Background

Herbivores feeding may trigger plant defense responses which leads to massive release of various volatile organic compounds (VOCs) (Jiang et al. 2019; Poelman et al.2012). The herbivore-induced plant volatiles (HIPVs) consist of terpenoids, green leaf volatiles (GLVs), aromatics and phenolics (Liu et al. 2020b). These volatiles can serve as foraging cues to attract natural enemies and elicit defense strategies against herbivores (Aartsma et al. 2017; Sun et al. 2014). Additionally, released volatiles from aromatic plants may potentially be used as repellents for tea green leafhoppers (Isman 2006). Therefore, the contents of HIPVs may provide critical information about the herbivory attack in the context of tritrophic interactions.

The tea plant Camellia sinensis (L.) O. Ktze is a major economic crop in China, and tea drink is also one of the three non-alcoholic beverages worldwide (Luo et al. 2019; Tan et al. 2020). Tea drink is hig2020hly rich in tea polyphenols, amino acids, tea polysaccharides and a variety of aroma components, and drinking tea has contributed largely to health benefits (Fan et al. ; Zhu et al. 2018). The incidence of pests during growth of tea plants is frequently observed especially in tea shoots. The tea green leafhopper Empoasca onukii Matsuda (Hemiptera: Cicadellidae, previously named as Empoasca vitis) belongs to a main pest for various tea producing areas in China, and is also a dominant species in southwest Anhui (Liu et al. 2020a; Meng et al. 2018). The infested tea leaves may initially scorch, curl up accompanied by browning, shriveling, and necrosis leading to decreased tea yields. Notably, there will be 7–20 generations of leafhoppers with an estimated 30–50% reduction in tea yields during seasonal outbreaks (Aartsma et al. 2017; Sun et al. 2014). Despite the predation by natural enemies in tea garden, chemical control is widely used for tea green leafhopper management. Excessive application of insecticides (chemical sprays reach 10–20 times in summer) may result in chemical resistance among leafhoppers and cause residue problems in tea. As a result, integrated strategies for pest management should be used for biological control over tea green leafhoppers (Cai, 2020; Cook et al. 2007; Pan et al. 2016).

In current work, the changes in volatile emission profile upon tea green leafhopper feeding was investigated using multivariate analyses. We found that volatile patterns in healthy and infested tea shoots could be well grouped into two clusters. Using supervised partial least square discriminant analysis (PLS-DA), several important volatiles such as 3-hexenal, α-pinene, benzaldehyde, α-phellandrene, 2-pentylcyclopentanone, acetophenone, longifolene-(v4) and E-5,9-undecadien-2-one,6,10-dimethyl- were identified. Such information can be employed to establish contamination-free regimes for pest management.

Materials and methods

Plants and insects

Tea green leafhoppers were collected from tea plant in a tea garden at the occurrence cycle of leafhoppers in Autumn 2020 at Yangting Village, Wuheng Township, Yixiu District, Anqing City from Southwest Anhui, China. The cultivar of tea plant is Camellia sinensis cv. ‘shifo quntizhong’. The leafhoppers were reared in generation on one-year-old tea plants of ‘shifo quntizhong’ in plastic pots within a greenhouse (L14/D10 cycle, 25 ± 3 °C, relative humidity: 65–75%) at Anqing Normal University.

Ten healthy tea plants were randomly selected from an experimental block (300 × 400 m in size) at a contamination-free tea garden from Yangting Village (Fig. 1). Notably, tea plants in this experimental block were raised in similar growing status. Then, simple random sampling method was used and ten random locations (sampling points) were generated using MATLAB (Fig. 1 and Table S1). The spacing between any two sampling points was at least 100 m to avoid the volatile communication between tea plants. Five tea plants were randomly selected as the infested group. For an infested sample, the 3-instar nymphs of tea green leafhoppers were reared on the seedlings with one bud and five leaves, with 20 leafhoppers per tea shoot. The leafhoppers were raised in greenhouse and tea shoots were covered with mesh fabric to avoid evasion of leafhoppers. After piercing damage for 24 h, the leafhoppers and molts were removed by brush and 45 g tea shoots with one bud and five leaves were collected. The tea shoot terminal was covered by moistened cotton ball for volatile extraction. For a healthy sample, 45 g healthy tea shoots covered with mesh fabric were used as control. Five healthy tea plants were used in the control group.

Fig. 1
figure 1

The workflow in our current work

Full size image

Volatile collection

The volatiles were collected by a push–pull aeration system with a glass cylinder (2L volume and 80 mm internal diameter) as previously described (Mu et al. 2012) (Fig. 1). The tea shoot samples were placed into the glass cylinder chamber with frosted ends. The filtered air by active charcoal was pumped into the cylinder through a caliber with a diameter of 8 mm and the air was pumped out through the other end with a caliber diameter of 8 mm. The outlet end was sequentially connected by 60 mg of Super Q absorbent, a flowmeter (240 mL/min) and a pump. After 12 h’ extraction, elution was performed with 200 μL dichloromethane. The eluting solution was kept in a 2 mL glass vial, mixed with 2 µL ethyl decanoate (10–4 g/mL) as the internal standard (IS) and preserved in 5 °C refrigerator.

Analysis of the tea shoot volatiles

The gas chromatograph-mass spectrometer (GC–MS) device from Products Quality Inspection Supervision Center at Northwest Anhui in Anqing City, Anhui Province and fitted with a HP-5MS (30.0 m × 250 µm × 0.25 µm) quartz capillary column. The column effluent was splitless (Mu et al. 2012). 1 µL sample was placed into microinjector by hand. The inlet temperature was 250 °C, and the GC/MS interface temperature was 280 °C. The temperature was initialized from 50 °C for 5 min and then increased by 3 °C /min to a final 190 °C for 5 min. The solvent delay was 3 min. The mass spectrometer was used in the EI (at 70 eV) scan mode with a scanning frequency twice/s. Helium was used as the carrier gas by a constant flow of 1.0 mL/min. The spectrum library was NIST08.L.

Volatile compounds were identified by comparing retention time with those of authentic standards, the mass spectra standards, and the related literatures (Han and Chen 2002; Sun et al. 2010). The relative contents of volatile compounds were calculated based on the total ion current peak areas divided by the peak area of the internal standard.

Silhouette value

To evaluate the similarity of samples within each cluster, silhouette value was calculated (Ozkan and Ozdemir 2015). The silhouette value is defined as:

$$S_{i} = \frac{{(b_{i} - a_{i} )}}{{\max (a_{i} ,b_{i} )}}$$

where ai is the average distance from the ith sample to the remaining points in the same cluster, bi is the minimum average distance from the ith point to all points in a different cluster.

Multivariate statistics

Statistical significance of volatile release between healthy and infested groups was determined by Mann–Whitney test. The principal component analysis (PCA) is performed in MATLAB using ‘pca’ function. The partial least square-discrimination analyses (PLS-DA) were performed using SIMCA-P software (v11.5, Sartorius Stedim Biotech, Umea, Sweden).

Results

Volatile organic compound profiles identified by GC–MS

To explore whether the volatile profiles from tea shoots were affected by tea green leafhopper feeding, the gas chromatography mass spectrometry (GC–MS) was used and experiments were performed for totally ten samples (five healthy samples versus five infested ones) (Fig. 1). The relative contents to the internal standard reached 505% (± 37%) in tea green leafhopper infested tea shoots. However, it was only 145% (± 54%) in healthy tea shoots. The volatile release from infested tea shoots was significantly elevated compared with that from healthy tea shoots (Mann–Whitney test, p = 0.0079). In healthy tea shoots, 23 volatiles were identified, whereas a total of 29 volatiles were detected from infested tea shoots (Fig. 2 and Table 1). The specific volatiles in infested tea shoots were α-pinene, decane, α-phellandrene, 2-pentylcyclopentanone, longifolene-(v4) and tetradecanal (Table 1). The 29 VOCs were classified as five categories: green leaf volatiles (GLVs), aromatics, terpenes, alkanes, and other volatiles (Table 2). We found that terpenes underwent a highest fold increase in infested tea shoots (1106%, Fig. 3A). Even for alkanes, the fold elevation in VOC expressions was 208% (Fig. 3A). We next investigated the proportion of five VOC classes in their relative abundances. A slight increase in fraction of GLVs and terpenes could be observed in infested tea shoots (Fig. 3B). Proportions of alkanes and aromatics, however, were markedly decreased (Fig. 3B). These results suggested that volatile release is enhanced when tea shoots are infested with leafhoppers. We found that differences of 17 volatiles were extremely significant (p < 0.01, Table 1). This group of 17 variables was selected as a more compact data set for further analysis.

Fig. 2
figure 2

The total ion chromatograms of the volatile components from healthy and infested tea shoots by tea green leafhopper feeding. A Healthy tea shoots; B Infested tea shoots. The numbering is the same as those in Table 1

Full size image
Table 1 The relative concentration of VOCs from healthy and infested tea shoots
Full size table
Table 2 Classification of volatile organic compounds from tea plants
Full size table
Fig. 3
figure 3

The categories and proportions of volatile compounds in healthy and infested tea shoots. A Five kinds of volatiles in healthy and infested tea shoots; B The proportion of five volatile compounds in healthy and infested tea shoots

Full size image

Clustering analysis for tea shoot volatiles

The silhouette score for each point is a measure of similarity within its own cluster, when compared to points in other clusters (Froehlich et al. 2007). A high silhouette score indicates that a sample is well-matched to its own cluster, and poorly-matched to neighbors. The silhouette scores showed that VOC profiles between healthy and infested tea shoots were well separated (Fig. S1). Especially, the VOC patterns from healthy tea shoots had higher silhouette values compared to those from infested group (Fig. S1), implying that VOC distribution was more dispersed during infestation. These data suggested that VOC profiles in healthy and infested tea shoots are well clustered.

Principal component analysis

The principal component analysis (PCA) is a strategy for dimension reduction. Then, PCA was conducted to identify important variables which contribute to variations in VOC patterns. Results suggested the first principal component (PC1) could explain 91.27% variations whereas PC2 was only responsible for 5.79% total variations. As a result, PC1 could clearly differentiate VOC profiles from healthy and infested tea shoots (Fig. 4A). We next used bootstrap approach to empirically estimate the sampling distribution for the loadings of first two components. The nonparametric approach was applied and resamples are drawn with replacement, and each resample is of the same size as the original one (Thompson 2010). 1000 resampling procedures were performed. The results showed that only a fraction of variables (VOCs) had relatively large loadings (Fig. 4B). We found that several volatiles such as 3-hexenal, α-pinene, benzaldehyde and acetophenone had relatively larger mean loadings (Fig. 4B, blue). All variables had mean loadings larger than zero suggesting that PC1 could be possibly used as the evaluation metrics (Jolliffe and Cadima 2016). The empirical eigenvalue distributions for the first two principal component again indicated that PC1 had significantly larger eigenvalues (p < 0.0001, Kruskal–Wallis test, Fig. 4C). Furthermore, bootstrapping showed that PC1 always had more contributions compared with PC2 (minimum in PC1 is ~ 77%, Fig. 4D). These results suggested that the VOC profiles for healthy and infested tea shoots could be well grouped.

Fig. 4
figure 4

Principal component analysis. A The principal component scores for VOCs from five healthy (No. 1-5, black) and five infested (No. 6-10, grey) groups. B Bootstrapping for loadings. Totally, 1000 resampling was performed and loadings for the first two principal component (PC1 and PC2) were shown. Numbering was shown in Table 1. C Bootstrapping for the first two eigenvalues. Eigenvalues were sorted in descending order. D Bootstrapping for contributions of PC1 and PC2

Full size image

Partial least square discrimination analysis (PLS-DA)

Partial least square based method is more suitable in multivariate analysis for small samples or samples with collinearity (Lee et al. 2018). Therefore, a PLS-DA model was used. The PLS-DA method resulted in a two-component (latent variable, LV) model which accounted for 97.5% variations in responses with a predictive power of Q2 = 91.8% which suggested that the model was not over-fitted. We noted that all volatile profiles of healthy tea shoots were assigned negative scores on the horizontal axis (LV1, Fig. 5A). Volatiles from infested group, however, had positive scores on LV1 (Fig. 5A). In addition, all samples were distributed within the tolerance eclipse based on Hotelling T2 statistics suggesting that there were not outliers (Fig. 5A). The biplot further showed that LV1 and volatile profiles from infested group had positive correlations with all variables (Fig. 5B). The variable importance for the projection (VIP) was then calculated and the results identified eight important variables with VIP > 1: 3-hexenal (C2, 1.1588), α-pinene (C7, 1.0616), benzaldehyde (C8, 1.0187), α-phellandrene (C12, 1.0013), 2-pentylcyclopentanone (C13, 1.0745), acetophenone (C14, 1.0902), longifolene-(v4) (C24, 1.0193) and E-5,9-undecadien-2-one,6,10-dimethyl- (C26, 1.0717) (Fig. 5C and Table 1). To ascertain whether the eight volatile features were conserved, we conducted logistic regression and orthogonal PLS-DA (OPLS-DA, Fig. S2). Although the exact order of importance for volatiles might differ, the selected feature with eight volatiles was conserved (Fig. S2). Collectively, these data showed important variables which are associated with the variations in volatile profiles.

Fig. 5
figure 5

The partial least square discriminant analysis for volatiles from tea shoots. A Scores for the healthy and infested samples in LV1 and LV2 coordinates (The eclipse represents the 95% confidence eclipse based on Hotelling T2). B The biplot in PLS-DA (The radius for solid or dashed circle is 1.0 and 0.5, respectively); C VIPs for each volatile compound (Numbering was shown in Table 1, dashed line is a guide for 1.0)

Full size image

Discussion

Herbivores are typically recognized via herbivore-associated molecular patterns (HAMPs) or damage-associated molecular patterns (DAMPs) (Felton and Tumlinson 2008; Heil and Land 2014). Increased release in HIPVs can be found when plants are confronted with herbivores (Schmelz 2015). HAMP-induced HIPVs can provide information about herbivores in attraction of natural enemies (Turlings and Erb 2018). In current work, we investigated the volatile release patterns using a novel tea cultivar from southwest Anhui after tea green leafhopper infestation.

Patterns of volatile release were dynamically changed after infestation

From the volatile release profiles, we found that the relative contents and the number of volatile components were significantly increased in infested tea shoots compared with healthy ones. In addition, the fraction of GLVs and terpenes are elevated in infested tea shoots. Notably, GLVs are released at early stage of infestation possibly owing to the constitutive expression of GLV-associated enzymes (Matsui 2006). However, terpene release is remarkedly delayed from several hours to days during infestation (Liu et al. 2020b; Pare 1999). Meanwhile, the proportion of aromatics is significantly elevated from tea shoots collected in autumn compared with those obtained in spring (Bian et al. 2014), implying that the defensive strategies against tea green leafhopper feeding might be different.

By supervised PLS-DA, several volatiles were identified as important factors. (Z)-3-hexenal can be used as an attractant for male Ectropis obliqua moths (Sun et al. 2016). Another isomer 2-hexenal could also attract bruchid beetles to host plant volatiles in the field (Wang et al. 2020). The α-pinene is an attractant for natural enemy multicolored ladybug (Harmonia axyridis Pallas) (Xue et al. 2008). Meanwhile, α-pinene can also repel the pest walnut twig beetles (Audley et al. 2020). Its isomer β-pinene could be used as a lady beetle attractant for biological control of tea plant pests (Zhao et al. 2020). The benzaldehyde acts as natural enemy-attracting semiochemicals in response to herbivore damage (Simpson et al. 2011). In addition, benzaldehyde also has a complicate attraction effect for natural enemies the coccinellid (Coccinella septempunctata), the parasite (Aphidius sp.) and the lacewing (Chrysopa sinica) (Han and Chen 2002, 2010). The α-phellandrene has been recently identified as an attractant for natural enemy Encarsia formosa (Hymenoptera: Aphelinidae) (Ayelo et al. 2021). Remarkable increase in volatile blends including α-phellandrene is observed in infested Pinus massoniana Lamb (Li et al. 2017) suggesting that α-phellandrene might be implicated in plant defense responses. A 2-pentylcyclopentanone analog cyclopentanone is a major component of volatile cocktail for attracting woodwasp suggesting that it might be used as pest attractant (Wang et al. 2019). The acetophenone can significantly decrease the fraction of predated patches against herbivore attack (Kergunteuil et al. 2012). Longifolene-(v4) emission is usually increased in infested plants (Zong et al. 2017) and serves as lures to wood-boring wasp, Sirex noctilio (Xu et al. 2019). E-5,9-undecadien-2-one,6,10-dimethyl- is involved in volatile aroma compositions in tea plants (Fan et al. 2020). Therefore, the combinatorial emission of volatile blends may provide multiple strategies against herbivory attack.

Volatile patterns provide robust responses against herbivores

Notably, the types and relative contents of volatiles in infested tea shoots are substantially increased. By unsupervised (principal component analysis) and supervised (PLS-DA) analyses, volatile profiles from healthy and infested tea shoots are well classified into two clusters. Given a priori knowledge of infestation status (i.e. healthy and infested tea shoots), supervised PLS-DA method may provide more accurate estimation (Lee et al. 2018). From PLS-DA results, we noted that volatile profiles from healthy tea shoots are distributed in the third quadrant. On one hand, the volatile profiles from infested tea shoots are more dispersed. Therefore, the response of tea plants to tea green leafhopper feeding might be multidirectional. On the other hand, we can find that the identified important factors also serve as analogs or isomers of odor trap stimulants or natural enemy attractors. For example, 3-hexenal is the isomer for synomones 2-hexenal and attractant (Z)-3-hexenal. Cyclopentanone is the analog of 2-pentylcyclopentanone. The β-pinene, an isomer for α-pinene, is also an attractant for biological control of tea plant pests (Zhao et al. 2020). These might be reminiscent of ‘robustness’, a property which allows a biological system to maintain its functional states despite perturbations (Kitano 2004). Robustness can be obtained if multiple strategies can be combined to achieve a specific function (Kitano 2004). ‘Heterogeneity’ refers to a situation in which specific functions could be acquired by diverse means. The coexistence of multiple analogs to synomones implies that the attraction to natural enemies can be achieved in diverse ways. This diversity might enhance the foraging cues to natural enemies or retain the attraction when synthesis of certain volatiles is compromised. Therefore, the volatile release profiles of tea green leafhopper infestation might be a representation of biological robustness.

Conclusions

Eight potentially important volatiles for plant responses against infestation were identified in current study. Owing to the small sample size, this might be only a preliminary model for discrimination. Increasing the sample size will be required to achieve a more accurate model. Meanwhile, dynamic measurement may provide additional information. Whether all the identified important volatiles can serve as synomones should be investigated in future.