Introduction

Anthraquinoid dyes are aromatic compounds derived from 9,10-anthraquinone and they have been the most common organic colorants in art history applied to obtain red hues ranging from orange to pink shades. Their easy availability in nature and great resistance to photo-oxidation are plausible explanations for their employment since ancient times [1]. Owing to the solubility of these organic compounds in water and in other binding media, they have to be either precipitated as metal complexes or adsorbed on an inert substrate for painting purposes. The resulting lake is insoluble in water media and can be used in the most common paint binders [2].

Over the centuries, artists have applied lakes with different painting techniques and on several supports: on paper as inks for illuminated manuscripts [3], on walls for mural paintings [4], on wood panels or canvas for glazes [5], and even for decorating ceramic objects [6]. The glazing technique, consisting in the overlapping of colored paint layers on already dried paint layers, was exploited by several European painters from the XIV to the XIX century in order to increase the chromatic range of their palettes [7]. Thus, lakes, characterized by a low coloring power, were applied with a large amount of oil or egg as binder over inorganic pigment layers to achieve specific nuances and transparent colorations [8]. The most important red anthraquinoid lakes used for painting purposes were: madder lake (traditionally extracted from roots of different species of Rubiaceae and considered to be the most beautiful and stable among all lakes) [8], carmine (unspecific name often used for cochineal lake and extracted from different species of Coccus insects since prehistory) [9], kermes (extracted from eggs of a species of Coccus insect called Kermes vermilio Planchon, it has similar chemical proprieties and coloration but different composition to carmine) [10], and Indian lac (obtained from stick-lac, a resinous substance produced by Kerria Lacca Kerr insect, and spread in the East since 250 BC) [11].

The characterization of organic dyestuffs and lake pigments is of paramount importance from the artistic and archeological point of view, as it may permit dating or locating a work of art, to determine the artists’ technique, to evaluate technological skills of civilizations, and to establish ancient trading and commercial routes. The study of the interactions of dyes with the other materials present in the matrix and of the degradation phenomena due to aging is fundamental to prevent or slow down deterioration processes and to choose the most suitable procedures and materials for restoration campaigns [1–7, 12, 13].

From an analytical point of view, identifying lakes in paint samples is particularly challenging because of their instability related to low light-fastness and great sensitivity to atmospheric agents and to pH variations, if compared to most inorganic pigments [14]. Moreover, the analyses are further complicated by the high amount of binding media in which they are usually dispersed, the simultaneous presence of several organic materials and non-original compounds as a consequence of aging and environmental contamination, the low percentage of dyes used in traditional lakes (1–3% w/w), and most of all the difficult extraction of dyes from the matrices [5]. Thus, any analytical procedure aimed at detecting anthraquinoid dyes in very low concentration has to be optimized and the possible interference of the inert support of the lake or of the other painting materials (pigments, fillers and binders) on the analysis of the lakes should be investigated.

Micro-destructive analytical techniques such as high performance liquid chromatography coupled with diode array or mass spectrometric detectors (HPLC-DAD-MS) are the most suitable set-ups for dyes characterization because of their high sensitivity and selectivity, good chromatographic efficiency, and possibility to achieve quantitative detection [1]. Nevertheless, liquid extraction and separation of the chromophore-containing molecules from the matrix must be achieved before their analysis. Extraction of the compounds of interest from the matrix is an important step in every analytical procedure, and the selection of the most appropriate method is particularly challenging if we have to deal with unique micro-samples containing tiny amounts of colorants [15]. Among the extraction methods optimized and described in the literature, the key factors are to avoid oxidation phenomena, conversion of glycosides into aglycones, esterification of carboxylic compounds, and to achieve a high yield for all the molecular markers of the colorants present [16]. Until now, the analytical techniques and procedures developed for the analyses of dyes have been mainly focused on textiles [17, 18], whereas only few efficient protocols for the analyses of lakes in painting micro-samples have been described. In this work, we applied one of the so-called mild methods optimized by Sanyova [19] based on hydrofluoric acid (HF) and, for the first time, we coupled it with MALDI-MS detection for the analysis of painting samples.

While laser desorption/ionization mass spectrometry (LDI-MS) has already proven to be a promising analytical tool for the study of colorants [20–23], matrix assisted laser desorption/ionization mass spectrometry (MALDI-MS) has been mainly applied for the analysis of large molecules such as proteins, glycerolipids, and synthetic polymers. Nevertheless, some MALDI-MS methodologies have been recently developed in order to analyze smaller molecules such as organic dyes [20–26]. The mild extraction method based on the use of HF is particularly suitable for unstable colorants, and it has already been exploited for the investigation by chromatographic techniques of a wide range of colorant classes and mixtures applied on different supports, in various matrices or complexed with different types of metal cations such as Ba2+, Sn2+, Fe3+, Cr3+, Cu2+, Pb2+, Zn2+, and Ca2+ [19, 27].

MALDI-ToF-MS analyses of standard anthraquinones and derivatives are described in the literature [28, 29] but the anthraquinones typical of red lakes have never been investigated by this means, and no study on the possible interferences from the matrix is available. In particular, the possible influence of the complex painting matrix on anthraquinone identification in unknown samples has not been evaluated yet.

In this work, standard anthraquinones and reference anthraquinoid lakes and paint model samples were analyzed by a MALDI-ToF-MS method (entailing the extraction with HF and purification steps [19]) performed both in positive and in negative acquisition modes. The collected spectra are discussed in relation to the different ionization modes and the best one for each anthraquinoid dye has been assessed. Interestingly, the generation of singly charged double cation adducts has been observed in positive mode in agreement with the study of Lou et al. [28], in which the mechanism of this phenomenon has been thoroughly investigated. The analyses of freshly prepared paint model systems allowed us to evaluate the role played by the presence and the kind of paint medium on anthraquinones identification.

The MALDI-ToF-MS method employed in this study proved to be efficient for anthraquinoid lakes identification because the rapid spectra acquisition and high sensitivity made it possible to obtain analytical information from microscopic paint samples, avoiding possible interferences from the inorganic support or the binding media with a minimal sample treatment. The method was successfully applied to identify the kind of lake used in three archeological painted samples, confirming the results already obtained with different analytical techniques.

Experimental

Reagents and Materials

In order to prepare the solutions used for dyes extraction, clean-up, and spotting on the steel plate, the following reagents and solvents were used: hydrofluoric acid (HF, 38–40%; Merck, Darmstadt, Germany), formic acid (FA; Sigma Aldrich, USA), methanol (MeOH; LC/MS grade Sigma, Aldrich, USA), acetonitrile (ACN; Sigma Aldrich, USA), trifluoroacetic acid (TFA; Sigma Aldrich, USA), 2,5-dihydroxybenzoic acid (DHB; Sigma Aldrich, USA). The water used for the analyses was purified in a Milli-Q Direct Water system (Merck Millipore, USA). The clean-up and pre-concentration steps were performed using a reverse phase C18 sorbent in a microcolumn Zip Tip (EMD, Millipore, USA). For the preparation of paint model systems, bi-distilled water (Carlo Erba, Milan, Italy) and ammonia (NH3; Merck, Darmstadt, Germany) were used.

Standards and Reference Materials

The standard dyes were: alizarin (97% purity) and purpurin (80.9% purity) purchased from Fluka (USA) and carminic acid (90% purity) and laccaic acids (mixture of laccaic acid A and B) purchased from Sigma Aldrich (Milan, Italy). The following reference materials were used: madder lake (from Rubia tinctorum, 220 E, Zecchi, Florence, Italy), carmine (alum lake of carminic acid- Cochineal, C-1022, Sigma, USA) and Indian lac (from Coccus lacca, Zecchi, Florence, Italy). The reference materials used as binders for paint model systems preparation were: linseed oil (650 Maimeri, Milan, Italy), Arabic gum (Sigma, USA), casein (caseina lattice extra, ABT 53171 BP05A, Brescia, Italy), albumin from chicken lyophilized egg white (Sigma-Aldrich, USA), egg (bought in an ordinary supermarket), and rabbit glue (LeFranc, Le Mans, France).

Paint Model Systems

Forty paint model systems were prepared using three different lakes (madder lake, carmine, Indian lac) and six different binders (casein, rabbit glue, albumin, egg, Arabic gum, linseed oil). The binder/lake mixture was applied on glass slides (26 × 76 mm, 1 mm; Carlo Erba, Milan, Italy). The binder solutions used were: casein (0.1% w/w), albumin (1% w/w and 5% w/w), rabbit glue (0.1% w/w and 1% w/w), and Arabic gum (1% w/w and 10% w/w). Each solution was prepared in bi-distilled water, except for casein prepared in an ammonia solution (0.1% w/w). Whole egg was used with possible addition of water whereas linseed oil was undiluted. Each lake was dispersed in each binder in different ratios in order to obtain different shades of color. A list of the 40 analyzed paint model systems is reported in Table 1 with their relative composition and sample weight.

Table 1 List of Reference Lakes and Paint Model Systems Analyzed with Their Relative Composition and Sample Weight
Full size table

Archeological Samples

The samples analyzed were collected from different archaeological sites and historical periods. The description, provenience, dating, previous results collected with other analytical techniques for comparison purposes, as well as the sample weight for MALDI analysis are provided in Table 2.

Table 2 Name, Description, Provenience, Dating, Weight, and Previous Results Collected with Other Analytical Techniques in Archeological Samples
Full size table

Analytical Procedures

The method adopted for the analysis of standards, reference dyes, and lakes, paint model systems and archeological samples was the so called HF method [19]: a solution of 10 μL of ACN and 10 μL of MeOH was added to approximately 0.4 mg of each sample and sonicated for 20 min. The extraction of the dyes was carried out adding 20 μL of 4 M HF and the extract was left at room temperature overnight. The elimination of ionic and polar substances and of the excess of fluoride was performed with a reverse phase C18 sorbent in micro-columns using: ACN/MeOH (1:1) solution and H2O for the preconditioning, 0.01% TFA water solution after sample loading, and ACN/MeOH (1:1) + TFA 0.01% solution for the elution of dyes. 0.5 μL of eluted sample were mixed with 0.5 μL of DHB matrix (consisting of a solution of 8 mg of DHB in 150 μL of ACN added to 250 μL of 0.2% TFA and 100 μL H2O) and spotted on the steel plate, left to dry, and analyzed by MALDI-ToF-MS.

LDI-MS analysis was also performed in order to compare the relative abundance of the ions in the isotopic clusters of anthraquinones to those acquired by MALDI. For LDI-MS, less than 0.5 mg of each sample was dispersed in 1 mL of bi-distilled water; 1 μL of the resulting solution was spotted on the steel plate, left to dry, and then analyzed by LDI-ToF-MS.

MALDI-ToF-MS

A MALDI-TOF-TOF Autoflex Speed (Bruker) mass spectrometer instrument with a standard nitrogen laser (337 nm) in positive and negative reflector mode was used. Before each measurement, the instrument was externally calibrated. A minimum of 1500 laser shots was collected for each spectrum and laser power was adjusted each time. The acquisition range was set between 0 and 1000 Da (m/z) with an accuracy of 0.3 Da. The resulting spectra were processed with XMASS software, mMass (Bruker, Karlsruhe, Germany).

The instrumental calibration for the analysis of chromophore-containing molecules was performed on the known masses of DHB matrix, particularly suitable for organic compounds [31]. The positive mode is the most commonly used for spectra acquisition in MALDI because the protonation generally creates mono-charged species and the mass spectra obtained are characterized by a higher reproducibility in comparison to negative mode ones [32, 33].

Results and Discussion

First we will present the results for standard anthraquinoid dyes discussing the positive and negative modes separately. Secondly, the application to reference lakes, paint model systems, and archaeological samples will be shown in order to confirm the already observed behavior of the dyes both in positive and in negative modes.

Analytical blanks obtained with the adopted analytical procedure and acquired in positive and negative mode allowed us to establish which m/z peaks are not diagnostic for the samples analyzed (see Supplementary Figure S1a and b, respectively, and Supplementary Table S1).

Standard Anthraquinones

As reported in the introduction, alizarin and purpurin, carminic acid, laccaic acid A and B are the main chromophore-containing molecules of madder lake, carmine, and Indian lac, respectively, and they are considered to be their molecular markers [34].

The identification of all the molecular markers was successful with both analysis modes except for carminic acid, which ionizes in negative mode only. The differences and the potentialities of each ionization mode will be discussed.

Positive Mode

The peaks identified in the MALDI mass spectra of all the analyzed dyes (Figure 1a, Supplementary Figures S2a and S3a) are summarized in Table 3.

Figure 1
figure 1

(a) MALDI-ToF mass spectrum of alizarin; (b) isotopic cluster of alizarin (zoom of MALDI-ToF mass spectrum); (c) isotopic cluster of alizarin (LDI-ToF mass spectrum); (d) isotopic cluster of alizarin (simulated mass spectrum) acquired in positive mode. The peaks ascribable to analytical blank are indicated by the dots

Full size image
Table 3 List of Ions Identified in MALDI Mass Spectra of the Standard Dyes Analyzed in Positive Mode
Full size table

It is well known that the principal ions produced in MALDI are generated by proton transfers from matrix to the analyte ([M + H]+), reactions between cationized molecules and ubiquitous Na+ and K+ impurities ([M + Na]+, [M + K]+), ionization of the matrix followed by electron transfer from analyte to matrix radical cation [M]• + [21]. However, the formation of abnormal singly charged double cation adductions ([M + 2H]+, [M + 2Na]+) observed in our spectra for alizarin, purpurin, and laccaic acid A is particulary interesting. This phenomenon has already been observed in a previous study carried out on the MALDI-MS analysis of anthraquinone derivatives used as potential precursors for urea-anthraquinone-based foldamers [28]. It is important to highlight that Lou et al. [28] worked on pure, standard compounds; thus the behavior of anthraquinones in complex samples has not been assessed yet. The mechanism of the formation of [M + 2H]+ is unlikely ascribable to a reduction of the compound becuse the antharquinones have a very stable conjugated backbone, while it is more probable that the compound takes up first one electron upon laser irradiation and then two protons realeased by the photon-excited matrix or protonated ions. According to the literature [28], this unusual behavior occurs in these compounds because of the electron-deficient nature of anthraquinones, which enhances their electron capture ability. Accordingto Lou et al. [28], the kind of matrix used does not influence the phenomenon as long as it is a protic one. The formation of sodium or potassium ion single/double cation adducts is not surprising, as [Na]•+ and [K]•+ peaks are detectable in the analytical blank (see Supplementary Table S1).

We compared the relative abundance of the ions of the isotopic cluster of alizarin, purpurin, and laccaic acids in our MALDI spectra with LDI mass spectra and with simulated ones (Figure 1b–d, Supplementary Figures S2 b–d and S3 b–d) in order to highlight the differences and confirm that the formation of singly charged double cation adduction is a property of the MALDI process. The hypothesis that these compounds ionize both as single/double protonated and as radical ions explains the peculiar isotopic pattern in MALDI mass spectra as the result of the contribution of all the ions generated.

Negative Mode

The mass spectra of alizarin, purpurin, and laccaic acids collected in negative mode are characterized by the presence of the radical ion of molecular markers and the absence of sodium or potassium ion adducts (Figures 2a, 3a, and Supplementary Figures S4a, S5a, and S6a). Peaks coming from the matrix ionization are more intense than in positive mode. These results are again in accordance with Lou et al. [28]. Conversely, carminic acid, which ionizes in negative mode only, shows a MALDI spectrum in which [M – H]− (m/z = 491) is the base peak (Figure 3).

Figure 2
figure 2

(a) MALDI-ToF mass spectrum of alizarin; (b) isotopic cluster of alizarin (zoom of MALDI-ToF mass spectrum); (c) isotopic cluster of alizarin (LDI-ToF mass spectrum ); (d) isotopic cluster of alizarin (simulated mass spectrum) acquired in negative mode. The peaks ascribable to analytical blank are indicated by the dots

Full size image
Figure 3
figure 3

(a) MALDI-ToF mass spectrum of carminic acid; (b) isotopic cluster of carminic acid (zoom of MALDI-ToF mass spectrum); (c) isotopic cluster of carminic acid (simulated mass spectrum) acquired in negative mode. The peaks ascribable to analytical blank are indicated by the dots

Full size image

The isotopic clusters of alizarin (Figure 2b), purpurin, and laccaic acid A (see Supplementary Figures S4b, S5b, and S6b) show the expected pattern in which [M]• − is the most intense peak.

Lakes and Paint Model Systems

Reference lakes and all the paint model systems were analyzed in both modes, except for carmine and samples prepared with this lake, which were only analyzed in negative mode.

It is important to note that madder lake contains several minor components having the same molecular formula as alizarin or purpurin. Thus, the identified peaks in the mass spectra of madder containing samples might refer to alizarin and purpurin or their isomers. In the following discussion, we will mention alizarin and purpurin only, keeping in mind that the presence of other isobaric anthraquinones cannot be excluded.

The MALDI-ToF mass spectra of all the reference lakes (see Figure 4, and Supplementary Figures S6, S7, and S8a, b) and all paint model samples analyzed did not exhibit any substantial difference from those acquired for the respective molecular markers and the same considerations apply (see section related to Standard Anthraquinones). The MALDI mass spectra acquired in positive mode are still characterized by the ions reported in Table 2. In addition, [M + 2Na]+ ions were detected both in madder lake and in Indian lac (Figure 4).

Figure 4
figure 4

MALDI-ToF mass spectrum acquired in positive mode for madder lake. The peaks ascribable to analytical blank are indicated by the dots

Full size image

It is worth noticing that the mass spectra obtained for the paint model systems, prepared using madder lake, carmine, and Indian lac admixed with six different binders (casein, rabbit glue, albumin, egg, Arabic gum, linseed oil) in different lake to binder ratios, show no peculiar differences with respect to the reference lakes. In Figure 5a, b, and c, the MALDI-ToF mass spectra of A1ML2 (1% w/w albumin and madder lake), C01C (0.1% w/w casein and carmine), and C01IL (0.1% w/w casein and Indian lac) are shown as representative of the spectra obtained. The kind of binding medium used and the lake to binder ratio do not influence either the identification of the m/z peaks corresponding to the molecular markers of the dyes or their intensity. Therefore, the binding medium seems not to interfere in the detection of the lake by our procedure.

Figure 5
figure 5

Comparison of MALDI-ToF mass spectra of (a) A1ML2 (1% w/w albumin and madder lake) and madder lake acquired in positive mode; (b) C01C (0.1% w/w casein and carmine) and carmine acquired in negative mode; (c) C01IL (0.1% w/w casein and Indian lac) and Indian lac acquired in positive mode. The peaks ascribable to analytical blank are indicated by the dots

Full size image

Summarizing, the analytical procedure chosen allowed us to characterize both lakes and paint model systems. On the basis of S/N ratios of the marker compounds and of the other adducts identified, we assessed positive mode as the ionization method of choice to identify madder lake and Indian lac, whereas negative mode is indicated for carmine (in view of the impossibility to ionize carminic acid in positive mode).

Archeological Samples

Three archeological painting samples (“Delos,” “Kor 26,” and “Palmetta”) were analyzed using both positive and negative modes.

In positive mode (Figure 6a) all the samples show a peak corresponding to [M + 2H]+ (m/z = 258) and another to [M + Na]+ (m/z = 279) in accordance with double cation adduction ions found for standard purpurin. In “Delos” and “Kor 26” a low intensity peak [M + 2H]+ (m/z = 242) due to alizarin is also present.

Figure 6
figure 6

MALDI-ToF mass spectra of samples of “Delos,” “Kor 26,” and “Palmetta” compared with madder lake, acquired in (a) positive mode; (b) negative mode

Full size image

In the mass spectra of all the samples acquired in negative mode (Figure 6b) no peaks corresponding to carmine were detected, thus ruling out its presence in the samples. Interestingly, the peak corresponding to [M]• − (m/z= 256) of purpurin is clearly visible, whereas that of alizarin is not detectable. This is due to the low alizarin content of the samples, as demonstrated by previous analyses [30, a].

Even if positive mode confirms to be the method of choice for madder lake detection, as deduced from the analyses of the reference materials discussed in the previous sections, negative mode provides useful data for a more reliable assignment of the pink hue being due to madder lake in “Delos,” “Kor 26,” and “Palmetta.”

The MALDI-ToF-MS results are consistent with some previous studies [30, a] in which alizarin and purpurin have been already detected after micro-destructive, time consuming chromatographic procedures (Table 2).

Conclusions

The analytical approach tested in this study allows for identifying dyes and lakes in all the samples listed in Tables 1 and 2 with at least one of the ionization modes.

The HF method, applied in the literature to HPLC analyses for lake characterization, was coupled with MALDI detection for the first time. The extraction with HF proved reliable for anthraquinone characterization, by releasing the dye so that the molecular markers can be ionized by MALDI-ToF-MS.

The best identification is achieved in positive mode for the molecular markers of madder lake and Indian lac, whereas in negative mode for carmine. Nevertheless, performing both analysis modes by MALDI-ToF on the HF purified sample does not require any additional effort in terms of time and sample consumption, while it can provide complementary information useful for a more reliable identification of the lake used. Most importantly, results have proved that the presence and the nature of the binder do not influence the identification of organic lakes in reference samples. The formation of abnormal singly charged double cation adductions has already been studied relating to standard anthraquinones derivatives ionized in positive mode [28]. Differently, in our work the same phenomenon was observed for alizarin, purpurin, and laccaic acid A both in standard dyes and in reference lakes and paint model systems, revealing that the painting matrix does not interfere in this peculiar MALDI process.

The methods selected were applied for the analysis of organic materials in archeological samples revealing the presence of madder lake in all of them.

MALDI-ToF-MS shows some limitations for dye characterization, such us the impossibility to distinguish isomers in comparison with chromatographic techniques. Nevertheless, the minimal and simple sample preparation required, the rapidity of spectrum acquisition, and the small amount of sample needed make MALDI-ToF-MS a truly attractive method for lake identification. Moreover, the potential of this methodology is increased by the possibility to re-analyze the same sample at any time (after it has been spotted on the steel plate). In addition, the agreement between MALDI-ToF-MS and chromatographic techniques in identifying madder lake in the two wall painting samples and on a clay vessel sample analyzed validates the potentialities of our analytical approach also in the case of archaeological samples.

Although further analyses have to be carried out in order to improve the database of reference materials and to increase the number of case studies, including also artificially aged model paint systems, the results of this paper are a good starting point for the development of a micro-destructive analytical protocol that maximizes the information obtained while minimizing the amount of sample needed.