Characterization of Botanical and Geographical Origin of Corsican “Spring” Honeys by Melissopalynological and Volatile Analysis

Pollen spectrum, physicochemical parameters and volatile fraction of Corsican “spring” honeys were investigated with the aim of developing a multidisciplinary method for the qualification of honeys in which nectar resources are under-represented in the pollen spectrum. Forty-one Corsican “spring” honeys were certified by melissopalynological analysis using directory and biogeographical origin of 50 representative taxa. Two groups of honeys were distinguished according to the botanical origin of samples: “clementine” honeys characterized by the association of cultivated species from oriental plain and other “spring” honeys dominated by wild herbaceous taxa from the ruderal and/or maquis area. The main compounds of the “spring” honey volatile fraction were phenylacetaldehyde, benzaldehyde and methyl-benzene. The volatile composition of “clementine” honeys was also characterized by three lilac aldehyde isomers. Statistical analysis of melissopalynological, physicochemical and volatile data showed that the presence of Citrus pollen in “clementine” honeys was positively correlated with the amount of linalool derivatives and methyl anthranilate. Otherwise, the other “spring” honeys were characterized by complex nectariferous species associations and the content of phenylacetaldehyde and methyl syringate.


Introduction
The specificity of Corsican honeys is linked with the environmental characteristics of the island (biodiversity of flora, bioclimatic conditions and topography), the endemic black honeybee and typical hive management. Organoleptic and melissopalynological analysis have permitted Corsican honeys to be classified into six ranges: "spring", "spring maquis", "honeydew maquis", "chestnut grove", "summer maquis" and "autumn maquis", according to the harvest season and the geographic location anthranilate, lilac aldehyde and p-menth-1-en-9-al, were therefore suggested as the chemical markers of citrus (species not specified) unifloral honey [13][14][15]. Moreover, Alissandrakis et al. [16] showed that the volatile fractions of citrus flowers (four species) and the corresponding honeys were dominated by linalool derivatives. The phenolic compound hesperetin was also proposed as a botanical indicator of Spanish citrus honeys for its high levels in nectar and honey [17]. Methyl syringate and/or phenylacetaldehyde were identified as characteristic components of nectar from A. microcarpus Salz et Viv. and corresponding unifloral honeys [18,19].
Several techniques (HS-SPME, infrared spectroscopy and 1 H-nuclear magnetic resonance spectroscopy) have been used to distinguish Corsican and non-Corsican honeys, but these studies did not provide results for the differentiation of the botanical origin of different ranges of Corsican honey [20][21][22].
According to the geographical and botanical origins of Corsican "spring" honeys certified by melissopalynological analysis, the chemical composition of volatile fractions of honey samples was established using HS-SPME, GC and GC/mass spectrometry (MS). The aim of the study is to establish for the first time a multidisciplinary method for the qualification of Corsican "spring" honeys, based on relationships between the pollen spectrum, volatile chemical markers and some physicochemical parameters.

Honey and Flower Sampling
In total, 41 Corsican "spring" honeys (samples 1-41) were selected from our reference bank of honey with AOC and PDO appellations. All these samples were directly packaged in a sealed pot and stored below 14 °C according to the optimal conditions of honey conservation indicated by Gonnet et al. [23]. The honey samples of three years of harvest (2004)(2005)(2006) collected in April to June were provided from 12 Corsican producers. The apiaries were located from littoral to 400 m (principally under 100 m) in the oriental cultivated plain or in ruderal and/or maquis zone of thermo-and meso-Mediterranean levels. Clementine (Citrus sinensis × reticulate, six samples) and Asphodel (Asphodelus ramosus subsp. ramosus, six sample locations) flower specimens were collected in March-May 2009-2012. The nectar secretion during harvest period was ensured by the observation of foraging nectar by honeybees. Flowers samples were analyzed within 48 h.

Melissopalynological Analysis
In this study, melissopalynological analysis was performed using the method described by Yang et al. [24]. Identification of pollen in the "spring" honey was based on the comparison with laboratory's own reference pollen-slides library and also carried out with the palynological expertise practice [5,11] developed for the characterization and the AOC and PDO control of Corsican honeys. Pollen analysis was allowed to establish a total pollen spectrum (qualitative analysis) and pollen density (quantitative analysis) for each honey sample. The identified taxa in the pollen spectrum were expressed in term of relative frequency (RF) and the pollen density was expressed as the absolute number of pollen grain in 10 g of honey (PG/10 g).

Physicochemical Analysis
According to the description of Corsican honeys [1,5], two physicochemical parameters, coloration and electrical conductivity were chosen to complete the botanical origin characterization of Corsican "spring" honey. The honey coloration was measured using a Lovibond Comparator apparatus [25]. Results were expressed as millimeters (mm) Pfund. Electrical conductivity was measured at 20 °C with a conductivity meter micro CM2210 (CRISON, Spain) following the method described by Bogdanov [26] and expressed as milliSiemens per centimeter (mS/cm).

HS-SPME Extraction
Volatile fractions of honey and flower samples were extracted by HS-SPME with a divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS, 30 μm) fiber (Supelco Sigma Aldrich). The optimization of HS-SPME parameters was performed using two honey samples (9 and 24) and two flower samples (clementine and asphodel flowers). These samples and subsequent analyses (all honey and flower samples studies) were performed in triplicate to ensure that the coefficient of variation (CV: ratio of standard deviation to the mean) of the major compounds and the sum of the total peak areas were always <15%. The samples analyzed were placed in a 20 mL vial. The parameter optimization was based on the sum of the total peak areas measured using a gas chromatography-flame ionization detection (GC-FID) system. For each sample (both honeys and flowers): the temperatures (25 °C, 50 °C and 70 °C), the equilibration times (30, 60 and 90 min) and the extraction times (15,30 and 45 min) were tested in various experiments. The honey concentration in distilled water was optimized after six different experiments (0.5 g/mL, 1 g/mL, 1.5 g/mL and 2 g/mL) with Na 2 SO 4 addition (1 g and 2 g). The maximum sum of the total peak areas was obtained from 4 g of honey sample with 4 mL of water and 2 g of Na 2 SO 4 at a temperature of 70 °C, an equilibrium time of 90 min, and an extraction time of 30 min. The flower weight was optimized after three different experiments (1 g, 3 g and 5 g). For the Asphodel flowers, the maximum sum of the total peak areas was obtained from 3 g of sample at a temperature of 70 °C, an equilibrium time of 90 min, and an extraction time of 30 min. Otherwise, the best sampling conditions of Clementine flowers were 1 g of sample at room temperature (25 °C) with an extraction time of 15 min. Before sampling, the fiber was reconditioned for 5 min in the GC injection port at 280 °C. After sampling, the SPME fiber was consecutively inserted into the GC-FID and GC-MS injection ports for 5 min for desorption of volatile components, both techniques using the splitless injection mode.

GC-FID and GC-MS Analysis
GC-FID analyses were performed using a PerkinElmer (Waltham, MA, USA) AutoSystem XL GC apparatus equipped with a FID system and a fused-silica capillary column (30 m × 0.25 mm, film thickness 1 μm) coated with Rtx-1 (PDMS). The oven temperature was programmed from 60 to 230 °C at 2 °C/min and then held isothermally at 230 °C for 35 min. The injector and detector temperatures were maintained at 280 °C. The samples were injected with an SPME inlet liner (0.75 mm i.d.; Supelco) using hydrogen as the carrier gas (1 mL/min). The retention indices of the compounds were determined relative to the retention times of a series of n-alkanes (C 5 -C 30 ) with linear

Statistical Analysis
The statistical analysis of melissopalynological data was carrying out the methodology previously described by Battesti et al. [11]. In the case of "spring" honey, the inclusion of Citrus and Asphodelus pollen during the nectar foraging is low or very low because of pollen maturity or floral morphology. The "under-representation" of these pollen types and entire pollen spectrum were taken into account for the characterization and comparison of pollen spectrum from "spring" honeys. Principal component analysis (PCA) was carried out using the "PCA" function and canonical correspondence analysis (CCA) was performed with "CCA" function from R software (R Foundation-Institute for Statistics and Mathematics, Austria). CCA is a multidimensional exploratory statistical method in order to demonstrate the correlation between two sets of variables obtained from the same individual.

Determination of Geographical and Botanical Origins of Corsican "Spring" Honeys
The analysis of 41 Corsican "spring" honeys allowed the determination of 92 taxa, including 64 nectariferous taxa and 28 only-polleniferous taxa (Table 1). A biogeographical analysis (biogeographical code: BC [5]) showed the diversity of biogeographical origins of these taxa. Mediterranean species (28 taxa, BC 1-3) associated with Eurasian and Atlantic species (13 taxa, BC 5-6) were well represented in the pollen spectrum. Additionally, cultivated species (four taxa, BC 99) were reported in more than 40% of honey samples. This distribution was consistent with the database of the characterization of the Corsican honey taxa directory [5,11].
In the case of honey with the "under-represented" pollen type, the contribution of other nectariferous species could not be discounted. It had to note that some honeys samples possessed dominant nectariferous taxa (RF > 45%): Trifolium sp. T9 for sample 2 and 3, Echium sp. T13 for sample 4 and Lotus sp. T7 for sample 38. The nectar contribution of these taxa could not be neglected. Otherwise, several taxa might take part in the honey composition for their high RF in the pollen spectrum: Trifolium sp. T9 and E. arborea T5 were characteristic for both groups (RF max 53.5% and 35.5% for group I and 44.1% and 29.9% for group II, respectively); Echium sp. T13, Prunus form T11 and Viburnum tinus T26 possessed a higher RF max in group I (71.1%, 24.1% and 16.2%, respectively) than in group II (30.1%, 3.3% and 3.7%, respectively), while Lotus sp. T7, Genista form T6, Salix sp. T8, Lupinus angustifolius T41 and Apiaceae T14 were higher in group II (FR max : 52.8%, 31.5%, 29.9%, 18.9% and 17.5%, respectively) than in group I (FR max : 8.7%, 11.8%, 12.9%, 3.5% and 4.5%, respectively).

Physicochemical Characteristics of Corsican "Spring" Honeys
Corsican "spring" honeys possessed light to very light colors. The mean value of coloration was 30.0 ± 15.4 mm Pfund, with great variation between 11.0 and 71.0 mm Pfund ( Table 2). The two groups exhibited quite similar coloration values: 26.4 ± 13.6 mm Pfund for "clementine" honeys and 33.3 ± 16.3 mm Pfund for the other "spring" honeys. For each group, nine samples possessed a very light coloration value (<20.0 mm Pfund). Only one sample (17)  The average electrical conductivity value of the honey samples was 0.25 ± 0.08 mS/cm with a variation of 0.13-0.45 mS/cm ( Table 2). The electrical conductivity of the two groups was also quite similar: 0.25 ± 0.07 mS/cm for "clementine" honeys (range: 0.15-0.42 mS/cm) and 0.24 ± 0.09 mS/cm for other "spring" honeys (range: 0.13-0.45 mS/cm). Only three samples (17,34 and 41) of these honeys had medium electrical conductivity (>0.4 mS/cm).
The coloration and electrical conductivity values of Corsican "spring" honeys were lower than those of "chestnut grove" and "Erica arborea spring maquis" honey ranges [5,24,29].

Chemical Variability of Corsican "Spring" Honeys
GC and GC/MS analysis of the headspaces of Corsican "spring" honeys allowed the identification of 43 compounds that accounted for 71.5%-96.8% of the total volatile composition ( Table 3,  Table S2-supplementary materials). It should be noted that the volatile fraction of "spring" honeys is rich in aldehyde (22.1%-63.1%) and alcohol (2.8%-40.2%) components.
The decrease in linalool amount and the occurrence of other linalool derivates (hotrienol, linalool oxides, lilac aldehyde isomers and p-menth-1-en-9-al isomers) in honey samples could be explained by the enzymatic degradation of linalool by some pathways [15]: (1) linalool can be transformed to 8-hydroxylinalool isomers by enzymatic hydroxylation at the C8 position, and then hotrienol; (2) 8-hydroxylinalool can be transformed to lilac aldehyde via (E)-8-oxolinalool and lilac alcohols, or p-menth-1-en-9-al via 8-hydroxygeraniol and (3) linalool can also be transformed via 6,7-hydroxylinalool into furanoid linalool oxide isomers under acidic conditions or by heating. These results were in accordance with those previously reported on the volatile fraction of citrus flowers and corresponding honeys [16]. It demonstrated that the flowers from Citrus species (orange, tangerine and sour orange) had high amounts of linalool (51.6%-80.6%) and that the honeys consisted of more than 80% of linalool derivatives (lilac aldehydes and lilac alcohols).  The 23 "not-clementine" honey samples (group II) were dominated by phenolic compounds (23.2%-60.4%) followed by linear compounds (11.3%-53.4%). The main compounds were phenylacetaldehyde C14 (3.7%-36.2%), benzaldehyde C9 (2.5%-18.4%) and methyl-benzene C2 (1.5%-17.3%). Furanic compounds (average: 7.5%) were less abundant than in "clementine" honeys (average: 26.2%), and acid components (average: 10.3%) were more abundant than in the "clementine" honeys (average: 6.8%). To our knowledge, only one previous report focused on the volatile fraction of asphodel unifloral honeys from Sardinia [18]. Methyl syringate was detected in asphodel nectar in high concentrations and was therefore considered a marker of asphodel honeys [19]. A low content of this component (C42: 0.1%-4.1%) was reported in the volatile fraction of "spring" honey samples (18-21, 24-30, 32, 33 and 36-41). Additionally, the amount of methyl syringate was unrelated to the presence of Asphodelus pollen in the pollen spectrum. This result could be explained by the extreme "under-represented" type of Asphodelus pollen in Corsican "spring" honeys and/or by other nectar contributions in these honeys. The sample 18 exhibited the association of Citrus sp. and A. sinensis; it was grouped with the "not-clementine" honey. In this sample, the citrus nectar contribution was less important than in "clementine" honeys in accordance with the lower concentrations of lilac aldehyde isomers.
To our knowledge, the volatile composition of A. ramosus subsp. ramosus flowers is reported here for the first time ( Table 4). The HS-SPME volatile fraction of asphodel flowers was dominated by oxygenated compounds, especially linear compounds. Nonanal (16.5%-38.2%), (Z)-3-hexenyl acetate (5.2%-41.8%), octanal (3.5%-12.6%), 1-octanol (5.7%-8.8%) and heptanal (3.1%-9.5%) were identified as major compounds. The two main components of the honey volatile fraction (phenylacetaldehyde and benzaldehyde) were detected in low concentrations in the flowers. Moreover, methyl syringate (a marker of asphodel honey) was not detected in the flowers analyzed. This result showed that a direct relationship between the volatile fractions of asphodel flowers and the corresponding "spring" honeys could not be established using HS-SPME analysis.

Correlation of Melissopalynological and Chemical Data
To identify relationships between the melissopalynological analysis and volatile composition data of honey samples, CCA was applied on the matrix linked the relative amounts of the 17 volatile compounds (previously used in section "Chemical variability of Corsican "spring" honeys") and the relative frequency (explanatory variables) of eight nectariferous taxa (T7-T9, T11, T13, T14, T18 and T22).
The sample distribution showed the occurrence of two main groups, group I (17 samples: 1-17) and group II (24 samples: 18-41), which correspond to the groups defined in "Determination of geographical and botanical origins of Corsican "spring" honeys" Group I was characterized not only by the significant presence of lilac aldehyde isomers (C25, C27 and C28), p-menth-1-en-9-al isomers (C31 and C32) and methyl anthranilate C38, but also by the high abundance of taxa: Citrus sp. T18, Echium sp. T13 and Prunus form T11 (group I: 6.3%, 9.5% and 3.8% versus group II: 0.1%, 4.9% and 0.8%, respectively). According to the literature [15,16,[30][31][32][33], all these compounds had been considered as characteristic components of citrus honey. From these results, it appeared that the other nectariferous taxa Echium sp. and Prunus form displayed a polleniferous role in these honey samples.

Figure 2.
Correlation between melissopalynological and volatile data of "spring" honey by canonical correspondence analysis (CCA).
Variables: taxa number corresponding to those of Tables 2 and 3; volatile components number  corresponding to those of Table 3. In the CCA plot, location of each sample indicated its compositional similarity to each other; volatile components locations indicated the similarity of their distribution to each other; length of taxa indicated the importance to the ordination, and the direction of taxa vector indicated its correlation with each axes. The perpendiculars drawn from volatile components to taxa give approximate ranking of volatile components response to the taxa variables.
Analysis of the volatile fraction of "spring" honeys also demonstrated the existence of two main groups in this range. The volatile fractions were often characterized by high amounts of phenylacetaldehyde, benzaldehyde and methyl-benzene. However, the chemical composition of "clementine" honeys was dominated by three lilac aldehyde isomers that were absent in the "not-clementine" honeys. The statistical analysis showed clearly that the "clementine" honeys were characterized by high volatile content (total peak area), methyl anthranilate, lilac aldehydes, p-menth-1-en-9-al isomers and some cultivated taxa, while the "not-clementine" honeys were characterized by phenylacetaldehyde, methyl syringate and complex taxa associations. The richness of Finally, it appeared that melissopalynological analysis was necessary for the certification of geographical origin and was useful for the determination of botanical origin. Moreover, analysis of the volatile composition could be used to specify the characteristics of volatile compounds in relation to the predominance and/or complexity of botanical origins of the product, especially when nectariferous species have an "under-represented" pollen type in the pollen spectrum, such as Citrus sp. or Asphodelus sp.