Introduction

Chrysanthemum morifolium Ramat., a perennial plant belonging to the Asteraceae family, is one of the most important ornamental cut flower crops in the world (Teixeira da Silva 2004). In Korea, chrysanthemum is an economically important ornamental with a cultivation area of 309.1 hectares in 2019 (MAFRA 2020). In addition to ornamental importance, the flower of C. morifolium has functional properties such as antioxidant, anticancer, and anti-inflammatory activities and has been widely used as traditional herbal tea and medicine (Lee and Hwang 2011).

Previous studies indicated that the volatile compounds of chrysanthemum and its wild relatives possess antioxidant activity and antibacterial activity (Jang et al. 2010;Jung 2009;Woo et al. 2008). The aroma compounds in chrysanthemum cultivars and wild relatives contain camphor, α -pinene, chrysanthenone, safranal, myrcene, eucalyptol, 2,4,5,6,7,7ab-hexahydro-1H-indene, verbenone, β-phellandrene and camphene (Sun et al. 2015). Recently, there have been some studies on the volatile compounds in chrysanthemum cultivars. However, despite of these studies, it is difficult to find research on the floral volatiles of cut chrysanthemum cultivars in Korea.

An Electronic nose (E-nose) has been considered as an olfactory simulation test tool and is used for the high-throughput screening of volatile organic compounds in a complex matrix (Rock et al. 2008). E-nose allows the rapid detection and differentiation between types of samples as an analytical instrument (Wi´sniewska et al. 2016). Using nano sensor array of E-nose which reflect the qualitative and quantitative changes in volatile components with pattern recognition software, a fingerprint of the volatile components can be obtained (Fan et al. 2018;Huang et al. 2011). Untill now E-nose has primarily been used in environmental monitoring, the perfume industry, pharmaceutical industry (Śliwińska et al. 2014), assessment of the shelf-life of fruits, vegetables (Benedetti et al. 2008, Plotto et al. 2008), and determining the authenticity of beverages (Aleixandre et al. 2008). However, there is little attention about analysis of fragrance pattern from flower using E-nose. These applications include germplasm discrimination (Fujioka et al. 2012;Zhang et al. 2014), distinction of flowering stages (Ray et al. 2017), and flower organ differentiation (Kim et al. 2016). E-nose analysis has not been previously reported in the differentiation of floral scents among cut chrysanthemum cultivars.

Hence, the aim of this study was to evaluate the differences in floral scent pattern among the different flowering stages, floral organs, and discriminate the floral profiles among different Korean cut chrysanthemum cultivars with different floral shapes and colors.

Materials and Methods

Plant materials

A total of 12 cut chrysanthemum cultivars were used in the study (Table 1). All cut chrysanthemums were cultivated in the experimental greenhouse of the National Institute of Horticultural and Herbal Science (Wanju, Korea). The cut chrysanthemums were collected in October during the flowering period in 2019.

Sample preparation and extraction of volatiles

Different flowering stages and floral organs of chrysanthemum cultivars ‘Black Marble’ were evaluated for the qualitative and quantitative analysis of floral scent. The flowering stages were classified into 5 stages as follows: flowering stages with (I) slightly opened bud (initial flowering stage), (II) half opened ray floret, (III) fully opened ray floret and initial opening disc floret, (IV) elongated inner ray floret and half opened disc floret, (V) fully opened disc floret (Fig. 1). Each floral sample (1 g) of different stages were taken into a vial and sealed with magnetic caps with silicon septa. The floral organs were divided into 4 samples: ray floret, disc floret, a receptacle surrounded by bracts and a flower head (capitulum). These samples were placed in HS-100 auto-sampler (CTC Analytics, France). These vials were then incubated at 40℃, and agitated at 500 rpm (on, 5 sec; off, 2 sec) for 2 min. An aliquot (2.5 mL) was taken from the head space of the vials.

To discriminate floral scents among different cultivars and to characterize the pattern of their floral scents, 12 cut chrysanthemum cultivars were collected (Fig. 1 and Table 1). 1 g of flower head was transferred into a 20 ml vial with screw cap. These vials were incubated at 50℃ on HS-100 auto-sampler and agitated at the same condition for 2 min. 2 mL of gas from the headspace of sample was used for E-nose analysis. Each sample was prepared in triplicate for the analysis.

E-nose analysis

E-nose analysis of floral scent was performed using an Alpha MOS FOX-2000 E-nose (Alpha MOS, Toulouse, France) with a sensor array. The sensor array of this device consists of 6 MOS sensors. The MOS sensors were the P and T sensors as n-type semiconductors. Each floral volatile taken from headspace was injected into the sensor chambers of the E-nose with an injection speed of 1000 μL･s^-1. The data acquisition time was 120 sec and the default time period in-between samples injections was 500 sec for sensor signals to be the baseline. Analysis was performed using a ratio (R₀-R)/R₀ from 6 MOS sensors, which is the change in the rate of resistance readings during the vapor exposure of each sensor (Kim et al. 2016). R₀ is the sensor’s resistance baseline and R is the real-time resistance (Shao et al. 2015).

Data analysis

The data were processed using the Alpha Soft program (version 12.45; Toulouse, France). The relative changes in sensor resistance ((R₀-R)/R₀), were subjected to multivariate analysis, including PCA and DFA. The (R₀-R)/R₀ and the relative aroma intensity, the distance between the center of gravity of each group, was used for the comparison of differences of floral scent detected by each sensor and scent intensity among cut chrysanthemums. And HCA was performed to analyze the comprehensive relationship.

Results and Discussion

Comparison of floral scents pattern among different flowering stages

To comprehensively characterize and compare the pattern of floral scents in different flowering stages, the measurement of sensor responses and relative aroma intensities was conducted for various during flowering stages in ‘Black Marble’ cultivar. Each sensor showed distinct sensor responses according to different flowering stages. The MOS sensors were the P and T sensors as n-type semiconductors. Each sensor can detect different volatile compounds, which P10/1 and P10/2 can detect non-polar volatiles, methane, propane, hydrogen, bonding compound and aldehydes, PA2 and T30/1 are responsive polar compounds and organic solvents including alcohol. The P40/1 is specific to fluoride or chloride and aldehydes and T70/2 detects food flavors and volatile compounds including alcohol and aromatic compounds (Alpha 1998;Siripatrawan et al. 2006). The relative electrical resistance changes value from 6 sensors was used for the evaluation of floral scent among different flowering stages because it could give the most stable result and is more robust against sensor baseline variation (Siripatrawan 2008). The largest response change among that of different MOS sensors was PA/2 sensor, followed by T30/1 sensor at all flowering stages (Fig. 2A). Therefore, these results indicate that there were the largest volatile metabolic differences corresponding to polar compounds and organic solvents including alcohol in the scent of all flowering stages. Among different flowering stages of ‘Black Marble’, the response change of stage III was the largest. On the other hand, stage I had the smallest response change among all flowering stages (Fig. 2A). During flowering stages, response change increased to stage III and then decreased to stage V (Fig. 2A). In addition to these results, the relative aroma intensity (RAI) which was extracted by alpha Soft and expressed by distance between blank (air) and sample was the highest at stage III opening disc flower, followed by stage IV and the lowest at stage I (Fig. 2B). The RAI, euclidean distance between centroid of three replicates of samples and blank can be used to compare relative floral scent intensity between samples (Huang et al. 2011). Our result is consistent with a past study that the total amount of aroma in the head space gas was the highest at flower petal opening time during the process of flower opening from Bulgarian rose (Oka 1999). In Malus ioensis, the amounts of volatile compounds released during the initial flowering stage were the highest and the total amount of volatiles during each flowering stage showed the initial increase followed by a decrease (Fan et al. 2019). Among the plant organs in species having scent, flowers produce the most diverse and the highest amounts of volatiles, which peak when the flowers are ready for pollination. Volatile emission in flowers is increased during the early stages of organ development when flowers are ready for pollination and then pollination induces the decrease in emission of floral volatiles which starts after reaching of pollen tubes to the ovary (Dudareva and Pichersky 2000;Negre et al. 2003). Consistent with previous literature, the result of RAI also showed the similar pattern with the result of relative response change in each sensor (Figs. 2A and 2B).

Comparison of floral scents pattern among different floral organs

The response change of sensors and RAI was analyzed to compare the profile and intensity of floral scents among floral organs of ‘Black Marble’ cultivar. The response change of six sensors among different floral organs of ‘Black Marble’ is presented in Fig. 3A. This data showed that the response change of PA/2 and T30/1 was stronger than the rest of the sensors. Distinct differences were observed at the response change of sensors among different organs. For all 6 sensors, the changes from disc floret were the largest among the all floral organs, followed by flower head. And that from the receptacle surrounded by bracts was weaker than different organs. RAI result also showed that the intensity of disc floret was the strongest, which is in correlation to the result of response change (Fig. 3B).

Quantitative differences of the volatiles emitted from flower organs may be important to attract pollinators (Dobson et al. 1990;Hansted et al. 1994) as well as to be used as ingredients of perfumes, cosmetics and for aroma therapeutic applications (Ramya et al. 2020). The flower of oil-bearing rose, Rosa damascena Mill. having heavy rosy scent, could produce the most widely used essential oil in perfumery and cosmetics (Erbaş and Baydar 2016). In general, petals are the main source of fragrance in many flowers and are important economically (Guterman et al. 2002). However, other floral organs, such as the sepal, stamen, and pistil, also contribute to the emission of scent in certain plant species (Muhlemann et al. 2014;Hao et al. 2014). For example, some of Araceae species produces scent in the stamen (Lewis et al. 1988). And many reports showed that distinct pollinator attractants can also be emitted by pollen (Dobson et al. 1999). Similarly, in our result, it was observed that there was the strongest response in the disc floret which bears pollen-producing stamens. On the other hand, there was weaker a response in the ray floret and receptacle.

Discrimination of the floral scents among the cut chrysanthemum cultivars

To compare the profile of floral scent among cut chrysanthemum cultivars, the response change from sensors was expressed as a radar plot (Fig. 4A). The significant response change by floral volatile compounds was observed in all sensors compared to blank control. The radar plot shows that there was a distinct profile from each different cultivar. Different response changes indicate differences in the quantity and quality of floral scent volatiles.

RAI, euclidean distance between the samples and the control, was used for comparing the scent intensities of different cultivars in cut chrysanthemums. The highest RAI value of floral scent was in ‘Ilweol’, followed by ‘Orange Pangpang’ and ‘Moon Festival’, and the lowest value was in ‘Prima Donna’, followed by ‘Snow Pop’ and ‘Gold Rich’ among chrysanthemum cultivars (Fig. 4B).

Floral scents of 12 cut chrysanthemum cultivars were investigated using multivariate statistical analysis (PCA, DFA, and HCA) of E-nose sensor data (Fig. 5). PCA is a method of multivariate statistical analysis that can be used to reduce the dimensionality of multivariate data by eliminating inter-correlations among variables and identify the patterns in a data set (Yang et al. 2017). Referring to the results in this study, flowers were sampled at the stage III which have the strong scent to discriminate floral scents among the cut chrysanthemum cultivars. The flower head was used to reflect whole flower scent from cut chrysanthemum cultivars, respectively. A PCA score plot of E-nose response data from 12 cut chrysanthemum cultivars was displayed in a two-dimensional plot using significant principal components (PC1 and PC2), which explain 99.3% of the variation of total variation (96.0% and 3.26%, respectively) (Fig. 5A). PCA score plot showed that most samples belonging to the same cultivar were divided into the same envelope. This result presents that each floral scent of cut chrysanthemum cultivars has a distinctive characteristic from the other cultivars. However, cultivars were not grouped together according to the flower shape and the scent of cut chrysanthemum cultivars showed wide diversity in the PCA. DFA score plot was displayed in a two-dimensional plot using the first two discriminant functions (DF1 and DF2). As shown in Figure 5B, the first two discriminant functions explained 42.0% and 27.7% of total variance, respectively. In DFA plot, most cut cultivars were differentiated from each other as the score plot is similar to the PCA plot although it indicated that differences between some cultivars are less than in the PCA plot.

HCA based on the PCA for the response changes by E-nose was performed, which showed that the 12 cultivars of chrysanthemum were separated into three major clusters (Fig. 6A). The first major cluster consisted of ‘Princeling’, ‘Black Marble’, ‘Yellow Pangpang’, ‘Yellow Marble’, ‘Green Diamond’, ‘Orange Pangpang’, ‘Moon Festival’, and ‘Baekgang’. The second major cluster consisted of 3 cultivars including ‘Gold Rich’, ‘Snow Pop’, and ‘Prima Donna’, and the third major cluster was formed by ‘Ilweol’. The HCA dendrogram based on DFA of E-nose data is shown in Fig. 6B. HCA dendrogram from DFA also separated the 12 cultivars into 3 major clusters in the same manner as the dendrogram based on PCA. A slight difference between these dendrograms was observed only in branching pattern within the first cluster. This PCA dendrogram indicates that cultivars belonging to the same group have similar volatiles compared to the other groups. In HCA analysis, cut chrysanthemums were not clustered based on flower shape or color, which is consistent with the previously reported result (Sun et al. 2015).

The PCA, DFA, and HCA dendrogram showed that the 12 cut chrysanthemum cultivars could be clustered according to the profiles of floral scent. Therefore, this study showed that the E-nose and multivariate statistical analysis was able to discriminate the differences of volatile metabolites in chrysanthemum according to the characteristic of a cultivar. And we suggest that the differences of floral scent on the T30/1, PA/2 sensor obtained by E-nose could be used as metabolic marker to classify and identify the cultivar. Furthermore, this study could be applied in the development of new cultivars having strong or attractive scent as well as providing the basic data to develop the natural raw materials in cosmetics, medicine and perfume industry.