Headspace GC-MS Analysis of Spring Blossom Fragrance at Chungnam National University Daedeok Campus

Abstract

There are many types of spring blossoms on the Daedeok campus of Chungnam National University (CNU) at the area of 1,600,000 square meters. As an assignment for the class of Analytical Chemistry I for second-year undergraduate students, 2021, flower petals collected from various floral groups (Korean azalea, Korean forsythia, Dilatata lilac, Lilytree, Lily magnolia, and Prunus yedoensis) were analyzed using headspace extraction coupled to gas chromatography-mass spectrometry (HS-GC-MS) to study the aromatic profiles and fragrance compounds of each sample group. Various types of compounds associated with the aroma profiles were detected, including saturated alcohols and aldehydes (ethanol, 1-hexanol, and nonanal), terpenes (limonene, pinene, and ocimene), and aromatic compounds (benzyl alcohol, benzaldehyde). The different contribution of these compounds for each floral type was visualized using statistical tools and classification models based on principal component analysis with high reliability (R² = 0.824, Q² = 0.616). These results showed that HS-GC-MS with statistical analysis is a powerful method to characterize the volatile aromatic profile of biological specimens.

Introduction

Various types of spring blossoms appear on the Daedeok campus of Chungnam National University (CNU) at an area of 1,600,000 m², in particular, Korean azalea (Rhododendron schlippenbachii), Korean forsythia (Forsythia koreana), Dilatata lilac (Syringa oblata subsp. dilatata), Lilytree (Magnolia denudata), Lily magnolia (Magnolia liliiflora), and Prunus cherry blossom (Prunus yedoensis). A special project in the class of Analytical Chemistry I for second-year students in the 2021 Spring semester was performed to analyze the aroma and aromatic contents of these spring flowers at the CNU campus using headspace extraction coupled to gas chromatography-mass spectrometry (HS-GC-MS) and to identify the compounds that cause each flower to have a specific scent. In the class of Analytical Chemistry I, students learned about the general concepts of the analytical process, chemical measurements, statistics for experimental errors, quality assurance and calibration methods. Conventionally, qualitative and quantitative analysis methods were taught mainly with course materials (textbooks, lecture materials, etc.), and the corresponding experimental courses were designed based on these provided understandings. Unfortunately, experimental courses for second-year students had been suffering from an unexpected status due to the COVID-19 pandemic since 2020. All in-person classes had been closed for more than a year and substituted with on-line broadcasting alternatives. Therefore, this project was designed to have students experience experimental processes and methods for instrumental analysis, in specific, chromatography and mass spectrometry, in the analysis of practical specimens.

Aroma is usually composed of a complex mixture of various volatile organic compounds (VOCs), including amino acid-derived compounds, lipid-derived compounds, phenolic derivatives, mono- and sesquiterpenes.^1-4 Since a sense of smell is a physiological reaction caused by a chemical stimulus, the molecules must be significantly light enough (<300 Da) to travel to the olfactory system via airborne.⁵ Therefore, gas chromatography (GC) coupled with mass spectrometric (MS) detector is often regarded as a powerful tool for aroma compound analysis, being a technological combination of the separation power of GC with the detection and quantification capability of MS. While flame ionization detector (FID) is more reliable and sensitive for quantitative analysis, the qualitative identification power of MS could provide robust and dynamic identification of chemical composition in samples with unknown or sophisticated background matrices, such as biological specimens. Furthermore, headspace (HS) extraction is an optimal method for volatile compound analysis, in which the volatile analytes could be extracted and isolated from the contaminants in the sample matrix. Methods based on HS-GC-MS can reduce the number of sample preparation steps required prior to analysis, as well as enable automation and programmed qualitative and quantitative analysis of biological samples.^6,7

In this study, the petal samples of 6 different species of spring blossoms were carefully collected at 16 locations on the campus. All students were divided into small groups to collect the samples. The samples were pre-treated and analyzed by HS-GC-MS with the special assistance of the Core-Facility center at the Department of Chemistry, CNU. Statistical analysis was performed to understand the relationship between the identified volatile compounds to the aroma profile, and a classification model based on the GC-MS data was established to evaluate and visualize the aroma profiles according to each floral group.

Experimental details

Sample collection and Pre-treatment

A total of 62 undergraduate students in the two Analytical Chemistry I classes have participated in this project (Table S1). They were divided into 16 groups to collect the petal samples. All floral samples (Korean azalea, Korean Forsythia, Dilatata Lilac, Lilytree (aka., Yulan magnolia), Lily Magnolia, and Prunus Yedoensis) were collected from blooming flowers on the rooted plants before browning (Table S2). These samples were picked in April, 2021 across different sites on CNU Daedeok campus (Figure 1) and the exact picking time is different depending on the flowering time of each plant. The pistil, stamen, and leaves of the flowers were completely removed, and the petal samples were stored in a respective transparent close-tight Ziploc plastic bag at -20℃ before analysis.

Figure 1. The CNU campus map indicating the places where flowers were collected. The student team numbers are also indicated on the map

HS extraction conditions

The volatile organic compound profiles from the flower samples were extracted and injected into the GC-MS instrument using an automated static HS sampler (Agilent 7697A, Santa Clara, CA) at the CNU Chemistry Core Facility. The oven temperature was set at 100℃, the loop temperature was set at 110℃, and the transfer line was set at 120℃. Homogenized petal sample (1 g) was added to a 20 mL HS crimp vial with silicon septa, and equilibrated at 100℃ for 15 min. The equilibrated sample was pressurized to 3 psi prior to injection by a fill flow of 50 mL·min^-1, and the injection time was 0.5 min.

GC-MS Analysis

The extracted volatile constituents from the pre-treated sample were analyzed by GC-MS (Agilent 6890N gas chromatograph with a 5975B mass analyzer) equipped with a VF-WAXms column (30 m × 0.25 mm × 0.5 µm) filled with low-bleed polyethylene glycol (PEG) stationary phase. The inlet was kept at 200℃ in split injection mode (10:1 ratio), and the total flow rate was set at 13.6 mL·min^-1. The oven temperature was programmed as follows: initial column temperature was set at 50℃ for 2 min, then increased to 200℃ at a rate of 10℃·min^-1 and held at this temperature for 5 min (total running time 22 min). Ionization of analyte compounds was performed in positive electron ionization mode (EI+), in which 70 eV of ionization energy was acquired using built-in gain control. The MS source temperature was set at 230℃, MS Quad at 150℃, and the solvent delay was set to 2.5 min. The mass spectrometer was operated in scanning mode (m/z 40 to 359.0) with a scanning interval of 0.1 Da. Analyte compounds were identified using library search with NIST17 database, and the cut-off criteria for positive matches were set at 50% identification confidence.

Data pretreatment and analysis

The peak areas were integrated to calculate the peak area percentage (PA%) value which is related to the composition information of each compound in a sample. The PA% from each compound for each sample was calculated separately, and then averaged between each floral group to find the correlation between the compound profile and floral groups. For the multivariate data analysis, the raw chromatographic data (from RT= 2.6 min to 22 min, interval: 0.005 min) from each sample were used. Principle component analysis model was built using the SIMCA-P 11 program (Umetrics, Umea, Sweden), in which the log₁₀ transformation of the abundance from each data point was performed for normalization of the chromatographic data.

Results and discussions

HS-GC-MS chromatogram profiles of the spring blossom samples

Representative chromatograms of the HS-GC-MS analysis are shown in Figure 2. The volatile compounds identified by HS-GC-MS of each floral sample were quantified using the percentage peak area method, in which the PA% value was calculated using the integrated area of each compound divided by the total peak area of all compounds in the MS scan mode. This calculation is based on the mass balance model, in which the chemical concentration of a component is expressed as the linear sum of products of its abundance and contribution. It should be noted that while this method is simple and suitable for broad identification and quantification of unknown or complex samples, the quantitation accuracy is limited due to the differences in ionization efficiency of different compound types toward the MS detector. As observed on the GC-MS chromatogram of the petal samples by each floral group (Figure S1-S3), the identified compounds and their relative amounts in each sample varied depending on the location, which could be related to the differences in the growing environment and the blossoming state of each individual plant. However, larger sample size and time-domain monitoring should be performed to obtain a statistically viable conclusion of these effects on the volatile compound profile in each group. In summary, a total of 114 volatile compounds were observed in the spring blossom flowers samples, in which 34 compounds (29.8%) were found in at least half of the samples collected. Distinctively, homology series and isomers of straight-chain saturated and unsaturated alkanes, alcohols, and aldehydes were identified with high amounts specifically in some floral groups. Therefore, the categorization of these compounds and their amount could assist in establishing the aroma profile of spring blossom samples. A Venn diagram of the identified compounds according to each floral group was described in Figure 3a. In all sample groups, 10 compounds were commonly found and identified as tetrahydrofuran (THF), aldehyde compounds (2-methyl butanal, 3-methyl butanal, hexanal, nonanal, and benzaldehyde), sulcatone (6-methyl-5-hepten-2-one), and terpenes (α-pinene, trans-β-ocimene, and γ-terpinene). Furthermore, 11 other compounds were detected in at least 4 floral groups. These compounds mostly come from certain plant metabolization processes and define the odorous profile of flowers, which are replicated and utilized in various commercial products as antibacterial agents and perfumes.^8-10

Figure 2. Representative HS-GC-MS chromatograms of the petal samples from 6 floral sources collected in this study.

Figure 3. Statistical analysis on the volatile compound profile of the spring blossom samples. (a) Venn diagram showing the overlapping between the volatile content analyzed by HS-GC-MS of the floral groups, (b-d) Direct comparison of the relative peak area (%) of the identified compound groups: (b) saturated fatty aldehydes, (c) saturated alkanols, and (d) terpene and terpenoid compounds, according to each floral group.

Classification and comparison of volatile component profiles between each floral group

Statistical analyses were performed on the highly contributed chemical compound and compound groups between the five floral groups, in which the two magnolia samples were merged into one group, to further understand the differences between the volatile compound profile in each group. Firstly, the amount of benzaldehyde (RT = 12.362 min) was remarkably high in the sample of Prunus yedoensis, at 66.7% of the total VOC amount, while being less than 10% in the other floral groups, including azalea (9.6%), Dilatata lilac (2.1%), forsythia (1.0%) and Magnolias (0.57%). Possessing a strong sweet scent, benzaldehyde is formed under the enzymatic hydrolysis of amygdalin, a naturally occurring glycoside found in seeds and flowers of almonds, peaches or cherries. Previous reports also indicated the presence of high amount of benzaldehyde in cherry leaves or blossoms.^11,12 On the other hand, lilac aldehydes and lilac alcohol were found predominantly in Dilatata lilac, which averaged 8.2% of the total volatile contents in its samples.

Straight-chain saturated alkanes (C₁₀-C₁₆) were identified solely in Korean forsythia samples, in which their composition contribution was averaged at 6.34±0.73% of the total peak area. On the other hand, the other blossom groups showed only insignificant traces of these compounds in the volatile phase analyzed. The yellow-colored blossom, often called “golden bell”, is one of the affluent flowers in early Spring throughout the Korean peninsula. Forsythia trees are usually grown in scrubs and roadside on the CNU campus, and its blossoms frequently come in close vicinity of road surfaces and nearby motor vehicles. Therefore, we suggest that n-alkanes found in the forsythia samples were associated with the contamination from the exhaust gas of internal combustion engines. Meanwhile, saturated fatty aldehydes (SFAs), especially nonanal (C₉H₁₈O) were found in the volatile phase of all flower samples (Figure 3b). The C₅-C₁₁ homology of SFAs was found in Prunus yedoensis and Korean azalea samples, which account for 15.0 and 20.5% of the total volatile contents, respectively. Furthermore, the volatile contents of Korean forsythia samples contained up to 54.5% of C₅-C₁₂ straight-chain saturated aldehydes, which is the highest in this category. SFAs such as octanal and nonanal are highly fragrant compounds associated as the key odorants of various flowers and plants, such as rose orange¹³ or tea tree¹⁴, which could be responsible for the fruity and rosy odor of forsythia and azalea blossoms.

Saturated and short-chain alcohols such as ethanol and C₅-C₈ alkanols were found to be the major volatile components in Korean azalea samples. Floral nectars, especially in azalea and dandelion blossoms, usually contain Saccharomyces yeasts which could convert glucose and galactose into ethanol. Among all collected samples, the ethanol contents in azalea samples peaked at 42.0%, followed by Dilatata lilac and forsythia samples at 16.5 and 10.5%, respectively. Furthermore, the distribution of short-chain alcohols ranged from C₅ to C₈ according to each floral group were described in Figure 3c. C₅-C₈ alkanols contributed a total of 6.4% of the relative VOCs content in the azalea group, in which 1-hexanol was the main compound (3.6%). In comparison, 1-octanol was solely detected in Dilatata lilac samples at 3.4%. Finally, the presence of different terpene and terpenoid compounds in floral blossom samples (Figure 3d) was discovered as one of the key components to define the aromatic profiles. γ-terpinene (RT = 8.142 min) was found to be highest in Dilatata lilac samples (51.7%), followed by magnolias (3.1%), while their content in the other floral groups ranged below 1%. Contrastingly, β-myrcene (RT = 6.654 min) was found as the major terpenoid compound in magnolia samples. Furthermore, other compounds such as 3-carene, α- and β-pinene, camphene, D-limonene and α-cymene were identified at various amounts in all the floral groups. Therefore, significantly different patterns in the distribution of aldehydes and alcohol compounds in the HS-GC-MS results could assist in defining and identifying the aromatic profile of each floral type.

Multivariate statistical analysis for classification of flowers from aroma profile

Statistical data analysis techniques, specifically multivariate data analysis methods such as principal component analysis (PCA) and partial least-squares discriminant analysis (PLS-DA) were vastly employed to characterize and differentiate the aroma profiles obtained from HS GC-MS data.^15,16 These powerful tools allow swift and robust visualization of the trends and patterns in complex datasets obtained from advanced instrumental analysis techniques.^17-19 In this study, the raw chromatographic data obtained from HS GC-MS analysis of the collected flower samples were processed using PCA to classify their aroma profiles based on their biological origins. Three major botany groups in this study (Korean forsythia, Korean azalea, and Dilatata lilac) were selected for the analysis, but the floral groups containing only a single sample were excluded (Lilytree, Lily magnolia, Prunus yedoensis).

A classification model comprise of 3 principal components (PCs) was able to explain up to 82% of the total variance, including R² (PC₁ = 0.584, PC₂ = 0.140, PC₃ = 0.087). The number of PCs was selected so that most of the characteristics of the original data were described while avoiding overfitting the model. The 2D score plots of each pair of components (Figure 4a-c) show that complete separation between the Korean azalea and other groups was achieved using the first two PCs, while the additional figures of the third component PC₃ could provide the differentiation between the GC-MS data from Dilatata lilac from the other floral groups. Interestingly, the score plot of PC₂ versus PC₃ showed a complete separation of the three groups despite the low amount of variance covered. The corresponding loading plot of the PC₂/PC₃ pair (Figure 4d) reveals the contribution of the aroma profile to the observed group separation. As observed in this plot, signal intensities from highly volatile, low molecular molecules (elution time < 6 min) were located at the low end of PC₂ which contributes to the location of Korean forsythia and Dilatata lilac groups, while the intensities from high boiling point molecules (elution time > 14 min) were highly associated to the location of Korean azalea samples. Furthermore, a complete separation of the three floral groups was observed in the 3-D score plot of the model (Figure S4), and the partial separation of elution time points into different layers in the 3-D space was also noted on the loading plot with three components as well. Therefore, it can be deduced that the major differences in the aroma profile of the three floral groups are based on the proportion of the high-volatility molecules (straight-chain alcohols and aldehydes) compared to the low-volatility molecules (branched alcohols and aromatic derivatives). These results show that HS-GC-MS analysis combined with statistical data analysis is a powerful tool to study and characterize the fragrance compounds of blossoms, which is useful for many applications such as perfume or personal care products.

Figure 4. PCA analysis of the GC-MS data, showing the classification of the aroma profile according to the major flower sample groups: (a-c) score plot of the first 3 principal components (PCs) covering 82% of the total variance, and (d) loading plot of the PC₂-PC₃ pair, colored according to the elution time in GC-MS.

Conclusions

With the collaboration of 62 undergraduate students in the Analytical Chemistry I class, floral petals from Korean Azalea, Korean Forsythia, Dilatata Lilac, Lilytree, Lily Magnolia, and Prunus Yedoensis were collected in April 2021 at CNU Daedeok campus. The flower samples were pre-treated and chromatographically separated to identify the fragrance compounds using HS-GC-MS. From the analysis, major volatile and aromatic compounds were identified, and these proportions were different according to the flower species. These analytical data may probably be useful for further metabolic and plant physiological analysis of landscape trees and fragrance effect depending on the botanical origin and environmental variances. Furthermore, multivariate statistical analysis using the PCA method showed remarkable differentiation between floral origins based on the raw data of HS-GC-MS analysis, which may be largely useful for many industrial applications such as perfume or personal care products.