Introduction

Paint layers in polychrome artefacts are typically composed of a heterogeneous mixture of organic (binding media, varnishes, colorants) and inorganic (pigment, thickeners, stabilizers, dryers, and extenders) materials [1]. Traditional binding media are based on egg, milk, casein, glues, and oils, used alone or in a mixture. The main glycerolipid materials encountered in works of art originate from egg yolk fats and various (e.g., linseed, walnut, poppy seed, and tung) drying oils characterized by a high content (>50%) of polyunsaturated fatty acids.

Identification of the binder constituents may provide information of paramount importance to both art historians and restorers, allowing a correct approach to restoration treatments. However, due to the low amount and complexity of the samples, binder analysis is still a challenging analytical task.

Thin-layer chromatography [2, 3] and high-performance liquid chromatography coupled with UV-visible detection [4] have been used to some extent; gas chromatography (GC) generally in combination with mass spectrometry (MS) and/or pyrolysis [5–9] remains, however, the most used technique. Identification/discrimination of the source (e.g., egg and/or drying oils) of glycerolipids is typically based on the ratio between the relative content of azelaic acid and palmitic acid (A/P), the ratio between the relative content of palmitic and stearic acids (P/S), and the sum of the percentage content of dicarboxylic acids as determined by GC-MS analysis of the extract containing the lipid-resinous fraction after a suitable derivatization step [10].

An alternative marker for egg-based binders could be represented by cholesterol that constitutes a major component of the lipid fraction of eggs; however, due to its fast rate of oxidation upon light ageing, cholesterol is readily converted into by-products. Indeed, Van den Brink et al. [11] have shown that some cholesterol oxidation products (such as 5,6-epoxycholestan-3-ol and 3-hydroxycholest-5-en-7-one) are better markers for egg tempera than cholesterol itself. Using direct temperature resolved mass spectrometry, the above-mentioned markers were discovered in paints on baroque altar pieces from the sixteenth and eighteenth century and in a twentieth century glaze on a mural painting.

Only recently, matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) and electrospray (ESI) MS-based techniques [12–23], traditionally used in proteomics, have been adapted for protein identification in paintings and archeological findings providing very encouraging results. Triacylglycerols in archeological samples (oil lamps) have been identified by nanoESI and Fourier-transform MS [24]. However, up to date, only few studies have been dedicated to matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) analysis of other organic materials present in paint samples such as glycerolipids and phospholipids [25–27], colorants [28], terpenoid resins [29], and polyethylene glycols [30]. In particular, MALDI Fourier-transform mass spectrometry has been used in an ageing study [27] of oxygenated triglycerides and phosphatidylcholines in egg tempera paint strips showing that the degree of oxidation could be related to light exposure time, presence of lead- and copper-containing pigments, and exposure to NOx and SO2. Traditional processing methods of drying oils (i.e., heating, addition of lead-based siccatives) have also been investigated by MALDI-MS and indicated uptake of oxygen and oligomerization of the starting triacylglycerols (TAGs) [26].

Surprisingly, in spite of its reputation as a well-established technique for polar lipid analysis, MALDI-MS has moderately been applied for fingerprinting of lipid binders in artworks. This paper demonstrates for the first time the potential of MALDI-MS for detection of glycerophospholipid and triacylglycerol by-products, originated from hydrolytic and oxidative scission processes, as markers for egg yolk and/or drying oils in aged paint samples. Extraction conditions from very small samples from painted artworks were optimized, and ageing processes of test specimens were studied. Markers permitting an unambiguous identification of egg, drying oil, and a mixture of both were identified and characterized.

The developed approach was successfully applied to two case studies: the Polyptych “Madonna with Child and St. Bernard, St. Nicholas, St. Vito, and St. John the Baptist” (1490) by Bartolomeo Vivarini (Santa Maria Assunta-Polignano a Mare-Italy) and a seventeenth century French canvas painting.

Experimental section

Materials

2,5-Dihydroxybenzoic acid (DHB) and all other chemicals used were obtained from Sigma-Aldrich (Sigma Aldrich, St. Louis, MO, USA). Lead white [2PbCO3·Pb(OH)2] and red ochre (anhydrous Fe2O3) pigments were purchased from Kremer Pigmente GmbH & Co. (Aichstatten, Germany).

Trilaurin, trimyristin, tripalmitin, and tristearin were obtained from Supelco. Water, acetonitrile, trifluoroacetic acid, methanol (MeOH), acetone, and chloroform were HPLC-grade and were used without further purification. Linseed oil used in standards preparation was purchased from Maimeri (Milan, Italy). Fresh hen eggs were purchased at local supermarkets.

Samples

Different oils and egg samples were used as reference binders. Test specimens were prepared using simple recipes. Paint replicas were made by applying different mixtures of binder and lead white pigment on glass slides. The binders were composed of egg yolk, whole egg, linseed oil, or a combination of egg and oil. The samples were analyzed either fresh or aged for 1 and 2 years in laboratory at ambient temperature under natural light and air exposure. One sample composed of egg yolk and red ochre had been aged for about 20 years within a box in the absence of light.

For the case study paintings, microsamples were collected by gently scraping off a single paint layer using a scalpel. Regarding the Vivarini Polyptych, a first sample was taken from the red “bole” (iron oxides/hydroxides and clay minerals) layer covered by the gold leaf constituting the background; a second sample was taken from the green mantle of St. John the Baptist where azurite (2CuCO3·Cu(OH)2) and lead tin yellow type I (Pb2SnO4) have been identified as the main pigments. A small fragment of the dark green background was sampled from the seventeenth century French painting containing raw umber (iron and manganese oxides) and green earth (glauconite and/or celadonite) pigments.

Lipid extraction

Different protocols (i.e., chloroform, Bligh Dyer, and Folch extractions) have been tested for lipid extractions from artificial samples and compared in terms of number and abundance of ions observed in the mass spectra. The Bligh-Dyer (BD) [31] method was selected as the most effective. Briefly, a fragment (50–100 μg) of either paint replica on glass slides (see Experimental section) or painted artwork was crushed using a pestle and a mortar into a fine powder. Then, 150 μL of CHCl3/MeOH (1:2) was added followed by vigorous vortex-mixing and ultrasonic bath (20 min). Then, 50 μL of CHCl3 followed by 50 μL of H2O were added; the solution was vortexed and ultra-sonicated again (20 min) after each addition. Finally, the solution was centrifuged (10 min at 3,000×g) and the lower organic layer was collected and dried under a stream of nitrogen. The residue was reconstituted in 100 μL of a CHCl3/MeOH (1:1) solution and mixed (1:1 v/v) with DHB matrix; 1 μL of the sample/matrix solution was spotted, allowed to dry and analyzed by MALDI-TOF-MS.

MALDI-TOF-MS

MS experiments were performed using a Micromass M@LDI™-LR time-of-flight mass spectrometer (Waters MS Technologies, Manchester, UK) equipped with a nitrogen UV laser (337 nm wavelength), a precision flat target plate sample introduction system bearing a micro-titer target plate (96-well, 3 mm diameter, 4.5 mm pitch, rectangular 12 by A-H with additional 24 near-point calibration wells), reflectron optics with effective path length of 2.3 m, a fast dual micro-channel plate (MCP) detector, and a high magnification (×70) camera system.

Positive ion spectra were acquired in reflectron mode in the mass range 400–1,200 Da. The following voltages were applied: pulse, 2,610 V; source, 15,000 V; reflectron, 2,000 V; MCP 1,900 V. The laser firing rate was 5 Hz, and, unless otherwise specified, 80 laser shots were used for each spot. The 80 resulting spectra were averaged, background-subtracted, and smoothed by a Savitzky–Golay algorithm.

A time lag focusing delay of 500 ns was used. Mass calibration was performed using a TAGs mixture composed of trilaurin, trimyristin, tripalmitin, and tristearin.

Results and discussion

In order to verify the efficiency of the BD method in the simultaneous extraction of TAGs and PLs, mixtures of egg and linseed oil were prepared, processed, and analyzed by MALDI-MS. It should be stressed that the BD method (normally employed for lipids extraction from a variety of biological samples) has never been tested on complex, and unusual matrices, such as painted artworks, where a mixture of inorganic (pigments, gypsum…often >80%), and organic components (lipids, proteins, terpenoids, synthetic resins…) are present.

Figure 1 reports the matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) spectra of lipid extracts from linseed oil (a), and from egg yolk (b) and egg yolk mixed with linseed oil (c). The spectrum in Fig. 1a reveals, as expected, the presence of Na+ and K+ adducts of the major TAGs (see Table 1) of linseed oil. The spectrum of Fig. 1b is clearly dominated by PLs signals (see the low m/z range of the spectrum), whereas signals occurring in the higher m/z region could originate from phosphatidylinositols (expected m/z range, 881–907) even if this attribution could not be unequivocal due to possible overlap of TAGs signals.

Fig. 1
figure 1

MALDI-TOF mass spectra obtained on: a linseed oil extracted with CHCl3; b egg yolk and c linseed oil/egg yolk mixture extracted using the BD method

Full size image
Table 1 Attribution of the main ions observed in the spectra of Fig. 1
Full size table

The results in Fig. 1c reveal that both polar lipids, in the mass range m/z 700–820, and neutral lipids (i.e., TAGs), in the mass range m/z 850–950, from egg and linseed oil could be detected. Unfortunately, the BD method does not provide an efficient extraction of cholesterol and (likely) of its oxidation by-products. Nonetheless, a cholesterol-related signal [M-H2O + H]+ was clearly detectable in the MALDI-MS spectra (data not shown), even if it falls in an m/z region with plenty of matrix-related signals; cholesterol oxidation products, with m/z <400, should be even more difficult to recognize. For these reasons, cholesterol and its oxidation products were not included among the egg markers potentially detectable by our approach.

Figure 2 shows the MALDI spectra of egg yolk/lead white pigment mixtures freshly prepared (a) and aged for 1 year (b) under light and ambient air exposure.

Fig. 2
figure 2

MALDI-TOF mass spectra of different egg yolk samples mixed with lead white: a un-aged; b aged for 1 year in presence of light and air. Insets in the figure show the zoom of three different mass regions of spectrum (b). M indicates matrix-related ions

Full size image

The addition of certain pigments to the binding medium is known to accelerate the conversion of egg lipids into oxidized products [27], the degree of oxygenation being (obviously) directly related to the light exposure time. Indeed, fresh samples in presence of lead white (Fig. 2a) show the native glycerophospholipids (and to a lesser extent TAGs) without any degradation products (see Table 1 for the attributions). After 1 year of ageing, the occurrence of new signals in the m/z ranges 630–730 and 881–920, mainly due to oxidation reactions, can be observed (Fig. 2b). Comparison of both spectra clearly shows that two major changes occur upon ageing, leading to the formation of oxidized PLs and TAGs (mass shift of +16 or +32 Da) and to by-products containing a shorter acyl fatty acid (mass shift to lower m/z) likely formed through a β-cleavage mechanism [32].

For instance, in the TAGs mass window (m/z 850–920) of Fig. 2b, the signal at m/z 897.72 (present in the fresh sample—see Fig. 2a—and corresponding to the sodiated adduct of palmitoyl-diolein TAG, [POO + Na]+), is practically absent, whereas Na+ adducts of its oxidation products are clearly distinguishable at m/z values 895.75 (mono-oxidized oleic acid residue–mass shift +16 Da) and 911.76 (hydroperoxidized oleic acid residue and/or two mono-oxidized oleic acid residues–mass shift +32 Da), respectively. This behavior can be generalized for all the detected TAGs.

As for phospholipids, palmitoyl-lineloyl-glycero-phosphatidylcholine (PLPC), for instance (protonated and sodiated adducts at m/z 758.53 and 780.54 in Fig. 2a), is nearly completely converted, after 1 year ageing, into mono-oxidized species (protonated and sodiated adducts at m/z 774.57 and 796.56 in Fig. 2b).

It is worth noting that typically, for both TAGs and PLs, the higher the unsaturation degree the higher was the extent of conversion to oxidized species; however, oxidized PLs showed typically only mono-oxygenated derivatives (mass shift of +16 Da), whereas oxidized TAGs showed mono- and di-oxygenated adducts (mass shifts of +16 and +32 Da).

As can be seen in Fig. 2b, additional ions in the range m/z 600–750 were observed in aged samples. These can probably be attributed to short-chain aldehydic or carboxylic oxidized phospholipids originating from β-cleavage oxidation processes [33, 34] of the unsaturated fatty acyl residues. PLPC (see above) may generate, during ageing, short-chain products due to β-scission of the linoleic acid residue (the saturated palmitic acid is not expected to undergo radical oxidation). Linoleic acid with the allylic hydrogen at C-11 is oxidized to allylic hydroperoxide concomitantly with a shift of the double bond producing conjugated and non-conjugated hydroperoxides. Four possible hydroperoxides could result at C-9 and C-13 (abundance of 32% and 34%, respectively) and C-10 and C-12 (each with an abundance of 17%) [35]. The hydroperoxides are then readily decomposed, through a β-scission breakdown mechanism, into short-chain phospholipid products bearing aldehydic and dicarboxylic functions (see Scheme 1). Taking into account these considerations, the predominant ion observed in the MALDI spectrum (Fig. 2b) at m/z 666.50 was attributed to the [M + H]+ of 1-palmitoyl-2-(nonanedioic acid)-phosphatidylcholine with the corresponding less abundant sodiated adduct at m/z 688.45. The intensity of this C9 dicarboxylic acid short-chain product suggests that the intermediate radical at C-9 is more stable than the intermediate radical species at C-13 which probably decomposes quite rapidly due to the presence of two unsaturations in the chain.

Scheme 1
scheme 1

Proposed formation of short-chain products from palmitoyl-lineloyl-glycero-phosphatidylcholine (PLPC) after oxidative β-cleavage. The m/z values are referred to [M + H]+ ions. R: palmitic acid residue

Full size image

Other short-chain by-products detected in the MS spectra can be attributed to protonated or sodiated ions generated from glycero-phospholipids and can be identified as oxidation products either with terminal aldehydic and dicarboxylic groups or keto/hydroxy moieties (see Table 2).

Table 2 Attribution of the main ions observed in the spectra of Fig. 2b
Full size table

After 2 years of ageing (data not shown), the native and oxidized PLs and TAGs have completely disappeared and only degradation products could be detected. Other reactions that can deplete TAGs and PLs are cross-linking reactions which originate high-mass dimers, trimers, and so on.

An egg yolk/red ochre sample aged for 20 years in absence of light was also analyzed. Figure 3a shows the MALDI-TOF mass spectrum obtained on a BD extract of this sample; Figs. 3b, c show expanded views in the low- and high-mass range of the same spectrum, respectively. Data in Fig. 3 suggest that the most intense signals fall in the range of the lysophosphatidylcholines (LPCs) arising from the hydrolysis of the major PLs present in egg yolk. In fact, ions at m/z = 496.24, 518.24, 524.24, and 546.24 can be attributed to H+ and Na+ adducts of LPC(16:0) and LPC(18:0), respectively. In the m/z range 660–700, short-chain products (see above discussion) were also observed. Ions at m/z 652.40, 666.44, 686.44, 688.44, 690.48, and 694.46 are likely generated from PLs scission (see, e.g., Scheme 1), whereas ions at m/z 591.40, 619.45, and 663.44 likely originated from TAGs scission. Surprisingly, intact PLs are still detectable at m/z 725.58, 734.62, 784.64, 756.62, and 762.62 (see Table 1 for assignments) together with H+ and Na+ adducts of LPC(16:0) dimers (at m/z 991.85 and 1013.88) and LPC(18:0) dimers at m/z 1019.90 and 1041.93, respectively. This result could suggest that paintings stored in appropriate conditions, such as reduced light exposure, can maintain for a long time a “memory” of the binding medium composition, particularly in the absence of pigments with a catalytic activity towards oxidation. Indeed, both light exposure and presence of catalytic Pb- or Cu-containing pigments (lead white, azurite…) have previously been shown to accelerate oxidation [26, 27].

Fig. 3
figure 3

a MALDI-TOF mass spectrum of the BD extract from egg yolk mixed with red ochre sample aged 20 years; b, c expanded views of spectrum A in the low- and high-mass regions, respectively

Full size image

In conclusion, all the above experiments indicate that, for binding medium based on egg yolk, the MALDI spectra of aged samples are dominated by “marker signals” (see Table 2) related to short-chain by-products derived from the oxidation of PLs.

In order to assess the usefulness of this methodology in distinguishing (on the basis of lipid profiles) between egg yolk and drying oil binders, fresh and aged linseed oil were also investigated. Figure 4 reports the MALDI spectrum of the BD extract from linseed oil aged for 1 year in presence of lead white.

Fig. 4
figure 4

MALDI-TOF mass spectrum of the BD extract from linseed oil aged for 1 year in presence of white lead. TAG-related marker ions of are labeled with an asterisk. M indicates matrix-related ions. Inset A: zoom of MALDI-TOF mass spectrum (m/z 600–850) of fresh linseed oil

Full size image

A comparison with Fig. 1a and inset in Fig. 4 (fresh sample) clearly indicates that the original TAGs signal clusters are considerably depleted; at the same time, new clusters, shifted by +16 and +32 Da, appeared as expected for an ageing process causing lipid oxidation [27]; mass shifts of +48 and +72 Da were also observed indicating a higher oxygenation rate, compared with egg samples due to the presence of TAGs with a higher unsaturation degree. Moreover, products deriving from the hydrolysis of the more saturated TAGs as well as by-products of β-scission are observed in the m/z range 600–650 and 650–850, respectively (see, for example, the β-scission of palmitoyl-diolein TAG in Scheme 2).

Scheme 2
scheme 2

Proposed formation of short-chain products from triacylglycerols after oxidative β-cleavage. The m/z values are referred to [M + Na]+ ion adducts. R 1 , R 2 : palmitic and oleic acid residue, respectively

Full size image

For instance, the most abundant ions at m/z 663.45 and 685.42 (Fig. 4) can be attributed to the protonated and sodiated adducts of the by-product generated from the oxidative degradation of triolein according to Scheme 3. These and other ions (in the m/z range 650–850), originating from hydrolytic and oxidative β-scission of the most abundant TAGs, represent possible markers of drying oil binders. It is of note that no PL-related by-products could be detected. Then, very interestingly, a MALDI-TOF spectrum could provide a fingerprint of the binding medium in a painted artwork as will be demonstrated in the following.

Scheme 3
scheme 3

Proposed formation of short-chain by-products from triolein after oxidative β-cleavage. The m/z value refers to the [M + H]+ ion observed in Fig. 6

Full size image

Case studies

Minute samples taken from two paintings dated to the fifteenth century (Polyptych of Bartolomeo Vivarini; Electronic Supplementary Material Figure S1) and to the seventeenth century (French canvas painting) were properly extracted and analyzed using the proposed MALDI approach. Figure 5 compares the MALDI spectra (zoom on the marker mass range) of the extracts of (a) “bole” sample and (b) green sample from the Vivarini painting. The presence of ions related to both PLs and TAGs by-products was observed (see Table 3 and Electronic Supplementary Material Figures S2 and S3 for probable attributions). In particular, in the spectra of Fig. 5, native TAGs are absent (data not shown), while the amounts of hydrolytic and β-scission products are quite high. Some interesting observations can be made for the two different samples shown in Fig. 5. In the spectrum of Fig. 5a, referred to as the “bole”-containing sample, it is worth noticing that a small amount of native PLs can still be detected (see ions at m/z 725.49, 760.54, and 788.59). This result resembles that obtained on the red ochre tempera sample aged for 20 years, corroborating the finding that in adequate storage conditions and in the absence of catalytic pigments intact PLs can be preserved to some extent. In fact, the “bole” paint layer is covered by a gold leaf protecting the inner layers from direct light. On the contrary, in the spectrum of the green azurite/lead tin yellow type I-containing sample (Fig. 5b), no native PLs could be detected. This is probably due to a catalytic effect of the copper- and lead-containing pigments on oxidative degradation of egg tempera paint, as previously observed [27]. On the other hand, the short-chain products are very similar to those of the “bole” Vivarini sample (Fig. 5a).

Fig. 5
figure 5

MALDI-TOF mass spectra (region covering the m/z range of the proposed markers) of the extracts of a “bole” sample and b green sample from the Vivarini painting (see Electronic Supplementary Material Figure S1). Inset C: zoom of MALDI-TOF mass spectrum (m/z 880–1000) of “bole” sample. TAG- and PL-derived marker ions are labeled with asterisks and circles, respectively

Full size image
Table 3 Proposed elemental composition of the main marker ions observed in the spectra of Fig. 5
Full size table

In conclusion, the paint of the Vivarini Polyptych shows numerous marker signals of egg binder. Moreover, several ions, characteristic of TAGs degradation products, of comparable intensity with respect to the PLs by-products can be observed. This finding suggests the use of a mixture of egg and drying oil (tempera grassa) as paint medium. Furthermore, the TAGs high-mass region shows the presence of multiple oxidized products (data not shown) as observed for drying oil aged samples.

Pyrolysis-gas chromatography-mass spectrometry analysis of a Vivarini paint sample yielded an A/P ratio of 0.31, indicating the use of an egg/oil mixture as the binding medium [10]. This finding is corroborated by historical data on painting techniques indicating that, during the fifteenth century, drying oils started to be used in combination with egg binders, as confirmed by several analyses performed on paintings of that period [36].

Figure 6 shows the MALDI spectrum of a sample taken from the seventeenth century French canvas painting. As can be seen, short-chain products at m/z values 663.39, 685.35, and 701.38 are observed, suggesting the presence of drying oil as binder. Very low signals are detected at m/z ion 907.83, probably attributed to triolein and at m/z ion 855.78 attributed to 1-oleyl-2,3-dipalmitoyl TAG. As mentioned before, a certain residual degree of unsaturation is not completely unexpected since protective layers may preserve the paint from complete hydrolysis and further oxidation reactions [37]. Interestingly, the mass spectrum is dominated by the marker ions at m/z 663.39 and m/z 685.35. The complete absence of ions at m/z 640.45, 668.44, and 696.46 (see inset) permits to exclude the presence of egg and strongly corroborates the conclusion that this painting is based on a drying oil binder.

Fig. 6
figure 6

MALDI-TOF mass spectrum of a sample taken from the seventeenth century canvas painting. TAG-derived marker ions are labeled with asterisks. Inset: zoom of the m/z 696

Full size image

Conclusions

The capabilities of MALDI-TOF mass spectrometry for lipids fingerprinting in extracts of minute samples taken from painted artworks of different origins and historical periods has been demonstrated for the first time. The presence of specific markers, originated from hydrolytic and oxidative scission processes involving triacylglycerols and phospholipids, permits an unambiguous discrimination between egg- and drying oil-based binding media and assessment of their simultaneous presence as well.

At present, this information is generally obtained by the use of widely accepted indexes, such as the azelaic/palmitic acid ratio, whose determination by GC-MS, however, requires a much more complex sample pre-treatment and a derivatization procedure.

Interestingly, according to our protocol, the sample residue remaining after lipid extraction contains the protein fraction which can also be analysed by MALDI-TOF-MS. This implies that a single small sample can be simultaneously characterized from both a lipidomics and a proteomics point of view. Work in this direction is in progress.