Unifloral Autumn Heather Honey from Indigenous Greek Erica manipuliflora Salisb.: SPME/GC-MS Characterization of the Volatile Fraction and Optimization of the Isolation Parameters

For long heather honey has been a special variety due to its unique organoleptic characteristics. This study aimed to characterize and optimize the isolation of the dominant volatile fraction of Greek autumn heather honey using solid-phase microextraction (SPME) followed by gas chromatography-mass spectrometry (GC-MS). The described approach pointed out 13 main volatile components more closely related to honey botanical origin, in terms of occurrence and relative abundance. These volatiles include phenolic compounds and norisoprenoids, with benzaldehyde, safranal and p-anisaldehyde present in higher amounts, while ethyl 4-methoxybenzoate is reported for the first time in honey. Then, an experimental design was developed based on five numeric factors and one categorical factor and evaluated the optimum conditions (temperature: 60 °C, equilibration time: 30 min extraction time: 15 min magnetic stirrer velocity: 100 rpm sample volume: 6 mL water: honey ratio: 1:3 (v/w)). Additionally, a validation test set reinforces the above methodology investigation. Honey is very complex and variable with respect to its volatile components given the high diversity of the floral source. As a result, customizing the isolation parameters for each honey is a good approach for streamlining the isolation volatile compounds. This study could provide a good basis for future recognition of monofloral autumn heather honey.


Introduction
Honey bees (Apis mellifera L.) are primary pollinators with an important role in ecosystem conservation [1], offering many services and products, such as honey. Honey through the centuries has always been a vital food for humans, with many health properties [2,3]. The Mediterranean region, specifically Greece pronounces a set of several common and rare monofloral honeys in international markets [4]. Additionally, nowadays few rare honeys, like heather, have become increasingly well-known for their special characteristics and have received several awards in national and international food quality or taste competitions [5]. The term "heather" is used for plant species belonging to Erica and Calluna genera. However, this term is used to describe the honey produced from Calluna vulgaris (L.) Hull and not from other Ericaceae botanical sources [6]. In relation to honey from common species, including Erica arborea L., Erica carnea L., and Erica cinerea L., the given names are "Tree heath", "Spring heather", and "Bell heather", respectively [6].
The analyses of volatiles were carried out to 25 honey samples provided directly by Greek beekeepers. Samples were produced during the 2019-2021 harvest period. The botanical origin was assessed by the beekeepers and then confirmed by melissopalynological [22], and physiochemical analysis [23,24], as previously described [25]. Floral origin was confirmed firstly according to European [26] and secondly according to the more strict Greek [27] legislation (sum of fructose and glycose not less than 60% w/w; sucrose content not more than 5% w/w; moisture content not more than 20% w/w; electrical conductivity not more than 800 (µS cm −1 ); diastase activity (Schade scale) not less than 8; HMF not more than 40 mg kg −1 ; heather pollen not less than 45%). Honey samples were kept in the dark at 4 • C in hermetically closed glass bottles until further analysis.

Experimental Design
A central composite design (CCD) was used combined with RSM methodology by Box and Wilson [28]. A flexible design structure was constructed to accommodate a custom model, with numeric and categorical independent factors and irregular constrained regions. Five numeric factors (A, B, C, D, and E) and one categorical factor (F) were analyzed by a quadratic design domain. A total of 38 runs were determined by a selection criterion chosen during the experimental design (Table 1).  The responses of the volatile compounds expressed as chromatographic area (%) were used as dependent variables. For this purpose, a randomly selected sample was used for response prediction. The model's fitness was confirmed by analysis of variance (ANOVA) and the determination coefficient (R 2 ) using p-values. Dependent variables were also confirmed by the Box-Cox, correlations, and normality of residuals statistical tests. All possible optimized solutions, for (a) volatile profile and (b) each volatile molecule separately were evaluated by maximizing desirability indices. The robustness of the model was validated with response data of 24 samples according to the optimum SPME solution.

Isolation and Analysis of Volatile Compounds
Isolation of the volatile fraction was done based on experimental design layout run (Table 1) using a manual holder with triple-phase divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS) fiber 50/30 µm (Supelco, Bellefonte, PA, USA) with length of 1 cm. Before each analysis, fibers were conditioned at 270 • C. Moreover, a blank sample was performed for cleaning from previous volatile residues. Then, a predetermined volume ratio of water: Honey (v/w) was transferred in 15 mL screw top (22.7 × 86 mm) vials with PTFE/silicone septa and a portion of 20 µL (300 µg mL −1 in methanol) of benzophenone (Alfa Aesar, Kandal, Germany) was added as an internal standard.
RSM experiments were performed using a Trace Ultra gas chromatograph (GC) (Thermo Scientific Inc., Waltham, MA, USA), coupled with a mass spectrometer (MS) (DSQII, Thermo Scientific Inc., Waltham, MA, USA). GC-MS was performed with a Restek Rtx-5MS (30 m × 0.25 mm i.d., 0.25 µm film thickness) chromatography column with helium as carrier gas at a 1 mL min −1 rate. The chromatography conditions and temperature program have been previously described [29]. In brief, the GC inlet temperature 260 • C in the splitless mode for 3 min, with a 0.8 mm injector liner (SGE International Pty Ltd., Ringwood, Australia). Oven temperature was adapted to 40 • C for 6 min, then increased to 120 • C at a rate of 5 • C min −1 , followed by an increment of 3 • C min −1 up to 160 • C and up to 250 • C with a step of 15 • C min −1 . Finally, the temperature of 250 • C was kept constant for 1 min. The transfer line and injector temperatures were maintained at 290 and 220 • C, respectively. Electron impact was 70 eV, and mass spectra were recorded at the 35-650 mass range.
The peak identification was achieved with the Wiley 275 mass spectra library, and the arithmetic index provided by Adams [30]. Retention Index (RI) values of volatile compounds were calculated using n-alkane (C8-C20) standards (Supelco, Bellefonte, PA, USA). The isolated compounds were semi-quantified against the internal standard (benzophenone) and expressed as mg kg −1 of honey. All samples were analyzed in triplicate.
Esters have been encountered almost always in all blossom and honeydew honeys with some of them being dominant volatiles [18]. In our results, most esters were detected in small amounts except for methyl nonanoate. However, methyl nonanoate has been reported at much higher concentrations in honeydew honey, like fir and pine [29]. This presence could occur in the collection period of pine honey by the bees in October. Moreover, methyl octanoate and methyl dodecanoate can be related to the above conjecture [29]. Ethyl 4-methoxybenzoate was a derivative coming from p-anisic acid which has been reported in Erica arborea L. honey [10] by a Likens-Nickerson steam distillation (L-N) isolation technique. However, ethyl 4-methoxybenzoate was worth studying as it has not been detected in other botanical sources yet.   Hydrocarbons were detected in most samples with undecane having the highest average compared to the rest. This class of volatiles is very common among honeys [18].
Other studies refer to hotrienol, cis-linalool oxide, and 2-phenylacetaldehyde as the main volatile compounds of Erica spp. honey [13,17,19]. However, this is not confirmed by our samples. Hotrienol was not detected in any of our samples. Oxide of cis-linalool was linked with hive atmospheres or combustion of wood/vegetation during beekeeping activity [32], and 2-phenylacetaldehyde also had been reported in relevant concentrations, while some studies attribute this molecule to long-term storage by enzymatic catalysis of phenylalanine or heat treatment [33]. Furan derivatives identified in some of our samples, emanate from thermal processing and/or prolonged storage [34][35][36] and cannot be related to honey botanical origin.

Optimization of Each Dominant Volatile Compound
Several SPME conditions (A: Temperature; B: Equilibration time; C: Extraction time; D: Magnetic stirrer velocity; E: Sample volume; F: water: honey ratio) were investigated to determine the most suitable conditions for each volatile compound. A total of 13 volatile compounds were chosen for optimization (responses R1-R13) ( Table 3). These compounds were selected as they constitute dominant and characteristic responses of autumn heather honey. Prior to undertaking the processing steps, data for each volatile compound were confirmed by normal distribution, Box-Cox test, determination of coefficient (R 2 ) and ANOVA ( Table 4). The condition number of coefficient matrix (<10) did not indicate multicollinearity. Additionally, all responses followed the normal distribution. Box-Cox test provides a guideline for selecting the correct power law transformation. If the 95% confidence interval around this lambda includes 1.00, it does not require a specific transformation. Table 4 shows the lambda values at the 95% confidence range, as well as the current lambda. R-square (R 2 ) constitutes a measure of the amount of variation around the mean explained by the model. The ANOVA in this case confirms the adequacy of the model (p-value < 0.05) and indicated whether the model terms were significant. Significant model terms may have a real effect on the response.
ANOVA results showed many considerable independent SPME conditions, while some of them could contribute in combination. At the same time, equations were developed in terms of coded factors that can be used to make predictions about the response for given levels of each factor. By default, the high levels of the factors are coded as +1 and the low levels are coded as −1. The coded equation is useful for identifying the relative impact of the factors by comparing the factor coefficients. However, these equations should be considered with caution because it is not safe to use them as panacea for modeling future responses. In this case, all these results are presented in Table 5.   After these steps, optimization models were developed based on each volatile molecule. The results are presented in Table 6 and were evaluated by desirability indices. A high level of ideal cases is coded as 1 and low level as zero. A predicted mean for each volatile response is also included. Experimental findings showed that optimum conditions of some volatiles required the maximum value of model terms. However, this conclusion is overturned by extraction time and magnetic stirrer velocity. As previously described, extraction time was a significant parameter, along with magnetic stirrer velocity, which in some cases, allowed better isolation of some compounds [20], whilst usually shortened the equilibration time.
Due to the nature and molecular structure variability of autumn heather honey volatiles, compounds with short equilibration time can be displaced from the headspace of the vial gradually, by compounds with higher equilibration time. However, it is not a factor that significantly affects the efficiency of the system [14], since it reacts with other parameters at a relative moderate impact, as observed from the ANOVA test (Table 5). Extraction time is critical for the sample to establish equilibrium with the SPME fiber coating. A typically 15 min extraction time showed a good performance. This parameter depends on the interactions of the molecules and can be lower when using the headspace technique and interact with high concentration samples. At the same time, it should be considered that many volatiles encumber the overall sensitivity and drive the specific compounds out of the fiber, which is easier to happen at prolonged extraction. On the other side, some studies indicated that the efficiency increases together with the extension of extraction time [14]. Obviously, the assessment is very difficult due to the intrinsic variability of honeys.
The magnetic stirrer speed, as previously referred to by Xagoraris [20], was confirmed to interact with most of the responses. However, this contribution is not always important as shown by the coded equation. Although one can assume that compounds with lower volatility require greater velocities, and in our case, 100 rpm gave satisfactory results. Thus, maintaining consistent agitation improves the accuracy and precision of the system.
The optimum sample volume was 6 mL, while this parameter had minimum impact on the isolation. Nevertheless, sensitivity is better when the headspace volume is small and fiber extracts faster compounds with higher volatility [37].
Finally, the ratio of water: Honey was proven a significant parameter. Honey is highly viscous and the addition of water enables the sample agitation, while also water evaporation drifts more easily the volatiles from honey. However, the excessive addition of water tends to dilute the honey reducing the concentration of specific molecules. The most favorable water: Honey ratio was 1:3 (v/w).
Reviewing the literature, in other botanical origin honey samples, different optimization conditions are reported. Ceballos [38] suggested the analytical conditions of the optimized SPME method by RSM as 60 µm PDMS/DVB fiber, 6 g honey, 3 mL water, and 20% w/w sodium chloride, 20 min for thermostatic time, 30 min for extraction at 60 • C. Plutowska [14] referenced the following conditions: CAR/PDMS/DVB fiber, 5:1 w/w honey to water ratio, 3 g sample, equilibration time 10 min, extraction time 30-60 min, and temperature at 60 • C. Bianchi [39] examined four different sets of conditions for volatile fraction isolation from thistle honey. Bianchin [40] proposed a new optimization strategy based on the use of three different extraction temperatures (60, 40, and 30 • C) followed by equilibration time (60, 36, and 6 min), respectively, in a single assay. Robotti, [41] reported as optimal extraction conditions for multi-floral honeys (extraction temperature: 70 • C; extraction time: 60 min; salt percentage: 27.50% w/w). Da Costa [42] reported the optimum condition for extraction of volatile compounds were as follows: equilibration time of 15 min, extraction time of 45 min, and extraction temperature of 45 • C.
On the basis of the above considerations, it is difficult to predict the factors that can affect the isolation of each volatile compound of honey. In a previous study, the optimized combination of isolation conditions of thyme honey was different with temperature (60 • C), equilibration time (15 min), extraction time (30 min), magnetic velocity speed (700 rpm), sample volume (6 mL) and water honey ratio (1:3 v/w) [20]. Each honey has its unique composition thus requiring different optimization conditions regarding volatile compounds.