4 DISCUSSION

The primary change made in the derivatization procedure of Hušek (1991) involved elimination of chloroform, water, and acid from the injection solvent, so that stationary phase stripping of the column would not occur. This procedure involved transferring the chloroform solution of the derivatives to a clean vial, extracting the aqueous acid solution a second time with chloroform, drying the combined chloroform layers over sodium sulfate, evaporating the chloroform to dryness, and using benzene as the injection solvent.

Column activity can result from partial stripping of the stationary phase and result in peak resolution problems and tailed peaks. To improve resolution, Hušek (1991) investigated columns coated with a number of different stationary phases and found that OV-1701 gave the best compromise between resolution and analysis time. We found this column to be prone to activity problems, so a number of polar stationary phases (OV-1701, DB-1701, DB-WAX, and HP-INNOWAX) were tested to determine which was most resistant to activity problems. Although all provided adequate resolution, HP-INNOWAX proved to be the most robust and could be heated to the highest temperature to bake out contaminants.

The HP-INNOWAX stationary phase is chemically very similar to CARBOWAX, a stationary phase used for separation of fatty acid methyl esters derived from oils and lipids. It is indeed fortuitous that the capillary column best suited to separate ECF derivatives of amino acids may also be used in the identification of drying oils and lipids. Conservation science laboratories already equipped with a gas chromatograph for oil identification may perform protein identification without the additional expense of obtaining a second column. Furthermore, the inevitable instrument setup delays due to changing columns are eliminated. Most important, it is possible to forgo the major expense of buying HPLC equipment to perform protein identification.

In general, vapor-phase hydrolysis performed satisfactorily for samples below 100 μg in weight. Weight percentages of protein, after correction for the hydrolysis yield by the rabbit skin glue reference standard, were accurate to within 10% relative. Errors were much larger for samples exceeding 100 μg, especially for egg yolk. Presumably the hydrophobic nature of the lipids in egg yolk prevented complete vapor-phase hydrolysis.

Very little contamination was observed in the blank analyses, due in large part to the fact that in the vapor-phase hydrolysis technique, preconcentration of the hydrolysate in order to accommodate low amino acid concentrations is not required. This feature is important for the study of microsamples removed from objects. Identification of the contaminants was not performed in this study.

Of the 20 common amino acids, a few are sensitive to acid hydrolysis conditions (Pickering and Newton 1990). Asparagine and glutamine are converted to aspartic acid and glutamic acid, respectively. Serine and threonine concentrations are reduced by approximately 10% through oxidation; if deemed necessary, corrections in yield can be made. Tryptophan is destroyed completely by hydrolysis. Methionine is converted to methionine sulfoxide unless first reacted with 2-mercaptoethanol. The yield of cysteine can be improved by conversion to cysteic acid with performic acid.

The concentrations of amino acids in protein hydrolysates can be affected by the presence of pigments and other inorganic compounds (Halpine 1992; Grzywacz 1994). These so-called matrix effects must be evaluated thoroughly for a given analytical method to understand which amino acids are free of them and which are most affected by them, so that a reliable protein identification scheme may be developed. The data in tables 1, 2, and 3 were all obtained in the absence of pigments; hence they are useful for reference purposes.

Furthermore, Karpowicz (1981) has shown that certain amino acids are strongly affected by aging (especially methionine and cysteine). Thus, the combined effects of pigments and aging may alter the normal profile of amino acids in paint samples to such an extent that results obtained from normal identification procedures may be compromised. These effects have been studied in relation to the ECF method and are the subject of another paper (Schilling and Khanjian, in preparation).

Compounds other than amino acids may be derivatized by ethyl chloroformate. For example, peaks from ethyl esters of palmitic, stearic, and oleic acid appear in the chromatograms of egg tempera (see fig. 2 for whole egg) and drying oils. The fatty acids originate from the triglycerides and are formed by acid hydrolysis and ECF-esterification. Azelaic acid, formed during the drying of oils rich in linoleic acid or linolenic acid, appears as a prominent peak in samples that contain drying oils (linseed oil, walnut oil, and poppy oil) and is much smaller in chromatograms for egg yolk and whole egg samples (Mills and White 1987).

Oleic acid may be present in sufficiently high concentration in some egg yolk samples to cause an overlap problem with threonine, resulting in erroneously high threonine concentrations. If an exact determination of threonine is required, we found that residual free fatty acids in sample hydrolysates may be extracted with chloroform prior to derivatization. Concentrations of free amino acids, which are nearly insoluble in chloroform, are not affected by the treatment. Extraction was not performed for the samples in this study, so the data for threonine in egg yolk should be considered approximate.

In general, good agreement is exhibited between the literature data and the data obtained by the ECF method, demonstrating that the level of accuracy is high when any of the methods is used to test pure proteinaceous substances. The small variations displayed in the tables may be a result of genuine compositional differences of the materials tested or of actual methodological systematic errors (such as matrix effects, derivatization yields, or interferences). To identify the reasons behind the small discrepancies would require analytical round-robin testing of a set of standard protein reference materials. This step is hardly justified in view of the overall agreement exhibited by the data.

The compositions of gelatin and fish glues tend to be more variable than those of the collagen samples, especially regarding the concentrations of hydroxyproline, alanine, threonine, phenylalanine, and proline. This variability may be due to improper container labeling or may reflect different sources for the gelatin. A few samples of fish glue and isinglass exhibited reduced proline and hydroxyproline, in keeping with observations made by White (1984).

We originally attributed the presence of amino acids in the plant gums to contamination of the samples, a conclusion that Keck and Peters (1969) also reached. But Sharon (1975) noted that amino acids are normally present in plant gums in the form of glycoproteins. Glycoproteins consist of polypeptide chains to which carbohydrate moieties are attached. The amino acids form bonds to specific carbohydrates (for example, arabinose bonds preferentially to hydroxyproline). It was, therefore, not a coincidence that high levels of hydroxyproline should be detected in gum arabic samples, because arabinose is one of the principal sugars of gum arabic.

This information has a substantial impact on the identification of proteins using amino acid composition. One of the principal indicators that glue or gelatin may be present in a sample is a high level of hydroxyproline. Lower levels of hydroxyproline in paint samples are often assumed to originate from particles of glue-based ground inadvertently sampled with the paint. In light of the above results, care should be taken when interpreting amino acid composition data. One way to avoid difficulties is to consider the relative amounts of all amino acids in the sample. As mentioned above, glue contains very high concentrations of glycine and proline in addition to hydroxyproline, while gum arabic does not.

Considering the results from testing of proteinaceous contaminants, and given the likelihood that an object may someday become the subject of an analytical investigation, a few simple precautions should be taken to minimize the accumulation of unwanted proteins. Deposition of amino acids from skin can be reduced by wearing gloves when handling objects (already a common routine). The practice of cleaning painted objects with saliva should be discouraged. Furthermore, the wetting of paintbrushes with saliva to pick up samples for testing must be completely avoided. Unfortunately, little can be done to prevent exposure of outdoor objects to proteinaceous contamination.

An important component of any scheme for identifying proteins is a collection of data for reference materials (such as the composition of eggs or gums). These data may be obtained by in-house testing of proteinaceous materials (if a suitably large collection of proteinaceous materials is available) or by consulting the literature. Only the literature data expressed in terms of weight (or molar) percentages are useful; peak areas or peak area ratios are practically useless outside the laboratory in which they were compiled. Thus, if one is engaged in amino acid analysis of samples removed from art objects, it would be of great service to the conservation community to report amino acid data in terms of concentration. In so doing, conservators can build a significant database that will deepen our understanding of artists and their methods.