Comprehensive Evaluation of the Volatomic Fingerprint of Saffron from Campania towards Its Authenticity and Quality

The volatile profiles of eight saffron samples (seven cultivated and one spontaneous) grown in different geographical districts within the Campania region (southern Italy) were compared. Using headspace solid-phase microextraction coupled to gas chromatography–mass spectrometry (HS-SPME/GC-MS), overall, 80 volatiles were identified in the eight landraces. Among them, safranal and its isomers and other related compounds such as isophorones, which are not only key odorants but also pharmacologically active metabolites, have been detected in all the investigated samples. Principal Component Analysis performed on the volatiles’ compounds revealed that the spontaneous sample turned out to be an outlier. In particular, the volatile organic compounds (VOCs) profile of the spontaneous saffron presented four lilac aldehydes and four lilac alcohol isomers, which, to the authors’ knowledge, have never been identified in the volatile signature of this spice. The multivariate statistical analysis allowed the discrimination of the seven cultivate saffron ecotypes in four well-separated clusters according to variety. Moreover, 20 VOCs, able to differentiate the clusters in terms of single volatile metabolite, were discovered. Altogether, these results could contribute to identifying possible volatile signature metabolites (biomarkers) or patterns that discriminate saffron samples grown in Campania region on a molecular basis, encouraging future biodiversity programs to preserve saffron landraces revealing valuable genetic resources.


Introduction
Saffron, the dried red stigma of flowers of Crocus sativus L., is highly valued for its unique aroma, taste and color. It is called "the red gold" and is currently considered to be the world's most expensive spice [1,2]. Saffron has been known for four millennia and probably originated in the Middle East or in Greece (the southern Aegean islands of Crete and Santorini) [3]. Although it is principally used as a natural dye and as a flavoring in culinary preparations, several pharmacological properties, including antioxidant, antiinflammatory, antidepressant and anticancer, have been confirmed for saffron [3]. In a very recent study, saffron has been also indicated as a promising anti-inflammatory and antiviral herbal medicine in the prevention of severe COVID-19 symptoms [4].
With a world production of 418 t per y-1, this spice can be cultivated in a range of different pedo-climatic conditions, which significantly impact its composition and quality. Iran is the principal producer with 90% of global production followed by India, Afghanistan, Morocco, and Euro-Mediterranean countries, including Greece, Spain and Italy [3]. In recent years, in Euro-Mediterranean basin countries there has been a drastic reduction in saffron by HS-SPME/GC-MS combined with chemometric tools could be used as an authenticity marker to define the variety and geographical origin, as well as to detect frauds for different foodstuffs, including saffron, providing local growers with numerous benefits [2,7]. In this context, comprehensive VOCs profiles of the abovementioned eight saffron samples have been obtained by HS-SPME/GC-MS method and the volatomic data were treated by multivariate statistical analysis, with the aim of contributing to the characterization of the volatile components and to identify possible signature metabolites (biomarkers) or patterns that discriminate the samples under examination.

Chemicals and Materials
All reagents used in this study were of analytical quality. Sodium chloride (NaCl, 99.5%) was supplied by Panreac (Barcelona, Spain), while 3-octanol and the C8-C20 nalkanes mixture were purchased from Sigma-Aldrich (St. Louis, MO, USA). Ultrapure water (18 MΩ cm at 23 • C), obtained from a Milli-Q system (Millipore, Bedford, VA, USA), was used throughout. The SPME fibers (divinylbenzene/carboxen on polydimethylsiloxane-DBV/CAR/PDMS, with 50/30 µm film thickness and 1 cm fiber length), the SPME holder for manual sampling and the amber glass screw cap vials for SPME with PTFE/silica septa were from Supelco (Bellefonte, PA, USA). Helium (ultrapure grade, Air Liquide, Algés, Portugal) was used as the carrier gas in the GC system.

Saffron Samples
The stigmas of saffron (Crocus sativus L.) of the eight landraces, harvested in 2021, were provided by local growers from different geographical districts within the Campania region (southern Mediterranean area of Italy), through local committee promoters. In  In particular, RRWT sample refers to a spontaneous ecotype.
Each saffron ecotype was grown, harvested and processed in agreement with the guidelines provided by the collective brand "Zafferano Italiano" (https://www.zafferanoitaliano.it (accessed on 1 December 2021)). Individual saffron samples, harvested at the end of the growth cycle, were dried at a temperature inferior to 45 • C with a final a moisture content not superior to 12%. All the saffron samples, stored in small glass jars, showed stigmas with a length ranging from 1 to 3.5 cm which in the extreme upper part had broad and cylindrical papillae and presented a purple-red color.

Sample Preparation and HS-SPME Procedure
The optimization of HS-SPME experimental parameters was carried out by analyzing commercial saffron samples obtained from a local market (Madeira Island).
For each variety, saffron stigmas were ground into a fine powder just before the analysis. The sample preparation procedure was the following: for each replicate, 1 g of each powdered saffron sample was mixed into a 20 mL amber glass vial to 0.5 g of NaCl, 6 mL of ultra-pure Milli-Q water and 5 µL of 3-octanol (0.4 µg/mL), used as the internal standard (IS). Successively, a magnetic stirrer was added into the vial which was capped with a PTFE-faced silicone septum and positioned in a thermostatic bath at 45 ± 1 • C to equilibrate the system. The SPME was executed in headspace mode (HS), where the fiber was fixed to a manual SPME holder and exposed to the HS sample for 50 min under constant magnetic stirring (800 rpm). The VOCs extracted by SPME were thermally desorbed by the direct insertion of the fiber into GC injector at 250 • C for 10 min, in splitless mode. SPME fibers were conditioned as suggested by the manufacturer, but below the maximum recommended temperature prior to their first use. Before the initial daily analyses, the fibers were conditioned at 250 • C for 10 min into the GC injector port and the blank level was checked. Experiments were performed in triplicate for all samples, except for RRWT saffron that was analyzed in duplicate.

Gas Chromatography-Mass Spectrometry Analysis (GC-MS)
Chromatographic separations of volatile metabolites from the eight saffron ecotypes were performed using an Agilent 6890N (Agilent Technologies Palo Alto, CA, USA) gas chromatography system equipped with a SUPELCOWAX ® 10 fused silica capillary column (60 m × 0.25 mm i.d. × 0.25 µm film thickness) supplied by SGE (Darmstadt, Germany), and coupled to a mass spectrometer 5975 C (Agilent, Santa Clara, CA, USA) with a quadrupole inert mass selective detector (Santa Clara, CA, USA). The oven temperature program was initially set at 40 • C for 1 min and then increased to 220 • C at 2.5 • C/min and held for 10 min, for a total GC run time of 83 min. The flow of He, the carrier gas, was kept at 1 mL min −1 , while the injection port operated in the splitless mode at 250 • C. For the VOCs detection, the operating temperatures of the transfer line, quadrupole, and ionization source were 270, 150, and 230 • C, respectively, while the electron impact (EI) mass spectra were recorded at 70 eV ionization voltages and the ionization current was 10 µA. The electron multiplier was set to the auto-tune mode and the mass acquisitions range was 30-300 m/z. The resulting chromatograms were processed by using the Enhanced ChemStation software for GC-MS (Agilent Technologies, Palo Alto, CA, USA).
Volatile metabolites were identified by matching their mass spectra with those reported in the standard NIST05/Wiley07 libraries, by comparing the retention indices (RI) (as Kovats indices), calculated relative to the RT of a series of n-alkanes (C8-C20) with linear interpolation, with of the data from the literature, and by comparison of the GC RT of the chromatographic peaks with those, when available, of pure standards run under the same conditions. For individual volatiles, the peak area was measured from the total ion chromatogram (TIC) and semi-quantified by relative comparison with the peak area of the IS (Relative Peak Area, RPA%).

Statistical Data Analysis
Exploratory data analysis was performed by Principal Component Analysis (PCA) [8]. PCA is an unsupervised multivariate technique able to summarize the data variation of a data table using a small set of score components, called principal components, calculated by linear combination of the measured features. The model of the data table obtained by PCA approximates the original data structure providing a less complex data-representation, where the similarity between observations and the correlation structure among the measured features are preserved. Moreover, since PCA discovers the structured data variation removing both random noise and redundant data variation due to multicollinearity, the space spanned by the principal components provides a more useful data representation for clustering than that obtained by the original features. Using the principal components as coordinates to represent the observations, Hierarchical Cluster Analysis (HCA) based on the Euclidean distance and the Ward's method was applied, to discover relevant clusters of observations. The optimal number of clusters was determined by Silhouette analysis [9]. PCA was also used for outlier detection applying the Hotelling's T2 and the Q tests.
The clusters discovered combining PCA and HCA were characterized in terms of single VOCs using the Kruskal-Wallis test controlling False Discovery Rate (FDR) by Benjamini-Hochberg procedure [10]. The distributions of the most interesting features were represented by boxplots. The relationships between VOCs and altitude were explored by correlation analysis. Specifically, Spearman's correlation coefficient was calculated for individual volatile and altitude. Controlling FDR by Benjamini-Hochberg procedure, a set of VOCs related to altitude was discovered and the results were represented using the Volcano plot.
Data analysis was performed by in house R-function implemented, using the R 4.0.4 platform (R Foundation for Statistical Computing, Vienna, Austria).

Volatomic Profile of Saffron Samples
The volatile profiles obtained from the eight saffron samples analyzed are quite different ( Figure S1).
Overall, 80 volatiles were identified by HS-SPME/GC-MS analysis in the eight saffron landraces belonging to the following chemical classes: monoterpene ketones (12), monoterpene hydrocarbons (10), monoterpene aldehydes (7), monoterpene alcohols (6), aldehydes (14), alcohols (7), ketones (7), hydrocarbon (7), furans (3), sesquiterpenes (2), esters (2), and others (3). The identified 80 VOCs, the abbreviation code, the experimental and literature Kovats index and the identification methods are reported in Table 1. For each identified VOC, the medians of the RPA% values and the corresponding range are reported in Table S1. Moreover, Figure 1 reports the volatile content detected in the saffron samples from different geographical sites, while Figure 2 indicates that monoterpene aldehydes, followed by monoterpene ketones, are clearly the most abundant VOCs identified in all the saffron ecotypes.
Among the C10 volatile components, besides saffron and its isomers, β-cyclocitral (V55) was present in the volatile patter of EC, TL and MC samples, while HTCC (V80) has been only found in the cultivated samples (Table S1). β-cyclocitral, also present in other plants, has been described as an important component of flavor and aroma in many fruits, vegetables and ornamental plants contributing in attracting pollinators [5,16].
Among the volatiles exclusively found in the RRWT sample, the compounds V49, V50, V51, and V52 correspond to the lilac aldehyde isomers A, B, C, and D, respectively, while the components V60, V62, V63, and V68 refer to the lilac alcohol isomers A, B, C, and D, respectively (Table S1). These oxygenated monoterpenoids are from the most complicated chiral floral scent compounds and have been identified among the main volatile components in many plants, most of all ornamental species, harvested all over the world, as important metabolites for the attraction of pollinators. In addition, these VOCs have been also described as key volatile markers in several honeys [22][23][24]. To the authors knowledge, lilac aldehyde and lilac alcohol isomers have never been detected in the VOCs profile of saffron.

Discrimination of Saffron Samples
To evaluate the potential of the volatile profiles obtained by HS-SPME/GC-MS in the discrimination of saffron samples according to the variety, a statistical analysis of the volatomic data matrix was accomplished. Specifically, the data set composed of 23 observations (3 technical replicates for each variety except for 2 replicates for the spontaneous variety, RRWT) and 80 VOCs was investigated by PCA. Data were log-transformed and auto scaled before performing data analysis. The model showed two principal components with R 2 = 0.65 and the relative biplot is reported in Figure 3A.

Discrimination of Saffron Samples
To evaluate the potential of the volatile profiles obtained by HS-SPME/GC-MS in the discrimination of saffron samples according to the variety, a statistical analysis of the volatomic data matrix was accomplished. Specifically, the data set composed of 23 observations (3 technical replicates for each variety except for 2 replicates for the spontaneous variety, RRWT) and 80 VOCs was investigated by PCA. Data were log-transformed and auto scaled before performing data analysis. The model showed two principal components with R2 = 0.65 and the relative biplot is reported in Figure 3A. The two observations of the RRWT sample are located at the limit of the plot, suggesting a relevant difference with respect to the other observations ( Figure 3A). Indeed, assuming α = 0.05, the replicates of the spontaneous sample were outliers, as it is possible to observe in Figure 3B, where the T2/Q plot is illustrated.
Excluding the RRWT sample, the data set composed of the remaining 21 observations were log-transformed and autoscaled and investigated by PCA. Considering a model with three principal components and R 2 = 0.69, HCA allowed us to identify four clusters of observations. In Figure 4 the dendrogram and the Silhouette plot used to define the optimal number of clusters is reported. It is worth noting that the data variation associated to the technical replicates is less relevant than that associated to the biological variety, as technical replicates of the same saffron sample are clustered together. Table 2 lists the composition of the identified clusters in terms of the saffron ecotype, while the biplots of the PCA model, where the observations are coloured according to the cluster membership, The two observations of the RRWT sample are located at the limit of the plot, suggesting a relevant difference with respect to the other observations ( Figure 3A). Indeed, assuming α = 0.05, the replicates of the spontaneous sample were outliers, as it is possible to observe in Figure 3B, where the T2/Q plot is illustrated.
Excluding the RRWT sample, the data set composed of the remaining 21 observations were log-transformed and autoscaled and investigated by PCA. Considering a model with three principal components and R 2 = 0.69, HCA allowed us to identify four clusters of observations. In Figure 4 the dendrogram and the Silhouette plot used to define the optimal Foods 2022, 11, 366 9 of 14 number of clusters is reported. It is worth noting that the data variation associated to the technical replicates is less relevant than that associated to the biological variety, as technical replicates of the same saffron sample are clustered together. Table 2 lists the composition of the identified clusters in terms of the saffron ecotype, while the biplots of the PCA model, where the observations are coloured according to the cluster membership, are described in Figure 5, which showed that the four clusters are in different regions of the plots and are well-separated. By Kruskal-Wallis test controlling FDR at the level δ = 0.01, 20 VOCs, able to characterise the clusters in terms of single volatile metabolite, were discovered (the p-values of the tests are reported in Table S2 of Supplementary Materials). In Figure 6, the boxplots of the selected VOCs are reported.     Since the maximum of SUMsilhouette is obtained for Ncluster = 4, the optimal number of clusters was 4. Compared to the other samples, in the cluster A, corresponding to the GB sample, 8 VOCs, including dimethyl sulphide (V1), β-myrcene (V15), α-phellandrene (V16), αterpinene (V17), γ-terpinene (V24), trans-β-terpinolene (V29), β-sesquiphellandrene (V66) and α-curcumene (V67) showed the highest amount, while 1,2,3-trimethylbenzene (V35), and 4-ketoisophorone (V59), were the less abundant components. Moreover, GB is the only sample in which eucarvone (V64) has not been detected (Table 2; Figure 6; Table S1). In particular, V1, V15, V17, and V24 have been found only in the GB sample (Table 1). To our knowledge, dimethyl sulphide (V1) and β-sesquiphellandrene have been identified for the first time in saffron. β-Myrcene (V15), described with a peppery odor, has been reported in Greek saffron stigmas dried by the traditional method [11], α-phellandrene (V16), characterized by a citrus, slightly green, black pepper-like odor, has been detected in the headspace analysis of a French saffron sample by using both GC×GC-ToF-MS and GC-ToF-MS [18], while α-terpinene (V17), γ-terpinene (V24), and terpinolene (V29), all having a citrus and herbaceous smell, have been first identified by Condurso et al., among the volatile constituents of Sicilian (South Italy) saffron samples [5]. Moreover, V24 has been recently detected in the volatile profile of a raw Spanish saffron sample [25]. Finally, α-curcumene (V67) has been found in the VOCs profiles of both stigmas and petals of Iranian saffron ecotypes with different geographical provenances and cultivated under different nutritional regimes [17]. Terpenes (mono and sesquiterpenes) are the most prevalent constituents in the essential oils of numerous plants, including saffron, and many of these metabolites have recognized potential biological effects. Specifically, V15, V16, V17, V24, and V29 and have been reported to exhibit antibacterial, antifungal, antiproliferative, antitumor, analgesic, and anti-inflammatory activity [26,27]. Several studies have also demonstrated that the combination of different monoterpenes produce a remarkable synergistic effect enhancing their biological properties [28].     Table 2.
Cluster D, composed of the RR and TL samples, presented the lowest amount of α-isophorone (V53) and 4-ketoisophorone (V59) which have been discussed above as being among the most abundant volatiles in samples of cluster C (Table 2; Figure 6; Table S1).

Conclusions
In this study, HS-SPME/GC-MS was used to characterize, for the first time, the volatile profile of eight saffron samples (seven cultivated and one spontaneous) grown in different geographical districts within the Campania region (southern Italy). Overall, 80 volatiles were identified in the eight landraces. Among them, safranal, its isomers, and isophoronerelated compounds were reported to be not only key odorants but also responsible for the pharmacological activity showed by saffron extracts, and have been detected in all the investigated samples.
PCA performed on the volatiles compounds showed that the spontaneous saffron (RRWT) was an outlier. Particularly, the VOCs profile of RRWT presented four lilac aldehyde and four lilac alcohol isomers, which, to the authors' knowledge, have never been detected in the volatomic fingerprint of this spice. Multivariate statistical analysis allowed the discrimination of the seven cultivate saffron samples in four distinct clusters according to variety. Moreover, 20 VOCs, able to differentiate the clusters in terms of single volatile metabolite, were discovered. Noticeably, the clear definition of these biomarkers would require dedicated studies, extending both the number of samples and the biological replicates.
Overall, our findings could promote future breeding programs aimed at safeguarding and preserving the production of the investigated saffron landraces from the Campania region despite the presence of biologically active volatile constituents characterized by relevant health and sensory properties.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/foods11030366/s1, Figure S1: Representative chromatograms of the eight saffron samples (see text for the sample code). Numbers above the peaks refer to the VOCs identified in Table 1.

Informed Consent Statement: Not applicable.
Data Availability Statement: Excluded statement.