Anal Bioanal Chem (2007) 388:201–206 DOI 10.1007/s00216-007-1206-2
ORIGINAL PAPER
Identification of proteinaceous binders used in artworks by MALDI-TOF mass spectrometry Stepanka Kuckova & Radovan Hynek & Milan Kodicek
Received: 4 December 2006 / Revised: 12 February 2007 / Accepted: 13 February 2007 / Published online: 6 March 2007 # Springer-Verlag 2007
Abstract Proper identification of proteinaceous binders in artworks is essential for specification of the painting technique and thus also for selection of the restoration method; moreover, it might be helpful for the authentication of the artwork. This paper is concerned with the optimisation of analysis of the proteinaceous binders contained in the colour layers of artworks. Within this study, we worked out a method for the preparation and analysis of solid samples from artworks using tryptic cleavage and subsequent analysis of the acquired peptide mixture by matrixassisted laser desorption/ionisation time of flight mass spectrometry. To make this approach rational and efficient, we created a database of commonly used binders (egg yolk, egg white, casein, milk, curd, whey, gelatine, and various types of animal glues); certain peaks in the mass spectra of these binders, formed by rich protein mixtures, were matched to amino acid sequences of the individual proteins that were found in the Internet database ExPASy; their
S. Kuckova (*) Department of Chemistry and Chemical Education, Charles University, M.D. Rettigové 4, 116 39 Prague 1, Czech Republic e-mail:
[email protected] R. Hynek : M. Kodicek Department of Biochemistry and Microbiology, Institute of Chemical Technology, Technická 3, 166 28 Prague 6, Czech Republic Present address: R. Hynek Enzyme and Protein Chemistry Group, BioCentrum-DTU, Technical University of Denmark, Søltofts Plads, Building 224, 2800 Kgs. Lyngby, Denmark
cleavage was simulated by the program Mass-2.0-alpha4. The method developed was tested on model samples of ground layers prepared by an independent laboratory and then successfully applied to a real sample originating from a painting by Edvard Munch. Keywords Matrix-assisted laser desorption/ionisation time of flight mass spectrometry . Artworks . Proteinaceous binders . Database
Introduction The colour layer of artworks is composed of pigments and organic binders. The identification of the organic binder is important for restoration because knowledge of it can help reveal the painting technique used by the artist (Table 1). The proteinaceous binders and some other organic pigments contain proteins characteristic for each material [1] and, consequently, the proteins can serve as identification matches of these materials. So far, this identification has been performed using high-performance liquid chromatography, gas chromatography (GC)–mass spectrometry (MS), pyrolysis–GC-MS and Fourier transform IR spectroscopy. These analytical methods have not always been successful, namely because of their relatively low sensitivity [2]. It is difficult and ambiguous to determine the different types of animal glues and gelatine, and to differentiate egg yolk and egg white from whole egg [3, 4]. Rabbit glue, fish glue, hide and bone glue are also idistinguishable using the aforementioned methods because of their similar content of the individual amino acids [5]. However, several new methods for reliable differentiation of proteinaceous binders have appeared in the last few years, namely enzyme-linked immunosorbent assay (ELISA) [6] and immunofluorescence [7].
202
Anal Bioanal Chem (2007) 388:201–206
Table 1 Natural binders and basic types of painting techniques Binder based on
Examples of material
Basic painting technique (binder usage)
Proteins
Whole egg, egg yolk, egg white, bovine casein Hide, bone, rabbit and fish glue
Oil Polysaccharide Terpenoid and other compounds
Linseed oil, poppy and walnut oil Arabic gum, honey, starch Dammar, mastic, copal, colophony, shellac
Tempera, illumination, casein tempera Isolation, ground layer, Chinese and Japanese screen and scroll paintings Oil Gouache paint, watercolour Varnishes and minor components of other binders
Recently we started to identify the proteinaceous binders by specific enzymatic cleavage [1, 8–10] and subsequent analysis of the peptide mixture by matrix-assisted laser desorption/ionisation time of flight (MALDI-TOF) MS. The enzyme (usually trypsin) used cleaves the peptide bonds only between certain amino acids. Consequently, the resulting peptide mixture is much more characteristic for each protein than the ratio of the individual amino acids. The peptide fragments are then analysed by the highly sensitive MALDI-TOF MS [11]. The mass spectrum obtained creates “a mass fingerprint”, which enables a reliable identification by comparison of the spectrum of the real samples with that of the reference sample from a library [8]. This method also solves the outstanding problem of the identification of the mixtures of proteinaceous binders, which is typical for the other commonly used analytical methods. The identification of the proteinaceous binders would be more reliable if the peaks in the mass spectra of the real sample could be assigned to the amino acid sequences of the individual proteins. The analysis of a real sample originating from the painting “Sitting nude and grotesque masque” by Edvard Munch reported here follows the previous analysis of other paintings by Edvard Munch [9].
Experimental Reagents Trypsin (TPCK) from Promega Corporation, trifluoracetic acid and 2,5-dihydroxybenzoic acid (DHB) both from Sigma, acetonitrile (p.a.) and ammonium hydrogen carbonate from Lachema Brno were used. Reference protein mixtures Proteinaceous materials often applied in artworks were used as references: rabbit glue (Grac), bone glue (Deffner & Johann 2406), hide glue (Deffner & Johann 2400), gelatine (Grac), bovine casein (Deffner & Johann 2410), fish glue (Kremer 63100), rotter bolus (Kremer 40503), earth pigment, bole, whole egg and egg yolk (both homeprepared), bovine milk, whey, curd and linseed oil (Umton, Děčín, Czech Republic). Model samples The ten model samples which simulated the most typical ground layers were prepared according to recipes from the nineteenth century [12] at the Academy of Fine Arts in Prague. The model samples were mixed and dried in air for 2–3 months on glass plates. They were supplied to us as “samples of unknown composition” (Table 2).
Table 2 Composition of the fresh model samples of the typical ground layers (recipes from the nineteenth century) Component
Rotter bolus (g) Linseed oil (ml) Whole egg (ml) Egg yolk (ml) Egg white (ml) Hide glue (ml) Water (ml)
Sample S1
S2
S3
S4
S5
S6
S7
S8
S9
S10
6.70 2.0 7.0 – – – –
6.80
6.95 – 9.0
7.33 3.0 –
7.18
7.56 1.5 2.5 – – 5 –
6.92 1.5 – 2.5 – 5 –
6.68 –
7.89
6.80 2.0
– 9.0 – –
–
6
–
–
9.0 – –
4.0 – 5 –
– 4.0 5 –
7.0 – – 2
Anal Bioanal Chem (2007) 388:201–206
203
Real sample Edward Munch’s painting “Sitting nude and grotesque masque”, dated to 1893, provided by Munch-Museet in Oslo, was analysed by the methods described in this paper. The sample of the painting layer was available as fragments of a few micrograms in total.
standard nitrogen laser (337 nm) in reflector mode. A 2-μl aliquot of the peptide mixture was mixed with 4 μl DHB solution (15 mg DHB in 1 ml of a mixture of acetonitrile and 0.1% trifluoracetic acid (1:2 v/v). A 1.5-μl aliquot of the resulting mixture was spotted on the stainless steel MALDI target and dried in air. At least 200 laser shots were collected for each spectrum and the date were analysed using the XMASS software (Bruker).
Digestion of samples by trypsin The reference proteins as well as model and real samples were digested by sequencing-grade trypsin (TPCK). Approximately 1 mg of the model samples and a few micrograms of the real sample were digested in 20 μl of 50 mM ammonium hydrogen carbonate containing approximately 10 μg/ml trypsin at room temperature for 2 h. MALDI-TOF MS All mass spectra were acquired using a Bruker-Daltonics Biflex IV MALDI-TOF mass spectrometer equipped with
Results and discussion Database description In a standard proteomics procedure, a peptide mixture of an individual protein is obtained (most frequently by trypsin digestion), the molecular masses of the peptides are assessed by MS and the set of these masses serves as the input for a publicly available database (e.g. ExPASy [13]). The number of agreements of the peptide masses between the unknown sample and a known protein contained in the
Table 3 Analysis of the model samples Sample
Proteinaceous binders
Values of m/z found in the database
Reality
Our results and scores achieved (%)a
S1
Whole egg
Whole egg (55%)
S2
Egg yolk
Whole egg (42%)
S3 S4 S5 S6
Whole egg Egg white Whole egg Whole egg
Whole Whole Whole Whole
Hide glue
Animal glue (rabbit) (30%)
S7
Egg yolk
Whole egg (yolk) (40%)
S8
Hide glue Egg yolk
Animal glue (hide) (40%) Whole egg (22%)
Hide glue
Animal glue (hide, rabbit) (15%, 25%) Whole egg (39%) Animal glue (hide, rabbit) (29%, 26%) Whole egg (33%)
S9
Egg yolk Hide glue
S10
Egg yolk –
egg egg egg egg
(yolk) (30%) (white) (38%) (29%) (52%)
Maybe animal glue or gelatine (9%, 10%)
1,020, 1,048, 1,050, 1,064, 1,085, 1,106, 1,114, 1,150, 1,164, 1,226, 1,345, 1,401, 1,406, 1,428, 1,436, 1,445, 1,560, 1,562, 1,859 900, 910, 915, 931, 935, 973, 989, 999, 1,048, 1,085, 1,099, 1,106, 1,114, 1,134, 1,150, 1,164, 1,168, 1,324, 1,342, 1,375, 1,401, 1,406, 1,435, 1,436, 1,567, 1,608 915, 1,099, 1,234, 1,255, 1,268, 1,445 1,234, 1,345, 1,687, 1,859 903, 989, 1,020, 1,153, 1,234 1,020, 1,048, 1,085, 1,088, 1,105, 1,177, 1,201, 1,240, 1,267, 1,303, 1,401, 1,406, 1,445, 1,563, 1,566, 1,587, 1,656 942, 1,088, 1,095, 1,105, 1,201, 1,221, 1,265, 1,267, 1,303, 1,453, 1,465, 1,469, 1,501, 1,566, 1,568 1,020, 1,105, 1,177, 1,240, 1,267, 1,303, 1,401, 1,406, 1,453, 1,560, 1,563, 1,566, 1,656, 1,859 1,267, 1,303, 1,453, 1,465, 1,469, 1,501, 1,560, 1,566 1,020, 1,105, 1,177, 1,240, 1,255, 1,267, 1,303, 1,454, 1,533, 1,562, 1,563, 1,567, 1,586, 1,656, 1,795 1,095, 1,099, 1,105, 1,105, 1,205, 1,221, 1,267, 1,303, 1,453, 1,465, 1,501, 1,550, 1,562, 1,586, 1,637, 1,795, 1,947 956, 979, 1,013, 1,020, 1,105, 1,153, 1,240, 1,267, 1,303, 1,305, 1,440 942, 967, 1,105, 1,162, 1,221, 1,267, 1,303, 1,455, 1,469, 1,501, 1,562, 1,568 999, 1,046, 1,048, 1,085, 1,106, 1,114, 1,150, 1,164, 1,195, 1,201, 1,324, 1,374, 1,401, 1,406, 1418, 1437, 1445, 1560, 1562, 1607, 1859 1,099, 1,104, 1,106, 1,114, 1,195, 1,201, 1,406, 1,560, 1,562, 1,816, 1,947, 1,980
In parentheses the type of proteinaceous binder is specified according to our experimental results; it can be seen that this attempt did not always give the correct result. a The scores are the ratios of proteinaceous binder to all relevant peaks.
204
Anal Bioanal Chem (2007) 388:201–206
Fig. 1 Mass spectrum of the real sample “Sitting nude and grotesque masque”
database gives the “score” of the protein; the higher is the score, the more reliable is the assignation of the unknown protein. It should be emphasised that this procedure can serve for more or less pure protein only. That is why we built a similar database of proteinaceous binders that always contain a rich mixture of individual proteins. This database facilitates quick assignment of mass peaks of unknown sample to the proteinaceous binder. The individual proteins of each proteinaceous binder (egg yolk, egg white, whole egg, casein, bovine milk, whey, curd, gelatine, rabbit glue, fish glue, hide and bone
glue) were found in the Internet database ExPASy and then their cleavage were simulated by the program Mass-2.0alpha4 [14]. The peak list, resulting from tryptic cleavage of the unknown sample followed by MALDI-TOF MS, is transferred into the database. Subsequently, values of the mass peaks are compared with values in the table containing masses of all proteinaceous binders and the scores are obtained. The final decision on the presence of the binder depends on the score and sometimes, as in all proteomic experiments, also on the personal experience.
Table 4 The assignments of protein fragments obtained after enzymatic digestion from “Sitting nude and grotesque masque” M/z (Da)
Material
Protein
Sequence
Position
1,088.61
Yolk Egg white Yolk Egg white Egg white Egg white Yolk Yolk Yolk Yolk Yolk Yolk
Low-density lipoprotein receptor-related protein 1 Ovomucoid Vitellogenin I Ovalbumin-related protein X Ovomucin α-subunit Ovomucin α-subunit Vitellogenin I Vitellogenin I Vitellogenin II Low-density lipoprotein receptor-related protein 1 Low-density lipoprotein receptor-related protein 1 Vitellogenin II
DSKRGKIER VEQGASVDKR LVTFEDPER VKVYLPQMK HCKSAAPVPVR CMYDTCNAEK QFSSRPAYRR LTELLNSNVRLR ILGIDSMFKVANK KPEHELFLVYGK SDEKQSYCSSRKCK LSSKLEISGLPENAYLLK
675–683 137–146 1058–1066 123–131 2030–2040 620–629 355–364 831–842 1083–1095 523–534 2548–2561 54–71
1,105.62
1,177.62 1,267.21 1,427.76 1,435.61 1,459.64 1,648.74 1,975.74
Anal Bioanal Chem (2007) 388:201–206
Model samples The model samples, the compositions of which were unknown to us at the moment of analysis, were digested by the trypsin. The mass spectra of the peptide mixtures were obtained. The m/z values of well-developed peaks were used for analysis by the database described earlier. The ratio of the number of peaks, typical for a given proteinaceous binder, to the number of all peaks in the spectrum is called the “score” and is given for each model sample in Table 3. The general identification was successful with one exception, for sample S10, where the second proteinaceous binder was falsely identified. The individual animal glues cannot be distinguished for certainty, because of the similar amino acid sequences in the animal glues (samples S6, S8, S9). The remnant of egg white on the yolk surface that is always contaminated makes the identification of “pure” egg yolk always uncertain (samples S2, S5, S7, S10). Real sample After testing of the method developed on the model samples, the real sample, obtained from the Edvard Munch painting “Sitting nude and grotesque masque”, was analysed. The resulting mass spectrum is shown in Fig. 1 and the assignments of protein fragments are given in Table 4. In the 100-year-old real sample, whole egg was identified owing to the fact that nine peptide fragments of egg proteins were identified with certainty.
Conclusions The enzymatic digestion of proteinaceous binders contained in solid samples and their identification by MALDI-TOF MS was tested on model samples that were prepared according to historical recipes; they were analysed without prior knowledge of the composition. All proteinaceous binders (individual binders and their binary mixtures) were successfully identified, with one exception. In principle, it was not possible to distinguish between egg yolk and whole egg with the method used, most probably owing to the extremely high sensitivity of MALDI-TOF MS to the presence of trace amounts of egg white proteins. Further, the identification of different types of animal glue is not always reliable because of the similar composition and sequences of the collagens (the main proteins in animal glues) in different vertebrates. The method developed for identification of proteinaceous binders in artworks is highly promising for several reasons. First, it needs very small samples, generally smaller than other analytical methods; as a matter of fact, in some cases it can even be denoted as “nondestructive”
205
because trypsin solution (e.g. 6 μl) can be applied on the surface of the sample (e.g. cross-section) and then carefully removed without damage. We have also demonstrated that this technique can distinguish among the main classes of the binders, i.e. among egg, animal glues, milk products, etc. There are some practical advantages of the method, namely its relative speed (the results can be obtained in several hours) and low-cost (once the MS instrument has been acquired, no great expenses are necessary–EUR 50 per sample); of course, the method needs well-trained and experienced personal. The variability of the scores achieved depends mainly (1) on the number of individual proteins existing in the reference sample of the binder, i.e. a higher number of peaks in the trypsin digest gives more chance for good assignment; (2) on the number of peptide fragments, i.e. on the appropriate cleavage of the proteins; (3) on the posttranslational modifications of the proteins, e.g. bovine casein produces not easily ionisable phosphopeptides that are difficult to identify in the mass spectra; (4) on some other circumstances, e.g. the assignment of amino acid sequences to the individual peptides obtained from fish glue was not possible because no fish proteins were found in the databases available. The process of ageing may negatively influence the identification of the binder; the storage conditions (moisture, light exposure, etc.) may have an even greater effect. The unknown mass peaks are observed in every spectrum of the samples. These peaks could belong to peptides that are modified by ageing processes or interactions with inorganic materials. Our method using peptide mass fingerprinting for the identification of proteinaceous binders was successfully tested on the real sample. As our database is continuously being extended, we can offer its “service” to any laboratory that is interested in identification of proteins in artworks: if you send the masses of tryptic peptides as an Excel table by e-mail to Stepanka Kuckova, we shall try to help you free of charge.
Acknowledgements The work was supported by a grant from the Czech Ministry of Education (project number 6046137305). We thank restorer Biliana Topalova Casadiego (Munch-Museet, Oslo, Norway) and Karel Stretti from the Academy of Fine Arts for the providing real samples and restorer Hana Kurkova from the Academy of Fine Arts for the preparation of model samples and professional discussions.
References 1. Kuckova S, Nemec I, Hynek R, Hradilova J, Grygar T (2005) Anal Bioanal Chem 382:275–282 2. Colombini MP, Modungo F (2004) J Sep Sci 27:147–160
206 3. Carbini M, Stevanato R, Rovea M, Traldi P, Favretto D (1996) Rapid Commun Mass Spectrom 10:1240–1242 4. Chiavari G, Gandini N, Russo P, Fabbri D (1998) Chromatographia 47:420–426 5. Bonaduce I, Colombini MP (2003) Rapid Commun Mass Spectrom 17:2523–2527 6. Heginbotham A, Millay V, Quick M (2006) J Am Inst Conserv 45:89–105 7. Ramírez Barat B, de la Vina S (2001) Stud Conserv 46:282–288 8. Hynek R, Kuckova S, Hradilova J, Kodicek M (2004) Rapid Commun Mass Spectrom 18:1896–1900
Anal Bioanal Chem (2007) 388:201–206 9. Kuckova S, Hradilova J, Kodicek M, Nemec I, Grygar T, Bezdicka P, Hradil D (2005) In: Non-destructive testing and microanalysis for the diagnostics and conservation of the cultural and environmental heritage, Lecce, Italy 10. Tokarski C, Martin E, Rolando C, Cren-Olive C (2006) Anal Chem 78:1494–1502 11. Ma Y, Lu Y, Zeng H, Ron D, Mo W, Neubert TA (2001) Rapid Commun Mass Spectrom 15:1693–1700 12. Massey R (1967) Formulas for painters. Watson-Guptill, New York 13. http://www.expasy.ch. Cited 21 September 2006 14. http://mmass.biographics.cz/. Cited 24 April 2006
