expertise of ancient peoples. In fact, it is malvidin, not syringic acid, which is present in modern red wine at 200 mg/L, according to Singleton (1995: 70) (see below).

• 2. It is very puzzling that the modern red wine, analyzed by Stern et al. (2008) (p. 2196, Fig. 3), is dominated by calcium tartrate. Tartaric acid, at a concentration of about 1400 g/L at 20 °C, is the principal acid in wine and should be detected. Moreover, any tartrate that has precipitated out should be in the form of potassium bitartrate, not calcium tartrate. Again, Stern et al. do not detail their extraction procedure, if any, which might explain these anomalies. Their explanation for the origin of calcium tartarte in ancient samples is equally suspect.

• 3. Stern et al. (2008) do not appear to have included “ancient reference samples” in their databases, comparable to the inscribed Malkata amphoras of our studies. If they had, they might have been able to detect differences in the FT-IR spectra of “aged” and modern wine samples.

  Together and apart from the lack of care in the interpretation of the FT-IR spectra, the subsidiary points highlight the confusing, ill-founded argumentation of Stern et al. (2008).