A Mass Spectrometry-Based Approach for Characterization of Red, Blue, and Purple Natural Dyes

Effective analytical approaches for the identification of natural dyes in historical textiles are mainly based on high-performance liquid chromatography coupled with spectrophotometric detection and tandem mass spectrometric detection with electrospray ionization (HPLC-UV-Vis-ESI MS/MS). Due to the wide variety of dyes, the developed method should include an adequate number of reference color compounds, but not all of them are commercially available. Thus, the present study was focused on extending of the universal analytical HPLC-UV-Vis-ESI MS/MS approach to commercially unavailable markers of red, purple, and blue dyes. In the present study, HPLC-UV-Vis-ESI MS/MS was used to characterize the colorants in ten natural dyes (American cochineal, brazilwood, indigo, kermes, lac dye, logwood, madder, orchil, Polish cochineal, and sandalwood) and, hence, to extend the analytical method for the identification of natural dyes used in historical objects to new compounds. Dye markers were identified mostly on the basis of triple quadrupole MS/MS spectra. In consequence, the HPLC-UV-Vis-ESI MS/MS method with dynamic multiple reaction monitoring (dMRM) was extended to the next 49 commercially unavailable colorants (anthraquinones and flavonoids) in negative ion mode and to 11 (indigoids and orceins) in positive ion mode. These include protosappanin B, protosappanin E, erythrolaccin, deoxyerythrolaccin, nordamnacanthal, lucidin, santalin A, santalin B, santarubin A, and many others. Moreover, high-resolution QToF MS data led to the establishment of the complex fragmentation pathways of α-, β-, and γ- aminoorceins, hydroxyorceins, and aminoorceinimines extracted from wool dyed with Roccella tinctoria DC. The developed approach has been tested in the identification of natural dyes used in 223 red, purple, and blue fibers from 15th- to 17th-century silk textiles. These European and Near Eastern textiles have been used in vestments from the collections of twenty Krakow churches.

The best separations to date have been achieved using reversed phase columns, mostly C18. However, since colorants have extended systems of conjugated double bonds, it seems reasonable to Table 1. High-performance liquid chromatography coupled with spectrophotometric detection and tandem mass spectrometric detection with electrospray ionization (HPLC-UV-Vis-ESI MS/MS) characterization of color compounds.

Dyes
Colorants extracted from indigo as well as from wool fibers dyed with nine other dyes were identified using an ESI MS/MS detector preceded by HPLC with a phenyl column. Full scan analysis and the subsequent MS/MS fragmentations of the predominant quasi-molecular ions were used to obtain information about the molecular weights of the colorants and for the structural evaluation of the sugar moieties, aglycones, and unglycosylated compounds.

Indigo
Chromatograms acquired for a DMSO extract of indigo showed two main peaks that corresponded to indigotin (65) ([M + H] + at m/z 263) and indirubin (67) ([M + H] + at m/z 263). They were observed with a spectrophotometer at 280, 550, and 600 nm and with an MS detector in positive ion mode. In addition, two small peaks were found at t R 17.2 and 18.4 min (Figure 1a). The MS investigation of these two compounds suggested that they might be isomers. Firstly, the even m/z values of both their [M + H] + ions were 262, which indicated an odd number of nitrogen atoms in their molecules. Secondly, their MS/MS spectra were very similar to each other although not identical (Supplementary Materials, Figure S1). Apart from the ions at m/z 245, 235, 219, and 190 present in both mass spectra, the first compound (t R 17.2 min) showed an intense signal at m/z 120, whereas the second one (t R 18.4 min) showed such at m/z 131. These MS/MS spectra almost completely coincided with those acquired for indigotin and indirubin, especially for m/z values above 150, indicating similarities between their structures too. Looking at the data taken together, the chemical formula of both compounds was defined as C 16 H 11 N 3 O; however, their molecular structures could not be determined. Nevertheless, it seemed that these compounds (called indigoid compound A (31) and B (37)) would not be crucial for identifying indigo in historical objects. , madder (f), brazilwood (g), logwood (h), sandalwood (i), and orchil (j) extracts acquired by UV-Vis detector; peak numbers are decoded in Table 1 (22), and dc4 (25) (Figure 1b) as well as pp6 (17), pp7 (20), and , madder (f), brazilwood (g), logwood (h), sandalwood (i), and orchil (j) extracts acquired by UV-Vis detector; peak numbers are decoded in Table 1. Colorants present in American and Polish cochineals were the subject of earlier detailed studies [24,25]. On this basis, apart from carminic and kermesic acids (52), flavokermesic acid (51), dcII (18), dcIV (30), dcVII (36), dcOfka (23), dc3 (22), and dc4 (25) (Figure 1b) as well as pp6 (17), pp7 (20), and deoxyerythrolaccin (57) (Figure 1c) were included in the presented method, as they were previously recommended as American and Polish cochineal markers, respectively.
Chromatographic and spectrometric data acquired for the lac dye extract proved the presence of several laccacic and xantholaccaic acids [39] (Figure 1d). Since they are animal-origin oxidized derivatives of kermesic and flavokermesic acids substituted at the C-7 position by a large functional group (Supplementary Materials, Figure S2)-that is, by N-acetyltyramine (in laccaic acid A and xantholaccaic acid A), tyrosal (laccaic acid B and xantholaccaic acid B), tyrosine (laccaic acid C), or tyramine (laccaic acid E)-their [M − H] − ions showed almost identical fragmentation pathways. The two most intense signals in each MS/MS spectrum (Supplementary Materials, Figure S3) corresponded to the loss of one or two CO 2 molecules from carboxyl groups, whereas the next one was a result of the further loss of a H 2 O molecule.
Apart from laccaic and xantholaccaic acids, the lac dye extract also contained kermesic acid (52)

Madder
Chromatograms of the madder extract acquired by the spectrophotometric detector showed several peaks. The most intense peak belonged to alizarin (55), but purpurin (64) and rubiadin (66) were also found ( Figure 1f). These colorants were identified by comparison with their standards, while others required the analysis of MS/MS spectra.
A compound eluted just before alizarin and observed in the chromatograms acquired at 280 and 400 nm was identified to be lucidin (

Madder
Chromatograms of the madder extract acquired by the spectrophotometric detector showed several peaks. The most intense peak belonged to alizarin (55), but purpurin (64) and rubiadin (66) were also found ( Figure 1f). These colorants were identified by comparison with their standards, while others required the analysis of MS/MS spectra.
A compound eluted just before alizarin and observed in the chromatograms acquired at 280 and 400 nm was identified to be lucidin Apart from that, all the other signals in both spectra were the same; the ions at m/z 211, 195, and 167 were formed by the further loss of CO, CO 2 , or both of these molecules together, respectively. The MS/MS spectra of xanthopurpurin and nordamnacanthal were identical to those ones found in the literature [40,41]. Similar congruence was found for the absorbance spectra of other identified compounds.
Another peak was observed at 32.1 min in the chromatograms acquired with spectrophotometric detection at visible range. This compound (coded as rt1) has not been reported to date. Simple MS/MS spectra of its [M − H] − ion at m/z 491 included only two significant signals (Supplementary Materials, Figure S4), the first one at m/z 239 and the second one at m/z 251. It led to the assumption that the colorant was an anthraquinone dimer, probably composed of one alizarin molecule and one lucidin molecule (76). Apart from the anthraquinones described above, some of their glycosidic derivatives were found, that is, lucidin O-primeveroside  Figure S5).

Brazilwood
According to the literature, brazilwood contained mainly brazilin (6) and brazilein (3). Peaks of these two compounds were found at t R 7.1 and 5.5 min, respectively, in the chromatogram of the brazilwood extract (Figure 1g). Their identification was based on MS/MS spectra.
The fragmentation of brazilin-type neoflavonoids has been discussed in only one publication to date [42], but the considerations have been devoted to one particular fusion, and they have not included the complete pathway, especially for brazilin. Nevertheless, the presented data were helpful in the identification of both neoflavonoids.
The fragmentation of brazilin and brazilein proceeded according to two mechanisms, that is, the loss of small neutral molecules or the cleavage of internal rings (their MS/MS spectra and the proposed directions of their fragmentation are shown in Figure 3a ) were rather low. On this basis, it can be assumed that compounds classified as brazilin-type homoisoflavonoids with a fused five-membered D-ring [43] decompose mostly via this D-ring, whereas their oxidized forms, with an extra unsaturated bond in a D-ring that prevent their fission, fragment via a heterocyclic C-ring.

Logwood
Chromatograms acquired using spectrophotometric detection for the logwood extract showed several not-very-intense peaks at shorter retention times ( Figure 1h). Two of them were identified to be hematein (1) and hematoxylin (2), a brazilin-type homoisoflavonoid [43]. Their MS/MS spectra (Supplementary Materials, Figure S7) showed that the decomposition of hematoxylin occurred via a D-ring, whereas hematein fragmented by the fission of a heterocyclic C-ring. Proposed fragmentation pathways are shown in Supplementary Materials, Figure S8.
The next four compounds observed in the chromatogram were found to be potential logwood formed by inner ring cleavage. The fragmentation of prottosappanin E, which is a combination of brazilin and protosappanin B molecules, was even simpler, since it resulted in only one signal at m/z 283 ( Figure 3d). Probably, it was formed by the decomposition of bonds between both moieties.
Moreover, two brazilin-like compounds (11,28) were also found (t  Figure S6) were like those of brazilin, but their structures remain unknown. Probably, they were brazilin isomers, or they belonged to homoisoflavans.
The compound eluted at 13.9 min was identified to be urolithin C (21) Figure 3e). Usually, it has been referred to as compound-type C, and its identity has been determined recently based on LC-ESI MS/MS, GC-MS, and NMR studies [36]. Since the structure of urolithin C is stabilized by resonance, the initiation of its fission required higher collision energies, and the fragmentation mostly led to the detachment of small molecules, such as OH (m/z 226), CO (m/z 215), CH 2 O (m/z 213), CO 2 (m/z 199), and double CO (m/z 187).

Logwood
Chromatograms acquired using spectrophotometric detection for the logwood extract showed several not-very-intense peaks at shorter retention times ( Figure 1h). Two of them were identified to be hematein (1) and hematoxylin (2), a brazilin-type homoisoflavonoid [43]. Their MS/MS spectra (Supplementary Materials, Figure S7) showed that the decomposition of hematoxylin occurred via a D-ring, whereas hematein fragmented by the fission of a heterocyclic C-ring. Proposed fragmentation pathways are shown in Supplementary Materials, Figure S8.
The next four compounds observed in the chromatogram were found to be potential logwood markers (Supplementary Materials, Figure S9). The first two of them, coded as hc1 (4) and hc2 (10) (7) and hc4 (15), respectively. The higher m/z values of their quasi-molecular ions, low fragmentation, and generated product ions suggested these compounds could be a combined structure of two colorants. Therefore, since the product ions acquired for hc3 corresponded to the quasi-molecular and product ions of hematoxylin, hc3 (7) was considered to be a hematoxylin dimer.

Sandalwood
Chromatograms acquired by a UV-Vis detector at 400 or 500 nm for the sandalwood extract showed four intense peaks (Figure 1i). According to the literature [44], it should contain santalins and santarubins, which was confirmed by MS/MS data. The first compound (t R 27.

Orchil
The chromatogram acquired in positive full scan mode by the ESI MS detector for the extract of wool dyed with orchil (Roccella tinctoria DC.) showed three significant as well as six minor peaks. According to the available literature data, these compounds corresponded to aminoorceins. To confirm the identity, the MS/MS spectra were acquired.
The most intense peak present in the chromatogram (Figure 1j) at 22.7 min was identified to be α-aminoorcein (49) ([M + H] + at m/z 363). On the basis of the literature and by analogy with α-aminoorsein, it was assumed that the other two minor peaks corresponded to α-aminoorceinimine (45) ([M + H] + at m/z 362) and α-hydroxyorcein (58) ([M + H] + at m/z 364). Product ion spectra acquired by a triple-quadrupole mass spectrometer for the precursor ions of all three α-orceins indicated far-reaching similarity. Nonetheless, the identity of the MS/MS signals and determination of the fragmentation paths were very difficult, as most neutral losses could not be positively determined without high-resolution data due to the variety of possible isobaric fragments. For example, the loss of 17 Da could correspond to OH or NH 3 , whereas the loss of 29 Da could correlate with the detachment of CHO, C 2 H 5 , CH 2 N, or CH 2 NH. Since these losses are indistinguishable by triple quadrupole MS, the orchil extract was next examined using quadrupole-time-of-flight tandem mass spectrometry (QToF MS). High-resolution and high-accuracy product ions are presented in Table 2.    (45), and α-hydroxyorcein (58), respectively. Since these colorants are stabilized by their resonance structures, clear and legible MS/MS spectra were acquired only using higher collision energy (CE) values, such as 30-45 V (Figure 4), and the fragmentations started from the detachments of small radicals. Although thermodynamic arguments preclude the possibility of the loss of radicals from even-electron ions, the high-resolution mass spectra of orceins contradict this generally known theory. The most intense signals (at m/z 348.1106, 347.1264, and 349.0946 for α-aminoorcein, α-aminoorceinimine, and α-hydroxyorcein, respectively) corresponded to the loss of CH 3 . Except that, the [M + H−OH] + ions were also present (m/z 346.1298, 345.1461, and 347.1152), but their origin was probably twofold. On the one hand, the hydroxyl radical could be detached from one of the two hydroxyl substituents of a phenyl ring ( Figure 5). The same loss might also be achieved for α-hydroxyorcein by the homolytic cleavage of the hydroxyl group at the C-7 position of phenoxazin-3-one (m/z 347.1152). Since the other two compounds, α-aminoorcein and α-aminoorceinimine, are substituted at the C-7 position by an amino group, the detachment of the C-7 substituent led to the loss of an aminyl radical (NH 2 ) and the formation of the m/z 347.1145 and 346.1318 ions, respectively. Even though radical loss from even-electron ions is rather unusual, this phenomenon has been already observed for prodiginines used as inks [45].
The fragmentation path also included the loss of small neutral fragments, such as H 2 O and CH 4 . It was observed mainly in the spectra of α-hydroxyorcein (m/z 348.0867) though, but the signals were less intense than those corresponding to the loss of radicals. The elimination of methane probably occurred between two methyl groups at the C-6 and C-9 positions, leading to the formation of the inner cyclopentadiene ring between the phenyl substituent and phenoxazin-3-one structure. A similar mechanism was responsible for the elimination of H 2 O from α-aminoorcein and α-hydroxyorcein (m/z 345.1232 and 346.1079, respectively) as well as of NH 3 from α-aminoorceinimine (m/z 345.1230). These losses occurred between the C-2-hydroxyl group of the phenyl moiety and the C-7 substituent of the phenoxazine skeleton. Moreover, due to the presence of a carbonyl group at the C-3 position of α-aminoorcein and α-hydroxyorcein, one of the possible fragmentation paths also led to the detachment of the CO molecule (the m/z 335.1400 and 336.1232 ions, respectively), which was not observed for α-aminoorceinimine ( Figure 5).   The next fragmentation stage was the further loss of the same small molecules (neutral and radical) from the primary product ions. The MS/MS spectra also showed the elimination of CHO or CH 2 N within the hydroxyl substituents or the amino group at the C-7 position of phenoxazine.
Apart from the loss of small neutrals or small radicals, the fragmentation of α-orceins also occurred with the detachment of larger fragments. One of them, the 2,4-dihydroxy-6-methylphenyl radical (C 7 H 7 O 2 ), was created via a homolytic cleavage of the C-C bond between the phenyl ring and phenoxazine skeleton. The signals corresponding to this loss were observed in the spectra of α-aminoorcein, α-aminoorceinimine, and α-hydroxyorcein at m/z 240.0892, 239.1055, and 241.0736, respectively. Other losses were a result of a cross-ring fission. The primary loss of CH 3 from the 2,4-dihydroxy-6-methylphenyl substituent probably triggered the ring fission and detachment of the C 4 H 3 O 2 radical that was followed by the furan or pyrrole ring formation. Moreover, analogous ion structures were created by the elimination of the C 4 H 4 O 2 molecules from the phenyl substituent, but these signals were present only in the MS/MS spectra of α-aminoorceinimine (m/z 278.1283) and α-aminoorcein (m/z 279.1128).
The next fragmentation path included the fission of the phenoxazine system and detachment of the structure between atoms 3 and 5. Since α-aminoorcein and α-hydroxyorcein are substituted at the C-3 position by a carbonyl group-and α-aminoorceinimine, by a primary ketimine group-this cleavage resulted in the loss of the C 3 H 2 O 2 or C 3 H 3 NO molecules, respectively. In the case of α-hydroxyorcein, the [M + H−C 3 H 2 O 2 ] + ion (m/z 294.1126) was very intense, hence its further fragmentation and subsequent detachment of the CO molecule (m/z 266.1173) and CH 3 radical (m/z 279.0892). Moreover, the alternative fission of the phenoxazine-3-one skeleton also led to the loss of the C 3 H 3 O 2 fragment from quasi-molecular ions of α-aminoorcein and α-hydroxyorcein (giving the m/z 292.1209 and 293.1049 ions, respectively).
The peaks of three βand three γ-orceins were also observed in the chromatogram, but since there were no significant differences between the MS/MS spectra of their βand γ-isomers, distinction between their two forms was not possible. Moreover, peaks that corresponded to βand γ-hydroxyorcein showed very low intensities.

Protocol for Analyzing Historical Samples
The analytical protocol for the identification of natural dyes in historic textiles using HPLC coupled with UV-Vis and MS detections was proposed herein. This approach considered sample

Protocol for Analyzing Historical Samples
The analytical protocol for the identification of natural dyes in historic textiles using HPLC coupled with UV-Vis and MS detections was proposed herein. This approach considered sample color (and thus also the extraction method), the UV-Vis and MS data acquired for the colorants (natural standards and markers in dyed fibers), and our expertise. Thus, according to the protocol (Figure 7), the detection of the colorants in the methanolic extracts from fibers should be conducted as follows: (I) yellow, orange, and black samples should be examined at 280 and 400 nm, in negative ion MS mode; (II) brown, blue, and green samples, at 280, 400, 550, and 600 nm, in both positive and negative ion MS modes; and (III) red and purple samples, at 280, 480, 550, and 600 nm, in both positive and negative ion MS modes; moreover, the DMSO extracts from (IV) brown, blue, green, and purple fibers should be analyzed at 550 and 600 nm, in positive ion MS mode. The protocol, recommended herein, was applied to identify the natural dyes used in historical samples, as described below.

Analysis of Historical Samples
The developed HPLC-UV-Vis-ESI MS/MS method was used to analyze 223 blue, purple, and red fibers taken from silk textiles dated from the 15th to 17th century and used in the vestments from the collections of twenty Krakow churches. The DMSO and methanol-water-formic acid extracts were analyzed using positive or positive and negative ion modes, respectively. The acquired results led to the identification of the natural dyes in the examined fibers, even though some of them were re-dyed with synthetic dyes. All the samples, identified colorants, and dyes are listed in Supplementary Materials, Table S1.
All the twenty-nine blue threads included in the set were dyed with indigo. Moreover, this dye was also identified in the next thirty-two samples of other colors (in fifteen at the trace level). Its use was proved by the presence of indigotin (65) in the extracts, always accompanied by isatin (13), a photodegradation product of indigotin. Indirubin was found only in some of these extracts. Since the composition of indigo colorants depends on the fermentation process for the indigo precursors, not on the origin of the plants used for this fermentation, indigo provenance could not be determined. Thus, the indigo could have been made from European or Asian plants from the genera Indigofera, Isatis, or others.
In 15th and 16th centuries, indigo was probably produced from woad (Isatis tinctoria L.), a native European plant that has been used on the Old Continent since antiquity. Although indigo from Indigofer tinctoria L. was imported to Europe after the discovery of the sea route to India, its use was much less likely at that time, since this dye was banned due to the allegedly "devilish origin" [46]. However, at the end of the 16th century, more and more Asian indigo came to Europe. Initially, it was used in combination with woad, but later on, during the 17th century, it replaced indigo from woad almost completely [44,46]. In consequence, threads taken from the younger textiles could be dyed using both Indigofera and Isatis species.
Although indigo was identified as a main dye in twenty-nine blue samples, it was used for individual dyeing in sixteen of them (in four of them, traces of other dyes were found as well). In another thirteen fibers, indigo was combined with other dyes and the fibers still remained blue. These dyes included American cochineal (one sample), weld (two samples), dyer′s broom (one sample), and, above all, orchil (ten samples, in two of them, together with wild madder (Figure 8a)).

Analysis of Historical Samples
The developed HPLC-UV-Vis-ESI MS/MS method was used to analyze 223 blue, purple, and red fibers taken from silk textiles dated from the 15th to 17th century and used in the vestments from the collections of twenty Krakow churches. The DMSO and methanol-water-formic acid extracts were analyzed using positive or positive and negative ion modes, respectively. The acquired results led to the identification of the natural dyes in the examined fibers, even though some of them were re-dyed with synthetic dyes. All the samples, identified colorants, and dyes are listed in Supplementary Materials, Table S1.
All the twenty-nine blue threads included in the set were dyed with indigo. Moreover, this dye was also identified in the next thirty-two samples of other colors (in fifteen at the trace level). Its use was proved by the presence of indigotin (65) in the extracts, always accompanied by isatin (13), a photodegradation product of indigotin. Indirubin was found only in some of these extracts. Since the composition of indigo colorants depends on the fermentation process for the indigo precursors, not on the origin of the plants used for this fermentation, indigo provenance could not be determined. Thus, the indigo could have been made from European or Asian plants from the genera Indigofera, Isatis, or others.
In 15th and 16th centuries, indigo was probably produced from woad (Isatis tinctoria L.), a native European plant that has been used on the Old Continent since antiquity. Although indigo from Indigofer tinctoria L. was imported to Europe after the discovery of the sea route to India, its use was much less likely at that time, since this dye was banned due to the allegedly "devilish origin" [46]. However, at the end of the 16th century, more and more Asian indigo came to Europe. Initially, it was used in combination with woad, but later on, during the 17th century, it replaced indigo from woad almost completely [44,46]. In consequence, threads taken from the younger textiles could be dyed using both Indigofera and Isatis species.
Although indigo was identified as a main dye in twenty-nine blue samples, it was used for individual dyeing in sixteen of them (in four of them, traces of other dyes were found as well). In another thirteen fibers, indigo was combined with other dyes and the fibers still remained blue. These dyes included American cochineal (one sample), weld (two samples), dyer s broom (one sample), and, above all, orchil (ten samples, in two of them, together with wild madder (Figure 8a)).
Orchil, as a red-purple direct dye produced from lichenized fungi in the genus Roccella, has been known in Europe since ancient times, but knowledge about its use was lost with the fall of the Western Roman Empire. The dye returned as a textile dye at the end of the 13th century and the beginning of the 14th century. Although its use was restricted in France in the second half of the 17th century due to its poor light resistance, it was still used in other European countries [46]. In consequence, orchil was identified in thirty-three blue, purple, and red fibers taken from 15th-to 17th-century textiles; this dye was always used in a mixture with other dyes, never individually.  Table 1.
Seven out of eight purple samples were dyed either with a combination of orchil and indigo or with a ternary mixture of orchil, indigo, and American cochineal. In one purple sample, orchil was not present at all. A mixture of indigo and American cochineal was used to achieve the intended color instead.
Kermes was identified in seven red threads, always together with orchil. All those samples were taken from 15th-to 16th-century European textiles. Polish (Porphyrophora polonica L.) or Armenian cochineal (Porphyrophora hamelii L.) were found in the next six samples dated to the same period, but the unequivocal determination of the dye was impossible, since both scale insects belong to the same genus and their compositions are very similar to each other. However, the largest group of samples was dyed with American cochineal (most likely Dactylopius coccus Costa), which after arriving to Europe in 1523 [44] quickly displaced from the market other red dyes of animal origin such as kermes and Polish and Armenian cochineals.
American cochineal was identified by the presence of carminic acid (19) together with minor colorants, such as dcII (18), dcIV (30), dcVII (36), carminic acid derivative (24), kermesic acid (52), and flavokermesic acid (51). Although similar compounds were also found for the samples dyed with Porphyrophora species, they stood out with clearly higher contents of flavokermesic acid (51) and kermesic acid (52) as well as a complete absence of dcII (18) and trace presence of pp12 (41), pp14 (53), and pp15 (60) instead. In consequence, American cochineal was found as the only dye in one hundred and one thread samples, whereas in the next eleven threads, it was used together with  Table 1.
Orchil, as a red-purple direct dye produced from lichenized fungi in the genus Roccella, has been known in Europe since ancient times, but knowledge about its use was lost with the fall of the Western Roman Empire. The dye returned as a textile dye at the end of the 13th century and the beginning of the 14th century. Although its use was restricted in France in the second half of the 17th century due to its poor light resistance, it was still used in other European countries [46]. In consequence, orchil was identified in thirty-three blue, purple, and red fibers taken from 15th-to 17th-century textiles; this dye was always used in a mixture with other dyes, never individually.
Seven out of eight purple samples were dyed either with a combination of orchil and indigo or with a ternary mixture of orchil, indigo, and American cochineal. In one purple sample, orchil was not present at all. A mixture of indigo and American cochineal was used to achieve the intended color instead.
Kermes was identified in seven red threads, always together with orchil. All those samples were taken from 15th-to 16th-century European textiles. Polish (Porphyrophora polonica L.) or Armenian cochineal (Porphyrophora hamelii L.) were found in the next six samples dated to the same period, but the unequivocal determination of the dye was impossible, since both scale insects belong to the same genus and their compositions are very similar to each other. However, the largest group of samples was dyed with American cochineal (most likely Dactylopius coccus Costa), which after arriving to Europe in 1523 [44] quickly displaced from the market other red dyes of animal origin such as kermes and Polish and Armenian cochineals.
American cochineal was identified by the presence of carminic acid (19) together with minor colorants, such as dcII (18), dcIV (30), dcVII (36), carminic acid derivative (24), kermesic acid (52), and flavokermesic acid (51). Although similar compounds were also found for the samples dyed with Porphyrophora species, they stood out with clearly higher contents of flavokermesic acid (51) and kermesic acid (52) as well as a complete absence of dcII (18) and trace presence of pp12 (41), pp14 (53), and pp15 (60) instead. In consequence, American cochineal was found as the only dye in one hundred and one thread samples, whereas in the next eleven threads, it was used together with annatto, brazilwood, dyer s broom, and weld in single samples, as well as with indigo or its combination with orchil, as has already been mentioned. Nevertheless, most often, American cochineal was combined together with an unknown ellagitannin dye used as an organic mordant. These two dyes were identified together in eighteen samples.
The last identified animal-origin dye was lac dye (Kerria lacca Kerr or K. chinensis Mahdihassan). This Indian dye had been occasionally used in medieval Europe, but it gained popularity only in the second half of the 18th century [46]. Lac dye had been more widely known, however, in Muslim countries, including the Ottoman Empire, where it had been used to obtain crimson red. It was confirmed by the results, as lac dye was found in eight thread samples taken from 15th-century European textiles and in four fibers from 17th-century Turkish textiles.
The Rubiaceae dyes were identified in only nine samples (mostly dated to the 17th century); nevertheless, the different compositions of the anthraquinones in the extracts led to the distinguishing of three different chromatographic profiles. The first one, with high signals of alizarin (55), purpurin (64), and nordamnacanthal (72), corresponded to dyer s madder (Rubia tinctorum L.). It was observed only in two extracts, which is understandable considering that, although madder as a European plant was widely cultivated on the continent, it was mainly used for dyeing wool, not silk. A similar chromatographic profile but without the peak of nordamnacanthal indicated the use of the Galium species, which was probably used in one sample. The high signal of rubiadin (66) together with the almost complete absence of alizarin corresponded to wild madder (Rubia peregrina L.). This dye was mostly used together with orchil and indigo to produce purple shades (Figure 8a).
Red color was obtained not only using anthraquinone dyes but also with flavonoid dyes. Thus, eleven red threads were dyed with brazilwood, which was used both individually as well as in combination with American cochineal (Figure 8b) or annatto (results published in [37,38]). Moreover, brazilwood was also found in nine samples of other colors (mostly yellow and orange). The origin of dyes, however, could not be determined precisely, since the same colorants were obtained from different tree species. One of them, sappanwood (Caesalpinia sappan L.), from southern Asia, was already known and used in medieval Europe, wherein the inner part of its trunks was used to produce brazilwood. In later centuries, the dye was also obtained from species imported from South America, such as Caesalpinia echinata L. and Haematoxylum brasiletto Karst. Furthermore, these trees were used for dyeing not only red but also yellow and orange.
Standard solutions of most colorants (0.2 mg·mL −1 ) were prepared in methanol. Only indigotin and indirubin (0.1 mg·mL −1 ) were dissolved in dimethyl sulfoxide (DMSO). Woolen yarns were mordanted, dyed, and extracted according to the procedure indicated previously [37]. Moreover, a 0.25 mg indigo sample was dissolved in 25 mL of DMSO. The solutions were kept in an ultrasonic bath for 5 min, and the obtained solutions was diluted 10 times with methanol.
The 223 silk fibers (red, purple, and blue) were taken from silk textiles that were dated to the 15th to 17th century and used in the vestments belonging to the collections of twenty Krakow churches (all the samples are listed in Supplementary Material, Table S1).
The purple and blue fibers were extracted twice, using two extraction methods consecutively, the first one with dimethylsulfoxide (DMSO), and the second one with acidic-methanol extractant. The red fibers were extracted only with the second procedure. The extraction procedures have been described in detail by Lech [37].
Chromatographic analyses were performed using a 1220 Infinity II LC System (Agilent Technologies, Waldbronn, Germany) with a Zorbax SB-Phenyl column (4.6 × 150 mm, 3.5 µm, 80 Å, Agilent Technologies); a Zorbax SB-Phenyl precolumn (4.6 × 12.5 mm, 5.0 µm, Agilent Technologies); and a mixture of methanol, water, and formic acid as a mobile phase. Detection was carried out with a 1220 Compact Variable Wavelength Detector and a 1200 Variable Wavelength Detector (Agilent Technologies, Waldbronn, Germany), as well as with a 6460 Triple Quad LC/MS system with JetStream Technology (Agilent Technologies, Waldbronn, Germany). The full scan chromatograms and spectra were acquired for m/z 100-1000. Quasi-molecular ions of the colorants were fragmented using 15, 25, 35, and 45 V of collision energy (CE). The MS/MS spectra were acquired from m/z 50 to the m/z-value of the precursor ion + 20 to achieve an upper limit of around 20 m/z above the m/z of each fragmented ion. The parameters of the method were described in detail by Lech [37]. The final method was developed in dynamic multiple reaction monitoring (dMRM) mode given the most intense precursor and product ion pairs (transitions) of the identified dye markers. The optimal collision energy for each transition was selected manually. Detailed settings (the retention times, fragmentor values for each precursor ion, MRM transitions, and CEs of the new colorants) are provided in Table 1.
The analyses were performed using the MassHunter Workstation software (Agilent Technologies, USA).

Conclusions
A phenyl HPLC column and UV-Vis-ESI MS/MS technique were used to separate and characterize colorants (flavonoids, homoisoflavonoids, anthraquinones, indigoids, and orceins) present in the extracts of ten natural dyes (American cochineal, brazilwood, indigo, kermes, lac dye, logwood, madder, orchil, sandalwood, and Polish cochineal). Tandem mass spectrometric detection provided information on the structures of unknown colorants eluted from the HPLC column. Several colorants were identified in this way, including protosappanin B, protosappanin E, santalin A, santalin B, santarubin A, nordamnacanthal, lucidin, erythrolaccin, and deoxyerythrolaccin, but the structures of some compounds (from brazilwood, logwood, madder, and sandalwood) are still pending. Moreover, complex fragmentation pathways of α-, βand γaminoorceins, hydroxyorceins, and aminoorceinimines extracted from orchil-dyed wool have been defined the first time according to our knowledge on the basis of high-resolution mass spectrometry data acquired by QToF MS. The results have shown that the fragmentation is twofold. It occurs by the loss of small neutrals and radicals, as well as by the loss or fission of the aromatic rings.
MS/MS data have been used not only to identify new colorants but also to expand the existing dMRM method with 60 new dye markers. It has resulted in the development, as far as we know, of the first universal and comprehensive approach, that includes 176 colorants, intended for the identification of natural dyes in historical objects. Furthermore, a general analytical protocol has been developed for the identification of the natural dyes used in historical objects, antiques, and works of art. It involves both extraction and analysis steps (including UV-Vis detection wavelengths and MS ionization modes) and also considers fiber colors and the physicochemical properties of presumed dyes. This approach has been used to analyze 223 red, purple, and blue fibers taken from the silk textiles used in the vestments belonging to the collections of twenty Krakow churches. It has led to the identification of several dyes, such as orchil, brazilwood, madder, wild madder, indigo, lac dye, kermes, and different species of cochineals. The results of this study have completed the picture of natural dyes used in the most valuable textiles of European and Near Eastern origin dated to the 15th to 17th century.
Supplementary Materials: Figure S1: Indigo: MS/MS spectra of indigotin, indirubin, indigo compound A, and indigo compound A acquired in positive ion mode. Figure S2: Chemical structures of laccaic acids. Figure S3: Lac dye: MS/MS spectra of laccaic acid E, laccaic acid C, xantholaccaic acid B, laccaic acid B, xantholaccaic acid A, and laccaic acid A, acquired in negative ion mode. Figure S4: Madder: MS/MS spectra of rt1 (alizarin-lucidin dimer). Figure S5: Madder: MS/MS spectra of rubiadin O-primeveroside, ruberthyric acid, and lucidin O-primeveroside acquired in negative ion mode. Figure S6: Brazilwood: MS/MS spectra of brazilin-like compounds in negative ion mode. Figure S7: Logwood: MS/MS spectra of hematoxylin and hematein acquired in negative ion mode. Figure S8: Proposed fragmentation pathways for a) hematoxylin and b) hematein. Figure S9: Logwood: MS//MS spectra of hc1, hc2, hc3, and hc4 acquired in negative ion mode. Figure S10: Sandalwood: MS/MS spectra of santalin-like compound acquired in negative ion mode. Table S1. Compounds and dyes identified in silk textiles dated to the 15th to 17th century.