Introduction

Maxillaria is a large genus of morphologically diverse neotropical orchids approximately 400 species (Dathe and Dietrich 2006; Davies et al. 2005) . Many emit strong vanillaor coffee-like scents, which are responsible for pollinator attraction (Flach et al. 2004). M. tenuifolia Lindl. called ‘coconut orchid’ for its consistently strong coconut-like scent, was recognized as the best scented orchid in the 18^th World Orchid Conference (Perraudin 2006). Diversity in floral scent has evolved to attract the pollinators (Armengol et al. 2015). Today, floral scent is an important trait. Industry drives the competition for flowers with novel scents, especially orchids that have not been smelled by the human nose for a long time (Kim et al. 2015; Park et al. 2014). In this respect, Maxillaria spp. leads valuable fragrance industry. However, despite the importance of scent, the study of distinct floral scents is limited. Floral scents are invisible and variable and it is difficult to investigate scent production/emission. Sometimes the quantity of scent produced renders the scent impossible to detect by human olfactory organs. When a species is identified as having an ideal scent, plants are further cultivated for scent. In this process, the scent of the original plant may be bred out and lost (Xiang et al. 2007). M. is difficult to research because of the number of species and their varied size (Dathe and Dietrich 2006). Most studies of M. fragrances have focused on determining the chemical composition of floral scent, instead of just the compounds detected by the human nose (Dudarecva and Pichersky 2000; Flach et al. 2004; Perraudin et al. 2006). Headspace sampling (HS), which can analyze solid samples directly, is a useful method to overcome this limitation (Stashenko and Martinez 2008). Other ways to investigate fragrance include sensory evaluation and gas chromatography/mass spectrometry (GC/MS). Sensory evaluation can reveal perceptible details about scent strength and differences, but it is limited by subjective analysis and requires the skill of a trained panel. GC/MS can determine the compounds and ratios involved in fragrance, but it cannot express interactions between each fragrance compound (Lee et al. 2003a). In the last decade, HS using an electronic nose (E-nose) has become influential in the analysis of scents and has been used to identify essential oils from scissile olfactive families and to assess the stability of perfumes (Branca et al. 2003). The advantage of the human sensory E-nose, like the human sensory system, is that it receives a signal and then integrates the data to allow for classification (Baldwin et al. 2011). Using this system, M. species and cultivars were analyzed to profile scent pattern and strength. This research establishes baseline data for scent analysis of M. and for the breeding of aromatics in orchids to promote floricultural industry.

Materials and methods

Plant materials and collection of volatile compounds

A total of 11 species and cultivars were analyzed (Fig. 1): M. tenuifolia, M. houtteana, M. porphyrostele, M. variabillis, M. variabillis ‘Alba’, M. variabillis ‘Orange’, M. variabillis ‘NxO’, M. variabillis ‘Brown’, M. variabillis ‘Red’, M. variabillis ‘Nana’, and M. sanguinea. All samples were purchased from Lee Won Orchid Nursery (Gimpo, Korea) and it were cultivated in the experimental greenhouse of the National Institute of Horticultural and Herbal Science (Wanju, Korea).

The experiment was conducted between January and May during the flowering season. The flowers were sampled at 13:00 on a clear day, at 200 ± 6 μmol·m⁻² s⁻¹, with a temperature of 25 ± 2˚C and relative humidity of 20 ± 3%. For each sample, one whole flower, except for the pedicel, was immediately sealed in a 10 mL vial. Three biological replicates from each cultivation were used for pattern analysis. Flowering characteristics research was conducted on each species and the cultivars (Table 1). The flowering data includes on each flower width/length, flower color, and leaf width/length, and scent sensory test results which was carried out in the same sampling condition.

Pattern analysis of flowering stages and floral organs in M. tenuifolia

M. tenuifolia, which exhibits the strongest and sweetest scent to the human nose among all M., was subjected to a comprehensive analysis of flowering stages and floral organs to determine the pattern and intensities of the scent. Three replicates from each sample were used. Flowering stages are divided as follows: the closed flower bud (Ⅰ); initial flowering stage (II); half flowering lip out of sight (III); full flowering (IV); loss of pedicel color (V); and wilting flower (VI). The floral organs were divided into four parts: p etals, s epals, l ip, and column ( Fig. 2) .

E-nose system analysis

All samples were analyzed using an E-nose (αFOX 2000, France) equipped with 6 metal oxide semiconductors (Alpha-MOS, France). These sensors reflect the concentration of specific compounds. Non polar volatiles are detected by P10/1 and P10/2, organic solvents by PA2 and T30/1, fluoride or chloride by P40/1, and food flavors and volatile compounds by T70/2 (Alpha MOS. 1998). In this system (Table 2), a dry air carrier gas flow was set to a constant flow of 150 mL/min. Then, vials were incubated for 2 min, heated to 40°C, and agitated at 500 rpm. Under these conditions, volatiles migrate into the headspace and are sent to the sensor via an auto sampler.

The data were processed using the statistical program Alpha Soft Version 12.45. Analyses were conducted using the chemical sensitivity of MOS (Delta R_gas/R_air), the change in the rate of resistance readings during the vapor exposure of each sensor (Park et al. 2014; Pearce et al. 2002). The interpretation of the complex data sets from the E-nose signals was accomplished using multivariate statistics, including principal component analysis (PCA) and discriminant function analysis (DFA), to reduce the high dimensionality of this multivariate problem (Baldwin et al. 2011). Intensity, which refers to the distance between the center of gravity of each group, such as the control (air) and a sample, is a convenient method to assess similarity between species and cultivars (Alpha MOS. 1998).

Results and discussion

Floral scent pattern of cultivars

Flowers have enormous diversity in their shapes, sizes, colors and scents. In over 900 flowerings plant species, more than 1,700 volatile compounds have been identified (Knudsen et al. 2006). These volatile compounds are usually specific to species, though the quality and quantity may vary (Raguso 2008). We investigated the scent pattern difference between species and cultivars, using the statistical methods PCA and DFA. It is difficult to discern the scent strength and distribution pattern from PCA (Fig. 3A). Analysis by discrimination index is typically performed to show how variations in a compound contribute to group separation. After the PCA evaluation, it is necessary to validate the DFA model for qualitative analysis. DFA is based on a search for vectors, along which some samples cluster together and others separate. Thus, this method is usually used to identify unknown samples (Alpha MOS. 1998). The pattern size indicated relatively concentrations among the sample (Huang et al. 2011) also x- and y- axis showed contribution in classification scent patterns (Hwang et al. 2015). A DFA score plot, which shows the differences between the 11 species and cultivars was able to evaluate 97% of the total variance and DF1 had influence on discrimination of scent patterns (Fig. 3B). Each scent pattern was different, but most patterns, with the exception of M. tenuifolia, are distributed in the center of the chart. In DFA M. tenuifolia scent pattern separated from DF1 as well as DF2. Although these species are all in the M. variabillis clade except M. porphyrostele- (Whittern et al. 2007), there are different ratios in the genetic character of the fragrances caused by cross-breeding. Note that the quadrants do not indicate any specific meaning, the overall chart results from graphs recorded by each E-nose sensor responsible for a particular part of the floral scent composition (Park et al. 2014). The pattern of M. variabillis cultivars, which are derived from M. variabillis, grouped with each other. However, the original M. variabillis species showed a quite different pattern compared to its cultivars. M. variabillis ‘Alba’ exhibited the strongest scent (Fig. 4). In Phalaenopsis, the scent patterns of six strains were similar, though minor variations were observed by Been (2010), similar to what was observed in the M. variabillis cultivar group. The distinction between M. variabillis and its cultivars may have been caused by cross-breeding. Genes that are related to in scent can transfer randomly, and the original scents may not be retained. If a species depends on one class of pollinator, members of that species have more similar scent patterns. However, studies show that dissimilar flower scents among closely related species may result from continuous variation, as interspecific similarities drift apart (Kundsen et al. 2006). The distance analysis shows that all species and cultivars have a distinct scent according to the overall sensory evaluation. Interestingly M. tenuifolia has a strong scent, and M. sanguinea is relatively unscented when tested with a human nose (Table 1), but the E-nose data do not show this trend (Fig. 4). Differences between sensory evaluation and E-nose evaluation may suggest that biological systems use countless sensors for volatiles, but the E-nose has a limited number of sensors that detect concentrations of compounds for analysis and fragrance characterization. For example, indole has an offensive smell, but when added to alcohol at concentration of 0.001% it has a scent similar to jasmine (Lee 2002).

Floral scent pattern of flowering stage in M. tenuifolia

Flowers emit the floral scents, to attract potential pollinators (Mehlemann et al. 2014). Once a flower has been pollinated, the post-pollination scent changes to prevent visits from other pollinators, which increases the chances of visitation to unpollinated flowers (Dudareva and Pichersky 2000). The scent intensity was different at each flowering stage and three days to transition from the fourth to sixth stage. The scent was strongest in the second of the six stage of blooming; the scent intensity gradually weakened as the stages progressed (Fig. 5). Our result is consistent with a previous study by Oka et al. (1999) that demonstrated that when Rosa damascene Mill flowers start to bloom, the total amount of aromas in the head space gas reached the highest level in the last stage. This result suggests that fragrance evolution from stage to has different mechanisms. Most flowers have floral fragrance expression sites on petal surfaces, so volatile compounds can escape directly into the atmosphere, and floral scents are produced upon flowering (Guterman et al. 2002; Lavid et al. 2002). The scents are then reduced by senescence and wilting (Dudareva and Pichersky 2000; Kolosoca et al. 2001). Fragrance intensity and composition as a function of flowering stage have been studied in wind orchids and lilies as well as in roses. In the wind orchid, the amount of fragrance was lowest at the bud and wilt stages and highest at full bloom by Been (2010). In Rosa hybrid ‘Orange Flash’ and ‘Blue Moon’, flowers at bloom exhibited the strongest scent by Lee et al. (2008b). Similarly, a study on lilies showed that when the flower was fully bloomed, the scent was the strongest, and the scent became weaker as aging progressed by Byun et al. (2007). However, the level of floral scents in Lilium Oriental Hybrida ‘Casa Blanca’ increased when old age began, due to increase in essential oil components by Rho and Pak (2001). This result indicates that scent production in the various flowering stages can be different between species and between cultivars.

Floral scent pattern of the floral organs in M. tenuifolia

It is difficult to determine which floral part emitted the most scent and to prove the specific composition of the scent (Dudareva and Pichersky 2000). In most flowers, petals are the main source of fragrance, and their expression is regulated developmentally (Guterman et al. 2002). Other tissues, such as the sepal, stamen, and pistil, also contribute to scent in certain plant species (Muhlemann et al. 2014; Hao et al. 2014). For example, several Araceae species produce scent in the stamen (Lewis et al. 1988). The PCA and DFA analyses showed that the scent pattern of the floral organs different slightly (Fig. 6). The pattern of each organ was distinguished by DF2 rather than DF1 which had influential in discrimination of scent patterns (Fig 6B). In M. tenuifolia, the sepal showed the greatest scent intensity, followed by the petal, column and lip in that order (Fig. 7). The different pattern of each flower part may suggest that the level of active enzymes related to fragrance is varied in different tissues (Dudareva and Pichersky 2000). A study on roses determined that the petals had five times the fragrance compounds compared to the sepal by Lee et al. (2013). In Sedirea japonicum, the sepal and petal emits almost all of the major fragrance compounds, whereas production in the column and spur is trace by Kim et al. (2015). In Phalaenopsis, the fragrance is produced at the petal and sepal, though the lip and column express fragrance to some degree by Been (2010). A study of three species of lilies by Byun et al. (2007) showed that the most diffusive organ part was different for each species, Lilium Asiatic Hybrids, it was the outer perianth, in Liliumlongiflorum, it was the pedicel, and in Lilium Oriental Hybrids, it was the inner perianth. These results further indicate that the floral organ part that exhibits the most scent can vary between species and cultivars.

These results on M. tenuifolia show differences in scent patterns compared to flowers in previous studies. The E-nose can simplify the patterns of scent compounds, but these experiments have limited in repeatability due to deviations in cultivar, rea, climate, and cultivation conditions. Additionally, there is a risk of changes in emission profiles due to autosampler use, because sampling involves removal of the flower from the plant for placement in the vial. Further studies are required to determine the degree of correlation between the E-nose system and sensory tests. It is also necessary to determine the specific volatile compounds that contribute to the concentration of fragrance analyzed by the E-nose; GC/MS can provide this type of qualitative analysis of the volatile components present in flowers.

Acknowledgments

This work was carried out with the support of “Cooperative Research Program for Agriculture Science & Technology Development (Project No. PJ011415032016)” Rural Development Administration, Republic of Korea.