Essential Oil of Cannabis sativa L: Comparison of Yield and Chemical Composition of 11 Hemp Genotypes

Cannabis sativa L. is an annual species cultivated since antiquity for different purposes. While, in the past, hemp inflorescences were considered crop residues, at present, they are regarded as valuable raw materials with different applications, among which extraction of the essential oil (EO) has gained increasing interest in many fields. The aim of the present study is the evaluation of the yield and the chemical composition of the EO obtained by hydrodistillation from eleven hemp genotypes, cultivated in the same location for two consecutive growing seasons. The composition of the EOs was analyzed by GC–MS, and then subjected to multivariate statistical analysis. Sesquiterpenes represented the main class of compounds in all the EOs, both in their hydrocarbon and oxygenated forms, with relative abundances ranging from 47.1 to 78.5%; the only exception was the Felina 32 sample collected in 2019, in which cannabinoids predominated. Cannabinoids were the second most abundant class of compounds, of which cannabidiol was the main one, with relative abundances between 11.8 and 51.5%. The statistical distribution of the samples, performed on the complete chemical composition of the EOs, evidenced a partition based on the year of cultivation, rather than on the genotype, with the exception of Uso-31. Regarding the extraction yield, a significant variation was evidenced among both the genotypes and the years of cultivation.


Introduction
Cannabis sativa L. is an annual herb belonging to the Cannabaceae family, which has been cultivated since antiquity as a source of fiber, seed oil, food, and medicine, as well as for recreational and religious purposes [1].
It has evolved as a dioecious species, with female and male flowers on different individuals, but selection processes have led to the development of monoecious genotypes that bear male and female flowers on the same individual. Thus, depending on the intended use, the morphology of the plants varies significantly between genotypes, in terms of height, biomass, and seed yield [1,2].
Cannabis female inflorescences and leaves are covered in glandular trichomes, which are considered as bio-factories of phytochemicals [3] due to their ability to synthesize and store different secondary metabolites, of which phytocannabinoids are the best known and studied [4]. On the basis of their cannabinoid content, in particular of their cannabidiol (CBD)/tetrahydrocannabinol (THC) ratio, Cannabis sativa L. genotypes are divided into Table 1. Agronomic trial data from the years 2019-2020. Height is expressed in cm, the diameter at the collar in mm, the density as number of plants/m 2 , and the dried inflorescences yield is g/m 2 .
Sowing was performed on 19 April 2019 and 23 April 2020, with a 25 cm distance between rows on randomized blocks with three repetitions, and a plot size of 20 m 2 .
The harvest date differed according to the flower behavior and flowering time of each genotype. Generally, monoecious and early-flowering strains were harvested earlier than dioecious and late-flowering genotypes in both years ( Table 1).
The comparison between plant heights showed differences between the two years, with higher plants obtained in 2020. The same trend was observed when comparing plant collar diameters. Plant density, defined as number of plants/m 2 , and dried inflorescences yield/m 2 were significantly lower in 2020, while dried inflorescences yield calculated for single plant was by far higher in 2020 than in 2019. These differences in plant growth in the two years of cultivation could be due to the lack of rainfall in the pre-and post-emergence period of 2020, compared to that of 2019, which could have caused a reduced germination rate and, therefore, a lower plant density; conversely, 2019 was very rainy in May-June, favoring seed germination and following plant density. This rainfall trend was inverted in June; indeed, the rainy period (July-August) in 2020 could have allowed for a biomass production greater than that observed in 2019, as is clearly shown in Table 1. Finally, September 2019 was rainier than 2020, and this could have influenced the EOs' yield.

Essential Oil Compositions and Yields
The complete compositions and the yields of the essential oils obtained from the dried inflorescences of 11 genotypes of Cannabis sativa L., cultivated by CREA-CI (Italy) in an experimental farm located in Rovigo and harvested during the years 2019 and 2020, are reported in Tables 2 and 3, respectively. In total, 116 compounds were identified, representing 90.6-99.4% of the total composition.
All the EOs, with the exception of Felina 32-2019, were characterized by a predominance of sesquiterpenes, in both their hydrocarbon and oxygenated forms, ranging from 47.1% in Carmagnola-2020 (Table 3) to 78.5% in Fibrante-2019 ( Table 2). As reported in Table 4, both the hydrocarbon and oxygenated sesquiterpenes presented significant differences in all the EO compositions as a function of the genotype, the year of cultivation, and their interaction.
In 2019 EOs, sesquiterpene hydrocarbons exhibited a significantly higher relative concentration in Futura 75, followed by Fibrante and Carmagnola. Lower relative abundances were, instead, found in Bernabeo, Eletta Campana, and Felina 32 ( Table 2). In 2020 samples, instead, this class of compounds was significantly more abundant in Carmaleonte, Codimono, Eletta Campana, Fibranova, and Futura 75, while their lowest presence was detected in Uso-31 (Table 3). Furthermore, sesquiterpene hydrocarbons were significantly higher in 2020 than in 2019 for each analyzed hemp genotype, except for Carmagnola and Uso-31. Among this chemical class, the main volatile compounds were β-caryophyllene and α-humulene, in accordance with Ascrizzi et al. [9], Vuerich et al. [18], and Menghini et al., 2021 [25], who reported these constituents as being typical of hemp varieties.
Regarding the oxygenated sesquiterpenes, among the 2019 EOs, Carmaleonte and Uso-31 presented higher relative abundances (59.6 and 60.9%, respectively), whilst the lowest was exhibited by Felina 32 (30.9%). Conversely to the hydrocarbon forms, oxygenated sesquiterpenes were significantly more abundant in all the 2019 samples compared to those of 2020. These secondary metabolites are degradation products deriving from the oxidation of the corresponding terpenes due to air exposure (i.e., during prolonged storage), and are considered responsible for the antioxidant activity of many EOs [18,26]. Moreover, higher oxygenated compounds' relative abundances reflect more favorable growth conditions, as was also confirmed by the higher plant density detected in the 2019 samples. The analyzed samples presented caryophyllene oxide and humulene oxide II as the main components of this chemical class, which were also reported as the main epoxides found in the hemp varieties analyzed by Micalizzi et al. [27]. Furthermore, 14-hydroxy-9-epi-(E)-caryophyllene, reported by Ascrizzi at al. [8], was present in low relative abundances.
0.08 ± 0.00 BC;b 1 Linear retention index on a HP 5-MS capillary column; 2 Not detected; 3 Compounds accounting for at least 1.000% of the dissimilarity rate (According to the SIMPER test, see Table 4) are evidenced in bold. For these compounds, for all chemical classes, and for the extraction yield, different superscript uppercase letters (A-G) indicate statistically significant differences between the each variety; superscript lowercase letters (a,b) indicate statistically significant differences among the cultivar withdrawn on different years (see Table 3 for 2020 samples). The statistical significance of the relative abundances was established by the Tukey's post-hoc test, with p ≤ 0.05. Table 3. Complete composition and extraction yield (% w/w dry weight) of the essential oil obtained from the dried inflorescences and floral bracts of 2020 hemp samples.

Extraction yield (% w/w)
EO hydrodistillation yield (% w/w) *** *** *** and the genotype. They accounted for up to 9.3% in Carmagnola 2020 (Table 3). Overall, 9 hydrocarbons and 14 oxygenated monoterpenes were detected: among the former, αpinene, β-pinene, and myrcene were the most representative, whilst in the latter, no compounds were revealed in appreciable relative abundances. Among the 2019 EOs, monoterpene hydrocarbons were significantly higher in Carmagnola and Fibrante (2.0 and 2.1%, respectively), and the latter presented a significantly higher content of oxygenated monoterpenes as well (1.3%). Fibrante-2019 also differed from the other genotypes in terms of its higher relative abundance in limonene (0.5%), which reached 0.8% in 2020. In the 2020 samples, hydrocarbon forms were found to be more abundant in Carmagnola (1.8%), while the oxygenated ones in CS (2.3%). However, with the exception of Codimono, 2020 samples were characterized by a significantly higher relative abundance of both hydrocarbon and oxygenated monoterpenes. The amount of the compounds belonging to this class of secondary metabolites is very different from those reported by Bertoli et al. [21], Benelli et al. [10] and Nissen et al. [28], in which the main detected chemical class was represented by monoterpenes, with significant differences between monoecious and dioecious genotypes; in these papers, however, the starting material consisted of the fresh inflorescences, rather than the dried ones. Assuming that the monoterpene content decreased in the drying process and storage [26,27,29,30], the predominance of sesquiterpenes in the examined EOs might be due, at least in part, to the drying process, which might have induced some changes in the chemical composition of the starting material, such as (i) evaporation of the low boiling-point compounds [27], and (ii) the induction of oxidative reactions, as in the conversion of β-caryophyllene in caryophyllene oxide, and α-humulene in humulene oxide II [30]. Nevertheless, few of the literature studies used the dried hemp inflorescences for the collection and characterization of the EOs.
Finally, cannabinoids were the second relevant class of compounds in all the EOs, excluding that of Felina 32-2019, in which they were the main chemical group, accounting for the 53.4% of the total ( Table 2). The detected cannabinoids were cannabidiol, cannabichromene, and ∆ 9 -tetrahydrocannabinol, but the first was the most abundant, with relative percentages ranging from 11.8 to 51.5% in Uso-31-2019 and Felina 32-2019, respectively (Table 2). In 2020 samples, these metabolites were detected in a greater relative amount in Carmagnola (37.7%) than in Bernabeo (15.0%, the lowest). In general, cannabinoids were meaningfully higher in the 2020 samples, except for Bernabeo and Felina 32, but according to the two-way ANOVA (Table 4), no significant differences were present between the two years of cultivation for this chemical class, which, on the contrary, showed significant differences among the genotypes. As for the sesquiterpenes, the predominance of cannabinoids in the EOs might be imputable to the drying process, during which the decarboxylation of cannabinoid acids into their relative volatile forms takes place [30].
Nevertheless, all the detected classes of compound, excluding the already mentioned cannabinoids, presented significant differences as a function of genotype (G), year of cultivation (Y), and genotype-year interaction (G × Y), as evidenced by the two-way ANOVA (Table 4), which was also conducted on the chemical compounds, selected by the SIMPER test, which contributed to at least 1.00% of the dissimilarities in the composition between 2019 and 2020 EO samples. The statistical analysis evidenced that 16 compounds were responsible for an overall dissimilarity contribution of 32.94%, and only seven accounted for a total dissimilarity contribution above 55%. They comprised (i) the two main sesquiterpene hydrocarbons, β-caryophyllene and α-humulene, (ii) four oxygenated sesquiterpenes, caryophyllene oxide, 14-hydroxy-9-epi-(E)-caryophyllene, and the two isomers of caryophylla-4(14),8(15)-dien-5-ol, and (iii) one cannabinoid, cannabidiol: all the selected chemicals had significant differences based on the genotypes, year of cultivation, and their interaction.
The essential oil extraction yield ranged from 0.03 to 0.12% w/w for the samples collected in 2019, and from 0.03 to 0.23% w/w for the 2020 samples, where the highest yield was comparable to those already published in the literature [20]. A statistically significant difference in percentage was found among genotypes, year of cultivation, and the interaction between genotype and year (Table 4). Regarding the dependence of the EO extraction yield on the hemp genotype, several studies are present in the literature [18,19,22,28,31], while no references are available for its dependence on the year of cultivation.
For the 2019 samples, the significantly highest EO extraction yield was obtained by Futura 75 (0.12%), followed by Carmagnola (0.09%) and Eletta Campana (0.09%), while the lowest (0.03%) from Bernabeo, Felina 32 and Uso-31. Regarding the 2020 samples, instead, Codimono, followed by Eletta Campana > CS > Fibrante (0.23% > 0.20% > 0.20% > 0.16%, respectively), were the most productive cultivars, whilst Uso-31 and Bernabeo were the least productive ones (0.03% and 0.05%, respectively). For Uso-31 (chemotype V), whose inflorescences are characterized by a negligible level of cannabinoids, the lowest EO extraction yield was found irrespective of the year. A positive correlation between the accumulation of total cannabinoids and total terpenes, both synthesized in glandular trichomes [22], was already reported [32][33][34], and explains the low EO yield obtained for Uso-31. Considering both the years of cultivation, all the 2020 samples were characterized by a significantly higher EO extraction yield than the 2019 ones (Tables 2 and 3), and this could be at least partly due to the different meteorological conditions occurring in the two subsequent growing seasons, particularly regarding rainfalls and average temperature ( Table 5). Higher temperatures might, indeed, cause a higher spontaneous evaporation degree of the EO from the plant trichomes.

Multivariate Statistical Analysis
The dendrogram of the hierarchical cluster analysis (HCA) performed on the complete compositions of the EOs, extracted from all the samples of hemp of both years of cultivation, is reported in Figure 1.
The samples were distributed in two macro-clusters: A and B. Cluster A, the more heterogeneous, was divided in two sub-clusters, A1 and A2; A2 (orange) was formed of a single sample (Felina 32-2019), whilst A1 comprised two sub-groups, A1.1 and A1.  The statistical distribution of the EOs samples evidenced a grouping based on the year of cultivation, rather than on the genotype, except for Uso-31, whose EOs of 2019 and 2020 occurred closely in the same cluster.
The score and loading plots obtained by the principal component analysis (PCA), performed on the complete chemical compositions of the EOs extracted from all the analyzed hemp samples, is reported in Figure 2a  The distribution of the samples in the score plot evidenced a partition based on the year of cultivation, as the samples of 2019 were plotted in the bottom quadrants (PC2 < 0) of the score plot, and those of 2020 in the upper quadrants (PC2 > 0), with the exception of Uso-31-2020, which was placed in the bottom left quadrant (PC1 and PC2 < 0) as Uso-31-2019.
Furthermore, the grouping of the samples obtained by the PCA was comparable to that of the HCA. All the samples of the B macro-cluster, evidenced in light blue, were plotted in the bottom left quadrant (PC1 and PC2 < 0) of the PCA score plot, due to their relevant relative abundances in the oxygenated sesquiterpenes caryophyllene oxide, humulene oxide II, caryophylla-4(14),8(15)-dien-5-ol (isomer 1 or 2), and 14-hydroxy-9-epi-(E)-caryophyllene.
The samples of the A macro-cluster, instead, occupied the remaining three quadrants of the score plot, with few exceptions. In particular, the samples of the blue sub-cluster were plotted in the upper left quadrant (PC1 < 0 and PC2 > 0), while the EOs of the green cluster were positioned in the upper right quadrant (PC1 and PC2 > 0). The blue and the green sub-clusters included most of the 2020 EOs samples, excluding Uso-31-2020, as previously mentioned, and Felina 32-2020, belonging to the pink cluster, but also plotted in the upper right quadrant. The β-caryophyllene and α-humulene vectors were responsible for the positioning of the 2020 samples in the upper quadrants, whilst the discrimination between the left and the right quadrants was determined by minor sesquiterpeneshydrocarbons and monoterpenes vectors, respectively.
The only orange sub-cluster sample was plotted in the bottom right quadrant (PC1 > 0 and PC2 < 0), as three samples of the pink sub-group: this positioning was probably due to the cannabidiol vector. The other samples of the pink group were plotted very close to the bottom right quadrant: Bernabeo-2019 was plotted towards the rightmost area of the bottom left quadrant, due to its relevant relative abundances in cannabidiol, caryophyllene oxide and humulene oxide II. Although Bernabeo is reported as a CBG-chemotype cultivar, its presence in the EO was not detected: this was likely due to the lower volatility of CBG compared to CBD, its higher boiling point, and its overall higher polarity due to the positioning of its two hydroxyls on the benzenic ring, as they are less likely to engage in the formation of a pseudocyclic structure, as could, instead, occur in CBD. Finally, Futura 75-2019 and Carmagnola-2019, due to their relevant relative abundances in β-caryophyllene and α-humulene, were plotted next to each other, close to the PC1-PC2 axis intersection, on the left side of the score plot.

Plant Material
Plant material was obtained from experimental fields of CREA-CI located in Rovigo, Italy (45 • 04 43.4 N 11 • 45 57.7 E). For the two subsequent years of cultivation (2019-2020) the same fields were used, characterized by a cultural precession with straw cereals. Nitrogen fertilization was applied before sowing (40-and 54-units of nitrogen fertilizer in 2019 and 2020, respectively). Meteorological information (temperature and rainfall) was collected, and monthly averages are presented in Table 5.
The site had total rainfall during the outdoor growing season (from April to September) of 428 mm and 288 mm for 2019 and 2020, respectively. In the same period, the 2019 had an average temperature of 21.06 • C, slightly higher than those registered for the same period in 2020, when the average temperature 20.48 • C. In both years, the maximum temperature never exceeded 32 • C in the hottest months. At harvest, each plant was cut at its lower third and air-dried at ambient temperature and in the dark, to avoid photo-oxidation reactions. Material for EOs analysis included inflorescences and floral bracts, which were separated manually from stems and seeds, with a 2 mm diameter sieve.

Essential Oil Hydrodistillation
The extraction of the essential oil from the plant material was performed by hydrodistillation with a standard Clevenger-type apparatus, for 2 h. For both 2019 and 2020 samples, the hydrodistillation was carried out in triplicate, on 100 g of dried and shredded inflorescences and floral bracts, previously macerated in water under mechanical agitation (150 rpm) for 24 h prior to hydrodistillation. The collected essential oils were stored in a refrigerator until analysis.

Gas Chromatography-Mass Spectrometry Analyses and Peaks Identification
The essential oils were diluted to 0.5% in HPLC-grade n-hexane and then injected into a GC-MS apparatus. The GC-EIMS analyses were performed with an Agilent 7890B gas chromatograph (Agilent Technologies Inc., Santa Clara, CA, USA) equipped with an Agilent HP-5MS capillary column (30 m × 0.25 mm; coating thickness 0.25 µm) and an Agilent 5977B single quadrupole mass detector. The analytical conditions were as follows: oven temperature programmed from 60 to 240 • C at 3 • C/min; injector temperature, 220 • C; transfer line temperature, 240 • C; carrier gas helium, 1 mL/min. The injection volume was 1 µL, with a split ratio of 1:25. The acquisition parameters were as follows: full scan; scan range: 30-300 m/z; scan time: 1.0 s.
The identification of the constituents was based on a comparison between the retention times with those of the authentic samples, comparing their linear retention indices relative to the series of n-hydrocarbons. Computer matching was also used against commercial (NIST 14 and ADAMS 2007) and laboratory-developed mass spectra libraries built up from pure substances and components of commercial essential oils of known composition and the MS literature data [35][36][37][38][39][40].

Statistical Analyses
The percentage of dissimilarity contribution of all the compounds of C. sativa EOs was evaluated by means of the Similarity Percentage Test (SIMPER) with the Bray-Curtis distance/similarity measure. The statistical significance of the difference in the relative abundances of the compounds accounting for at least 1.00% in the dissimilarity rate of the emissions was evaluated using the F-or t-test, for compounds with equal or unequal variances, respectively. The SIMPER, F-and t-tests were performed with the Past 4.03 Software [41].
The analysis of variance (ANOVA) was carried out using the JMP Pro 14 software package (SAS Institute, Cary, NC, USA). A two-way ANOVA was carried out in order to estimate the variance components of genotypes (G; Bernabeo, Carmagnola, Carmaleonte, Codimono, CS, Eletta Campana, Felina 32, Fibranova, Fibrante, Futura 75, and Uso-31), year of cultivation (Y; 2019 and 2020), and their interaction (G × Y). Averages were separated by Tukey's b post-hoc test. p < 0.05 was used to assess the significance of differences between means.
The multivariate statistical analyses were also carried out with the JMP software package. The covariance data matrix for the statistical evaluation of the EOs composition was a 116 × 22 matrix (116 compounds × 22 samples = 2.552 data). The principal component analysis (PCA) was performed selecting the two highest principal components (PCs) obtained by the linear regressions operated on mean-centered, unscaled data; as an unsupervised method, this analysis aimed at reducing the dimensionality of the multivariate data of the matrix, whilst preserving most of the variance. The principal component analysis (PCA) was performed, selecting the two highest PCs obtained by the linear regressions: the chosen PC1 and PC2 cover 60.30 and 26.40% of the variance, respectively, for a total explained variance of 86.70%. The hierarchical cluster analysis (HCA) was performed using Ward's method, with squared Euclidean distances as a measure of similarity. Both the HCA and the PCA methods can be applied to observe groups of samples, even when there are no reference samples that can be used as a training set to establish the model.

Conclusions
The essential oils obtained from hemp inflorescences constituted a high-value derivative by-product, contributing to a more sustainable agricultural system. The present study aimed to evaluate the chemical composition and the yield of the EO obtained from eleven genotypes of hemp, cultivated and collected for two consecutive growing seasons in the same cultural conditions, to promote their employment based on their peculiar characteristics.
The results of the present study showed that sesquiterpenes and cannabinoids were the main classes of compounds of the hemp essential oils obtained from dried inflorescences. Regarding the sesquiterpenes, the hydrocarbon form was more abundant in 2020 samples, while the oxygenated one was predominant in 2019 EOs: the major compounds belonging to this chemical class were β-caryophyllene, α-humulene, and their oxygenated derivatives caryophyllene oxide, 14-hydroxy-9-epi-(E)-caryophyllene, and humulene oxide II, all typical components of hemp EO. The cannabinoids identified in all the samples were cannabidiol, cannabichromene, and ∆ 9 -tetrahydrocannabinol, but only the former was revealed in relevant percentages. The EO extraction yield ranged from 0.03 to 0.12% w/w in 2019 samples, whilst in the 2020 ones, it was comprised between 0.03 and 0.23% w/w. The obtained data have shown that both the EO chemical profile and extraction yield were significantly influenced by the genotype of the starting material, the year of cultivation, and the interaction between these two factors.