Molecular species fingerprinting and quantitative analysis of saffron (Crocus sativus L.) for quality control by MALDI mass spectrometry

Herein we describe a rapid, simple, and reliable method for the quantitative analysis and molecular species fingerprinting of saffron (Crocus sativus L.) by direct MS and MS/MS analysis. Experimentally, powdered saffron was subjected to a brief treatment with a 0.3% TFA water/acetonitrile solution, and the resulting mixture was directly placed on the MALDI plate for analysis. This approach allowed the detection of the commonly observed crocins C-1–C-6 and flavonols, together with the identification of the unknown highly glycosylated crocins C-7, C-8 and C-9, and carotenoid-derived metabolites. The strategy endorsed the simultaneous detection and characterization of saffron and adulterant markers using crude extracts of the adulterant itself and synthetic sets of adulterated authentic saffron samples. The implementation of the strategy was to measure the amount of an unknown adulterant from the crude extract using curcumin as a non-isotopic isobaric internal standard. The relationship between the saffron and curcumin molar ratios were established with a correlation coefficient of 0.9942. The ANOVA regression model was significant, F(1, 72) = 13 595.82, p < 0.001, y = (0.0116 ± 0.0001)x + (−0.1214 ± 0.0086). No matrix effects were observed and good results were obtained with respect to instrumental repeatability (*RSD% < 2%) and LOD (1.1%). The analysis of commercial samples of saffron using the proposed approach showed the suitability of the method for routine analysis (minimal sample preparation and very short measuring time per sample).


Introduction
Saffron, which is the red dried stigmas of Crocus sativus L., is a member of the Iridaceae family. It is one of the most expensive spices in the world. The major components of saffron are the apocarotenoids cis-and trans-crocins, picrocrocin and its degradation product the odor-active safranal. It has been suggested that a good quality saffron must contain about 30% crocins, 5 to 15% picrocrocin and more than 0.5% volatile safranal compounds. 1, 2 The coloring strength of saffron depends on the content of crocins. 3,4 Crocins are glycosyl esters of crocetin (Ct), in which glucose, gentiobiose, neapolitanose, or triglucose (C-1-C-5) are the sugar moieties. 1,5 Many methods for saffron component analysis have been described in the literature including reverse-phase high-performance liquid chromatography (HPLC) in combination with UV, diode array. 6 Recently, studies of diffuse reectance infrared Fourier transform spectroscopy (DRIFTS) 7 and 1 H NMR combined with chemometric techniques 8 have been proposed in evaluating the adulteration of saffron with characteristic adulterants of plant origin. At present, there is a growing tendency to nd quick, simple, and powerful tools which enable detection of saffron metabolites for quality and fraud control. [9][10][11][12] Considering that the sensory characteristics (avor, taste, color) and particular properties (health benets) of saffron are essentially determined by its chemical content, the application of MS-based chemical component proling offers signicant opportunities to obtain precious detailed information that can be directly correlated to the spice quality. Mass spectrometry is a wellestablished method for the high-throughput detection and quantitative analysis 13 of metabolites, 14 amino acids and their synthetic analogues, 15 and proteins. 16 Direct MS analysis of foods or food extracts 17,18 has been found to be a useful and robust approach for the chemical ngerprinting when rapid classication of food-sample types or rapid screening of food adulteration is wanted. Matrix-assisted laser desorption ionization mass spectrometry (MALDI MS) and tandem mass spectrometry (MS/MS) techniques have seldom been considered for both direct analysis of saffron extracts, and for quantifying adulterants in saffron. 19 MALDI-TOF-MS has been reported to be a very useful analytical technique for authenticity assessment. 17 Its advantages include high sensitivity, accuracy, tolerance to buffers and fast data acquisition. 20 The specicity of MS and MS/MS experiments enable the identication of markers through the elucidation of their molecular structures. Furthermore, MALDI MS system allows quantitative determination of low molecular weight compounds, since special instrument setup and isotopically labelled internal standard are not required. 20 Chen et al. have recently reported a new label-free LDI MS imaging technique, that combine advantage of including both imaging and quantitative analysis. 21 Limitations of MALDI experiments, such as the strong matrix interference and poor reproducibility of the signal intensity, appear to be solved by averaging multiple analyses of the same spot. 22 In this work, a simple method MALDI MS based for the fast quality control of saffron was exploited. The strategy developed has different stages: devising of sample preparation protocol to obtain a chemical component prole of saffron by MALDI MS that can be evaluated and matched to retrieve data by simple spectra comparison. MALDI MS component prole followed by MS/MS was employed to detect simultaneously species arising from saffron and adulterating matrix with sufficient abundance for quantitative and MS/MS analysis.

Spice and sample preparation
Samples of saffron spice were directly obtained from producers with a guarantee of their origin and freedom from fraud. The dried Crocus sativus L. stigmas were obtained from Cooperative of Saffron, (Krokos Kozanis, Greece). Powdered saffron, Curcuma longa (turmeric) and Calendula officinalis L. (calendula) were purchased from a local market. Crocus sativus L. stigmas, turmeric and calendula crude materials were ground into a ne powder. Aliquots of powdered turmeric, or calendula were used to prepare adulterated saffron samples used in chemical composition experiments. Saffron was homogenized by crushing in a pressure rolling with metallic cylinder. A portion (5 mg) of powdered saffron and adulterated samples were extract with 1 mL of ethanol, methanol/water (50/50: v/v), and 0.3% TFA in H 2 O/CH 3 CN (40 : 60, v/v) solution at room temperature, under magnetic stirring for 2 min. Aer 2 min centrifugation at 12 000 rpm, the pellet was discarded. A 1 mL portion of each sample was spotted and dried at room temperature. Matrix solution (1 mL), was pipetted onto dried samples. Aer the crystals had dissolved completely, the spot was dried under a continuous air stream. Matrix solutions were prepared dissolving 20 mg mL À1 of SA in 60% CH 3 CN/0.3% TFA in water (v/v).

Standard solution
A stock standard solution was prepared by dissolving curcumin in CH 3 CN up to a nal concentration of 20 mg mL À1 .
Set of authentic adulterated saffron sample was prepared as follows: a portion (5 mg) of powdered saffron was extracted with 1 mL of the extracting solution under magnetic stirring for 2 min. The extracting solution was prepared for each level of adulteration by adding 0.3% TFA in H 2 O/CH 3 CN (40 : 60, v/v), 1 mM of KCl to the appropriate amount of curcumin solution (20 mg mL À1 ) up to nal volume of 1 mL. A 1 mL portion of each sample was spotted and dried at room temperature.
The limit of detection (LOD) and the limit of quantitation (LOQ) were calculated by applying eqn (1) and (2), following the directives of IUPAC and the American Chemical Society's Committee on Environmental Analytical Chemistry. S LOD is the signal at the limit of detection, S LOQ is the signal at the limit of quantitation, S RB is the signal of the blank "authentic saffron samples", and s RB is the standard deviation.

MALDI-TOF-MS and high energy CID-MS/MS analysis
Each sample was directly spotted three times on a 384-well insert Opt-TOF™ stainless steel MALDI plates (AB SCIEX, Darmstadt, Germany). Mass spectrometric analyses were performed using a 5800 MALDI-TOF-TOF Analyzer (AB SCIEX, Darmstadt, Germany) equipped with an Nd:YLF Laser with l ¼ 345 nm wavelength of <500 ps pulse length and p to 1000 Hz repetition rate, in reectron positive mode with a mass accuracy of 5 ppm. Mass spectra were acquired automatically in the positive reector mode between 200 and 2000 with a xed laser intensity. Spectra with signal-to-noise below 200 were discarded automatically by the instrument. The operation parameter was optimized for the mass region of interest. Laser intensity was adjusted manually to avoid detector saturation. At least 4000 laser shots are typically accumulated with a laser pulse rate of 400 Hz in the MS mode, whereas in the MS/MS mode spectra up to 5000 laser shots are acquired and averaged with a pulse rate of 1000 Hz. MS/MS experiments were performed at a collision energy of 1 kV, ambient air is used as collision gas at a pressure of 10 À6 torr. The potential difference between the source acceleration voltage and the collision cell was set as 1 kV. Aer acquisition, spectra were handled using Data Explorer version 4.11 (AB Sciex). All data presented in this work are averages of three replicate.

Nomenclature for crocetin esters
To abbreviate the names of crocetin esters in this paper, they were labeled as follows: the nomenclature were written with a hyphen separating the total number of the glucose moieties at both extremes of the basic molecule (C-n). Then, C-4 would indicate Ct (crocetin), with four hexoses (Hex, glucose residues). The fragment ions were designated according to the nomenclature proposed by Domon and Costello 23 for glycoconjugates, and referring to the nomenclature of avonoid fragmentations. 24 Results and discussion

Method design
The rst step of the strategy comprised the optimization of extraction procedure to perform the simultaneous detection and characterization of the major components (crocins and picrocrocin) of saffron (Fig. 1A). The second step of the strategy was focused on the simultaneous detection and characterization of saffron and adulterant markers using crude extracts of adulterant itself and synthetic sets of adulterated authentic saffron samples ( Fig. 1B and C). Finally, the implementation of the strategy was to measure the amount of an unknown adulterant from the crude extract of fortied authentic saffron samples and saffron sachets purchased from market ( Fig. 1D).

Sample preparation and MALDI MS analysis of saffron
The goal of this study was to use a MALDI-TOF/TOF-MS instrument as a sensitive detection system to perform a fast quality control of saffron in order to detect saffron adulteration without pre-or intermediate purication or desalting procedures. The stigmas of Greek saffron were considered as the model system to elaborate the potential of chemical component prole. The direct MALDI-MS analysis of the sample can be combined with a simple and selective sample preparation preventing analyte losses. The sample preparation can be tailored since saffron marker of authenticity are well established and can be used to optimize the extraction procedure. The hydrophilic properties of crocins suggest possible different accumulation sites inside the chromoplasts. 25 In cells of saffron mature stigmas the water-soluble crocetin glycosides can be sequestered within the central vacuole which eventually fuses with chromoplast. This led to the hypothesis that the use of a binary solvent system has the potential to disrupt the cell membranes, and should favor the release of crocetin glycosides in aqueous media. 26 Therefore, direct detection of crocins from saffron using binary solvent system for extraction was planned. The structure of crocins C-1-C-6 from saffron has been elucidated. 27 Focus was pointed out on ion signals related to C-2 (m/z 675.27), C-3 (m/z 837.32), C-4 (m/z 999.38) and C-5 (m/z 1161.45). They gave good MS signals enabling the optimization of the extraction procedure (matrix solution preparation, sample/matrix solution ratio). To lyse cells and selectively solubilize metabolites, stigmas were extracted with different solvents such as ethanol, methanol/water, and a mixture 0.3% TFA in H 2 O/ CH 3 CN (40 : 60, v/v) (Fig. 1S, ESI †). Among these, the latter solvent system was shown to be most suitable for the solubilization of crocins from in house powered stigmas. Further, 5 mg of powered saffron were extracted using a 0.3% TFA in H 2 O/ CH 3 CN (40 : 60, v/v) for two minutes under magnetic stirring, at room temperature. A small aliquot of the crude extract (1 mL) was directly placed on the MALDI sample plate and analyzed. When the extraction time was extended to 5 minutes, no signicant changes in the extent of crocins from sample were observed in the recorded MALDI spectra.

Identication of crocins
The MALDI MS spectrum crude extract showed the specic m/z spacing patterns (n 162) of crocetin glycosides, providing glycoforms within the 0.5-2.0 kDa mass range. Almost all observable peaks corresponded to C-1-C-9 were unequivocally determined by the accurate mass of each peak (Table 1) (Table 1). The glycosyl compositions (Hex) 1 Ct, (Hex) 2 Ct, (Hex) 2 Ct(Hex) 1 , and (Hex) 2 Ct(Hex) 2 were assigned to crocins C-1-C-4, respectively. A detailed investigation of the product ions was performed by product ion scan of the  24 H 42 NaO 21 ] + ) were diagnostic for a tetraglycosyl chain. These data indicated that the ion of m/z 1161.45 arose from an isomeric mixture of the pentaglycosylated crocetins (Hex) 3 Ct(Hex) 2 and (Hex) 4 Ct(Hex) 1 . Furthermore, the absence of ion signals due to the 0,4 A 2 and the 3,5 A 2 cleavages clearly indicated linear glycosidic chains. The elucidation of unknown crocins C-8 and C-7 is illustrated in Fig. 2A-B. Based on the accurate mass analysis, the quasimolecular ion [M + Na] + (m/z 1647.59) corresponded to a new crocin bearing eight glucose residues Ct(Hex) 8 (Table 1, Fig. 2A). The high energy CID-MS/MS fragmentation of neutral oligosaccharides oen occurs from the non-reducing end toward the reducing end. Therefore, the consecutive losses of Hex from the quasimolecular ion, produced ions of m/z 1485.5 and m/z 1325.5, respectively, indicating that Hex glycosyl residues were on the non-reducing ends of the crocin C-8. Accordingly, Ct(Hex) 8 was recognized as a bis-substituted glycosyl esters of crocetin. The complementary product ions of m/z 675.2 (Y a0 ) and 995.2 (B a6 ) suggested that one of the glycosidic chains is composed by six glucose residues. Furthermore, the ions sequence B a3 , B a4 and B a6 were sufficient to determine the sequence of the carbohydrate moiety and to conrm the glycosyl composition as (Hex) 6 Ct(Hex) 2 for crocin C-8, as shown in the inset structure ( Fig. 2A). The high energy CID-MS/MS spectrum of the unknown crocin C-7 ([M + Na] + at m/z 1485.54; Table 1) is shown in Fig. 2B. The losses of two Hex residues from the quasimolecular ion gave the product ions of m/z 1325.5 and 1161.4, indicating that this glycosyl unit was present at both the non-reducing ends. The highest B type product ion of m/z 995.3 was assigned as [6Hex + Na] + . Accordingly, the crocin C-7 contains a hexaglycosyl chain as substituent. Again, the ion sequence of the product ions B a4 , B a5 and B a6 indicated a linear structure for the hexaglycosyl chain. Furthermore, the formation of the product ions Y a0 (m/z 513.2) and Z a0 (m/z 497.2) retaining the aglycon moiety, conrmed the composition as (Hex) 6 Ct(Hex) 1 for the precursor C-7. Similarly, the structure (Hex) 4 Ct(Hex) 2 and (Hex) 7 Ct(Hex) 2 were assigned to crocins C-6 and C-9, respectively. Table 1  These considerations led us to recognize the rst group as composed by glycoconjugated of "trimethylcyclohexene" (TMC) skeleton, Hex-O-C type structures; while glycosylated metabolites, such as avonoids, were collected in the second group.

Identication of carotenoid-derived metabolites
The cation adduct of the picrocrocin [M + K] + was detected at m/ z 369.13 ([C 16 H 26 O 7 K] + , Dppm ¼ 6.2, DBE ¼ 3.5). High energy CID-MS/MS (1 kV) induced extensive cross rings fragmentation providing valuable information for the characterization either the basic "trimethylcyclohexene" (TMC) skeleton and the sugar moiety. In particular, the cleavage at the glycosidic O-linkages with a concomitant H-rearrangement led to the elimination of 162 Da (hexose) yielding the Y 0 ion of m/z 207.1, base peak of the spectrum. However, the most useful fragmentations in terms of TMCs identication are those that require cleavage of two C-C bonds of the C-ring, resulting in structurally informative i,j TMC ions. The ions of m/z 167.1, 149.1 and 109.1 were due to the done by 3,4 TMC + , 3,5 TMC + and 3,6 TMC + cleavages and indicated two, and one CH 3 group at C6 and C2, respectively. In addition, losses of small molecules and/or radicals from the [M + K] + ion was observed. Losses of 15 Da (CH 3 ), 18 Da (H 2 O), 30 Da (CH 2 O), and the successive loss of these small groups were detected.
These mass losses were helpful to conrm the presence of specic functional groups in the structure. Thus, the CHO group at C1 was detected by the loss of 30 Da (CH 2 O, m/z 309.1) from the [M + K] + precursor ion (Fig. 3) 15

Identication of avonols
Flavonol compounds, such as kaempferol-3-sophoroside, kaempferol-3-sophoroside-7-glucoside, kaempferol tetrahexoside and kaempferol-3-dihexoside have already been described in Crocus sativus L. 30   were assigned to curcumin, demethoxycurcumin and bisdemethoxycurcumin, which are the predominant curcuminoids in turmeric. 32 The fragmentation behavior of the major curcuminoids was extensively studied and the main fragmentation channels of these compounds were elucidated. 33 13 enabled the rapid and unambiguous distinction between curcumin and picrocrocin whose structures were univocally conrmed (Fig. 5S †). Since interference was observed for only one endogenous molecule (picrocrocin m/z 369.13) during the analysis of authentic adulterated samples, the developed approach was thought to be useful to determine saffron adulteration regardless of the type of adulterant.
In absence of any chromatographic step, quantitation can be performed using non-isotopic isobaric internal standard. The endogenous metabolite and the isobaric compound are measured by MALDI MS. Since the amount of internal standard is known, and the amounts of internal standard and analyte can be determined from the mass spectra, the amount of the endogenous analyte can be calculated. These consideration leads to use curcumin as non-isotopic isobaric internal standard for saffron authentication.

Calibration curve experiments
A synthetic set of 8 adulterated authentic saffron samples with curcumin in the range of 0-30% was prepared. Thus, eight sets of synthetic sets of adulterated authentic saffron sample spots were processed through a simple "sample list". MS analysis were performed in triplicate using three spots for each sample. The linear range was assessed by plotting the mole ratio of the analyte (MR s ) vs. the percentage of adulteration with curcumin. Molar ratio was evaluated by the equation MR S ¼ P ICA S / ( P ICA S + P ICA T ), where ICA S and ICA T are the isotope cluster areas of specic ion signals from saffron and curcumin, respectively. In all set only C-2-C-4 and curcumin ion signals were used for data analysis. The calibration curve built using triplicate of synthetic sets of adulterated authentic saffron samples showed good linearity (R 2 ¼ 0.996), and with condence intervals at 95% for intercept including the zero value. The ANOVA regression model was signicant, F(1, 72) ¼ 16 440.31, p < 0.001, y ¼ (0,0120 AE 0.0001)x + (À0,1482 AE 0.0081) (Fig. 6S †). The maximum adulteration percentage was estimated (98.9%) interpolating the limit of detection (LOD) in linear regression. LOD and LOQ were 1.1% and 6% respectively. The repeatability calculated on spiked samples was found to be lower than 1% (RSD%). Accuracy of the method was determined by using fortied samples prepared by adding known quantities of the adulterant (citrus leaves, calendula) to authentic saffron samples (Table 2).
Quantitative recovery and high reproducibility (*RSD%) highlighted the reliability of the method, suggesting that the developed approach is suitable for a rapid screening of saffron. Comparison between the determined LOD with values obtained as previously reported employing HPLC/PDA and/or ESI-MS detection, namely 5% (w/w) for calendula or safflower and 2% (w/w) for turmeric, 9 implied that the proposed approach may enable detection at lower levels of plant-derived adulterants in saffron. The recently published and highly sensitive methods that combine LC and MS to assess the authenticity of saffron through the analysis of a group of kaempferol derivatives 11 and through geniposide 12 have LOD ranging from 0.2-2% and 10 ng mL À1 respectively. However, the accuracy of the most sensitive method (LOD 10 ng mL À1 ), assessed evaluating the recovery obtained for geniposide in spiked saffron sample with 1 mg mL À1 of geniposide, was 89 AE 14%. 12 The relatively poor temporal resolution, due to sampling times of up to 20 minutes, is the major limit. Throughput of the LC MS systems is restricted to a limited number of samples. The MALDI MS method here described shows LOD comparable to that of LC/ MS methods. However, in our case, chromatographic separations or desalting are not required, and the recovery is quantitative. Moreover, the MALDI-MS method enables a high sample throughput because of the minimal sample preparation and the very short measuring time per sample. Since no interference was observed for both endogenous molecule and the non-isotopic isobaric internal standard during the analysis of authentic and spiked samples, the developed approach could be used to determine saffron adulteration regardless of adulterant type. Furthermore, the reliability of the proposed approach was conrmed by analytical parameters of accuracy calculated using spiked samples prepared by adding known quantities of citrus leaves and calendula (Table 2). Spiked samples given accuracy higher than 97% ( Table 2). Recovery of the analyte was quantitative because no chromatographic step is involved in sample preparation.
The good performance of the method was shown also by the repeatability of the measurements, with the *RSD% value being lower than 2%, in all the examined cases ( Table 2). The method was further applied to saffron sachets which were available at local stores (Table 3). These samples were chosen on the basis of their low cost and questionable origin.
MALDI MS analysis of suspicious saffron samples (S1-S3) The estimated adulteration percentage for saffron samples (S1-S3) ranged from 10% to 25% (Table 3). MALDI MS spectra of these samples (S1-S3) showed similar molecular prole. Representative MALDI spectrum of suspicious saffron samples (S3) are shown in Fig. 5A. MS/MS spectra of the ion specie of m/z 369. 13 were acquired in all case in order to establish the presence of turmeric as adulterant (Fig. 5B, sample S3). The   K] + , Dppm ¼ 5) were assigned to sodium and potassium adducts of geniposide, respectively (Fig. 5C). Geniposide, is a sensitive biomarker for the detection of adulteration of saffron samples, since represents the main component of extract of fruit gardenia. 12 The detection of geniposide in all samples S1-S3 indicated adulteration by Gardenia jasminoides.

Conclusions
MALDI MS of saffron extracts, which has the advantage of focusing on the more polar components, was found to be a fast and selective technique to control quality and to evaluate adulteration level. No matrix effects were observed and good results were obtained with respect to instrumental repeatability (*RSD% < 2%) and LOD (1.1%). The analysis of commercial samples of saffron using the proposed approach showed the suitability of the method in routine analysis (minimal sample preparation and very short measuring time per sample). The data here reported showed that the proposed approach enables quantitation of adulteration percentage in saffron, regardless of adulterant identity, and identication of the adulterant itself.

Conflicts of interest
There are no conicts to declare.