A sprayable solution containing adhesive RGO droplets was produced at Wageningen University & Research as described by van Zwieten et al. (2024). The solution contained 95.75 wt% tap water, 1.25 wt% oxidized RGO (type: R360) droplets, 1 wt% F-108 (a surfactant), and 2 wt% alginate. This RGO solution and a water control were tested in plant assay 1. The RGO solution and four control solutions: a 1% F-108 solution, 2% alginate solution, and a solution containing both 1% F-108 and 2% alginate, and a water control, were tested in plant assay 2.

A colony of Frankliniella occidentalis (obtained from a greenhouse in the Netherlands) was maintained in plastic cages (60×60×40 cm) at Leiden University for over 15 years on Chrysanthemum cut flowers (cv. Baltica Yellow) as described by Bierman et al. (2024). Climate room conditions were 25°C, 60% RH (70% inside cages) and 16-8 light-dark photoperiod (fluorescent TL-light).

Ten-day old Chrysanthemum × morifolium (cv. Baltica White) cuttings, rooted in small peat blocks (source: Deliflor B.V., Maasdijk, The Netherlands) were transplanted to 2.2 liter plastic pots filled with a mixture of 3:1 (v/v) autoclaved potting soil to vermiculite, with 2.5 g/l osmocote fertilizer pellets. Plants were grown in a climate room at 25°C day, 23°C night temperature, 70% RH and a 8-16 light-dark photoperiod (6:00am-10:00pm, fluorescent TL-light, 15,340 lm m^-2). When plants were 25 days old, plant assays were performed at Leiden University under the same climate conditions as during the plant growth phase.

Eighty Chrysanthemum plants of similar size (25 days old, 10 unfolded leaves, 14.3 cm [SD: 0.99] shoot length) were selected and divided into eight groups of ten plants. 40 plants were sprayed with 10 ml water per plant and 40 plants were sprayed with 10 ml solution containing RGO droplets using a paint spray gun (Einhell, Art. No. 41.330.00). Spraying was done from 20 cm above each plant under 1.4 bar airflow pressure. The spray pattern control valve was opened a quarter rotation. The flow control valve was opened two full rotations. Plant coverage was estimated from pictures of several leaves of five randomly chosen plants using ImageJ-Fiji V2.9.0 by counting adhesive droplets and by estimating the percentage leaf area covered ( Supplementary Figure S1 ). After 30 minutes of drying, plants were placed on felt circles located on plastic dishes (25cm diameter) and covered with thrips-proof cages made of transparent plastic cylinders (50 cm height, 22.5 cm diameter) that were closed off at the top with lids made of plastic rings with 120 μm thrips-proof nylon mesh (Leiss et al., 2009a). Felt circles were kept moist during the experiment to water the plants and to prevent thrips from escaping. Per spray treatment, 20 plants were infested with 20 female adult F. occidentalis at the base of each plant. One group of ten plants of each combination of spray treatment and thrips was harvested after 10 days, the other ten plants after 25 days. The cages with plants were placed in the climate room in ten blocks containing one plant of each treatment and harvesttime ( Supplementary Figures S2A, B ).

At harvest, the number of thrips caught in adhesive droplets, total thrips damage and necrosis damage (brown spots) on all leaves were scored in mm² and the shoot length (cm) and number of leaves were recorded. Then, for each plant a stack of sprayed leaves, leaf number 2 to 9 (counted from the bottom and up) was sampled in aluminum foil, flash frozen between 7 to 9 pm and stored at -80°C until freeze drying. After freeze drying, samples were ground into a fine powder using 2 ml Eppendorf tubes with 5 tiny iron balls and a Tissuelizer II Bead Mill (Qiagen, Hilden, Germany). The leaf powder would be used for ¹H NMR and headspace (HS)- GC-MS analysis. The remainder of the shoot of each plant was oven-dried at 60°C to obtain the dry weight (g).

Five groups of ten plants (25 days old, 11 [SD: 0.83] unfolded leaves, 15.5 cm [SD: 1.57] shoot length) were sprayed with 10 ml of one of five treatments: (1) water, (2) 1% F-108 solution, (3) 2% alginate solution, (4) 1% F-108 + 2% alginate solution, or (5) the full solution containing adhesive RGO droplets. Coverage (no. droplets and % leaf area covered) was estimated as before. Each plant was placed in a thrips cage, infested with 20 female adult thrips, and placed in the climate room in ten blocks with one plant per treatment. After 25 days, the number of thrips caught in adhesive droplets, total thrips damage (both on the adaxial and abaxial leaf side) ( Supplementary Figures S3A, B ) and necrosis damage (adaxial and abaxial), shoot length, number of leaves, and shoot total dry weight were recorded.

RStudio (v. 4.3.1) was used for data analysis. p values < 0.05 were considered significant. Assumptions for normality and homogeneity of variance were tested using Shapiro-Wilk tests and Levene’s tests. For plant assay 1: Total thrips damage was ln-transformed to meet normality assumptions, and a two-way ANOVA was used with spraying treatment and harvesttime as factors. Shoot length and dry weight were analyzed using three-way ANOVAs with spraying treatment, thrips presence (thrips or no thrips) and harvesttime as factors. Significant interactions of two-way and three-way ANOVAs were followed by simple effects analysis via pairwise comparison of marginal means, corrected for family wise error using Bonferroni adjustment. For plant assay 2, total thrips damage, the proportion of total adaxial leaf damage to total leaf damage per plant, shoot length, and shoot dry weight were analyzed using one-way ANOVAs, followed by pairwise comparison using Tukey tests.

The samples were prepared for ¹H NMR analysis following the protocol described by Kim et al. (2010) with some modifications as follows. Per sample (five replicates), thirty mg of freeze-dried and ground leaf material was extracted by 15 minutes of ultrasonication in 1 ml of CD₃OD-KH₂PO₄ buffer in D₂O (pH 6.0, 1:1, v/v) containing 0.29 mM TMSP-d₄ ) as an internal standard. The resulting extracts were centrifuged at 13,000 rpm and 300 µl of the supernatant was transferred to 3 mm NMR tubes. ¹H NMR measurements were performed on a Bruker Avance-III 600 MHz standard bore liquid-state NMR spectrometer with operating frequency of ¹H resonating at 600.13 MHz. A cryoprobe of type TCI H&F/C/N-D with Z gradient was used. The temperature was kept constant at 298 K. For internal locking, CD₃OD was used. For each proton measurement, a 30-degree pulse of 2.64 msec at 5.5 W power with a fid resolution of 0.36 Hz, 64 scans with a relaxation delay of 1.5 secs, and acquisition time of 2.7 sec, in total taking 5 min to complete the measurement. The water signal was suppressed using a pre-saturation method and low-power selective irradiation at 0.3 Hz H₂O at 4.87 ppm. Time domain data was transformed to the frequency domain by Fourier transformation with a window function of exponential function and a line broadening parameter set to 0.3 Hz for smoothening. The generated spectrums were manually phased, baseline corrected and calibrated to TMSP-d₄ at 0.00 ppm or HMDSO at 0.06 ppm using TOPSPIN V. 3.2 software (Bruker). The NMR spectra were bucketed using AMIX 3.9.12 (Bruker Biospin GmbH Rheinstetten, Germany). The bucketed data was obtained by integration of the spectra at 0.04 ppm intervals. The peak intensity of individual peaks was scaled to the total intensity recorded from δ 0.30 to δ 10.02. Due to the residual signals of HMDSO and CH₃OH-d₄ , the regions δ 4.7 – δ 5.0 and δ 3.28 – δ 3.34 were excluded from the analysis.

Headspace-GC-MS measurements (five replicates) were performed using a 7890A gas chromatograph equipped with a 7693 automatic sampler and a 5975C single-quadrupole mass detector (Agilent, Folsom, CA, USA). Volatile compounds were separated on a DB-5 column: 30 m × 0.25 mm, 0.25 μm film (J&W Scientific, Folsom, CS, USA), using helium (99.9% purity) as a carrier gas at a flow rate of 1.6 ml/min. Freeze-dried, ground and homogenized leaf material (100 mg per sample) was placed into 20 ml glass headspace vials. Each vial was incubated at 100°C for 30 minutes, then 500 µl headspace was sampled and injected (split mode 5:1, 8 ml/min flow) into the gas chromatograph. The oven temperature was programmed starting at 70°C, held for 1 minute, then increased at 3°C/min to 100°C, re-increased to 220°C at 7°C/min, and finally increased at 14°C/min to 300°C, held for 3 min. The ionization energy in EI mode was 70 eV, and the mass scan range was set to 50-550 m/z. Tridecane (500 ng/ml) was used as an internal standard. The obtained HS-GC-MS data files of the samples were converted to.mzml format using the MSConvert tool from the Proteowizard software suite. The auto-deconvolution of GC-MS data and multivariate analysis were executed according to the workflow outlined by the Global Natural Product Social Molecular Networking (GNPS) platform (Global Natural Product Social Molecular Networking, 2024). The detected peaks were identified by comparison of their retention times and ion spectra with those listed in Atomic Spectra Database, NIST 14 (libscore cutoff value of 70). The data was then processed using MassHunter (B.07, Agilent). The obtained GC-MS data were processed by GNPS.

Principal component analysis (PCA) and orthogonal projection to latent structures discriminant analysis (OPLS-DA) were performed using SIMCA P (version 18.1, Sartorius). PCA was used to analyze the inherent variation within the datasets, with all data subjected to Pareto-scaling. OPLS-DA models were then used to assess variation in metabolite profiles between spray treatments and to discern distinct chemical compounds. The quality of the OPLS-DA models was estimated by R²X and Q² values. Q² values were obtained from permutation tests (100 permutations). R²X indicated the model’s fitness and was defined as the proportional variance, whereas Q² was defined as the predictable variance (Villa-Ruano et al., 2019). S-plots were used to identify possible biomarkers.

Plant assay 1: RGO sprayed plants showed lower mean thrips damage than water sprayed plants, both after ten days (F_{1, 36} = 11.97, p < 0.01) and 25 days (F_{1, 36} = 49.12, p < 0.01; Figure 1A ). Furthermore, thrips damage on water sprayed plants increased faster over time than on RGO sprayed plants (F_{1, 36} = 6.30, p = 0.02). However, only one adult thrips was observed to be stuck in an adhesive droplet. Coverage with adhesive droplets was also rather low with on average 159 (SD: 13) droplets per leaf (108 with area 0.1-0.5 mm², 44 of 0.5-1 mm², 7 of 1 mm² or above) or 1.74% (SD: 0.1) of the total leaf area. The leaf surface also shimmered slightly and small white flakes were visible, indicating that some of the other compounds in the sprayable solution were also present on the leaf surface ( Supplementary Figures S4A, B ). Observation via a binocular confirmed droplets of presumably less oxidized rice germ oil were also present in the sprayable RGO solutions, likely being responsible for the shimmer ( Supplementary Figures S5A, B ).

As expected, shoot length ( Supplementary Figure S6 ) was higher on day 25 than on day 10 (F_{1, 72} = 1601.64, p < 0.01). Spray treatment did not affect mean shoot length (F_{1, 72} = 2.26, p = 0.14). In general, plants with thrips had a lower shoot length (F_{1, 72} = 44.67, p < 0.01) and thrips presence reduced shoot length more on day 25 than on day 10 (thrips × day interaction: F_{1, 72} = 9.60, p < 0.01). Other two-way interactions and three-way interactions were not significant for shoot length. Dry weight ( Figure 1B ) was also higher for plants on day 25 than on day 10 (F_{1, 72} = 1266.26, p < 0.01). However, significant interactions were detected between thrips presence and harvesttime (F_{1, 72} = 10.74, p < 0.01) and between spray treatment and harvesttime (F_{1, 72} = 4.09, p = 0.047). Pairwise comparisons of marginal means indicated that, only for day 25, for plants infested with thrips, dry weight was less in water-treated plants compared to plants treated with solution containing RGO droplets (F_{1, 72} = 17.77, p < 0.01). For plants without thrips, no difference in treatments was visible on day 25 (F_{1, 72} = 0.42, p = 0.52) and the dry weight of these plants was similar to that of plants treated with RGO solution and infested with thrips.

Plant assay 2: Only two thrips were found stuck in adhesive droplets. Coverage with adhesive droplets was estimated at around 199 (SD: 15) droplets per leaf (141 with area 0.1-0.5 mm², 46 of 0.5-1 mm², 12 of 1 mm² or above) or 3.3% (SD: 0.2) of the total leaf area. Thrips damage was higher in the water treatment than in all other treatments (F_{4, 45} = 12.23, p < 0.01; Figure 1C ). The proportion of adaxial to total damage was highest in the water treatment, intermediate in the F-108, alginate and F-108 + alginate treatments, and lowest in the RGO treatment (F_{4, 45} = 9.95, p < 0.01; Figure 1D ). No differences in mean shoot length (F_{4, 45} = 0.66, p = 0.62, Supplementary Figure S7A ) or shoot dry weight (F_{4, 45} = 1.10, p = 0.37, Supplementary Figure S7B ) were found between treatments.

In this study, we used ¹H NMR to examine the metabolite profiles of Chrysanthemum leaves of RGO sprayed and control plants. As shown in Supplementary Figure S8 , major detected compounds included flavonoids (such as apigenin glycosides, 5’,7’,3’,4’-tetrahydroxy flavanone glycosides, epi- or gallocatechin gallate), carbohydrates (glucose, fructose, stachyose, and sucrose), organic acids (including formic-, fumaric-, malic-, and acetic acid), amines (choline, betaine), amino acids (alanine, glutamine), triterpenoids, and steroids.

After initial visual inspection of the ¹H NMR spectra, multivariate data analysis was conducted. The spectra were binned at intervals of 0.04 ppm, yielding 243 variables. Principal component analysis (PCA) was initially applied to the binned data to assess the primary factors influencing the Chrysanthemum leaf metabolome. As depicted in Figure 2A , the rice germ oil (RGO) treatment had the most significant impact on the metabolome, while other factors, such as harvesttime (10 vs. 25 days) and thrips infestation, were less prominent in the major principal components (PC1 and PC2).

To further investigate these minor factors, orthogonal partial least squares-discriminant analysis (OPLS-DA) was applied. The data were classified into three groups: (1) water vs. RGO treatment ( Figure 2B ), (2) control vs. thrips infestation ( Figure 2C ), and (3) harvesttime (10 vs. 25 days) ( Figure 2D ). The OPLS-DA results confirmed that the RGO treatment had a distinct impact on the metabolome, with a high Q² value of 0.97, while harvesttime also had a significant effect (Q² = 0.45). In contrast, thrips infestation showed only a marginal separation with a Q² value of 0.30 ( Figure 2C ).

To identify the specific metabolic changes associated with each factor, S-plots from the OPLS-DA models were examined ( Figure 3 ). The RGO treatment was associated with higher levels of sugar alcohols assumed by the increase signal in the range of δ 3.5 – 3.7 and oxygenated fatty alcohols and acids likely present on top or inside of the leaves ( Figure 3A ). Thrips infestation led to increased levels of flavonoids, sucrose, glucose, glutamine, and acetic acid, but decreased levels of fatty acids and malic acid ( Figure 3B ). Regarding the harvesttime, older leaves (25 days) exhibited higher levels of sucrose and malic acid, while fatty acid levels were lower ( Figure 3C ).

Finally, to assess the potential interactions between RGO treatment and other factors, separate OPLS-DA models were constructed for water-treated and RGO-treated samples ( Supplementary Figure S9 ). In RGO-treated leaves, the separation by harvesttime was more pronounced, while the effect of thrips infestation on the metabolome was reduced with a Q² value of 0.14 ( Supplementary Figure S9 ).

The use of ¹H NMR with CD₃OD-KH²PO⁴ buffer extraction effectively identified a variety of metabolite groups influenced by different factors. However, some volatile compounds were not detected. To address this, headspace GC-MS was applied to the same set of Chrysanthemum leaves. A typical headspace GC-MS chromatogram is shown in Supplementary Figure 10 . The resulting data, processed through GNPS for multivariate analysis, yielded 853 variables. The separation between groups in the GC-MS data ( Figures 4A–C ) was consistent with the results from ¹H NMR analysis. A notable difference, however, was that in the GC-MS analysis, the primary factor influencing the metabolome was harvesttime, whereas in ¹H NMR, RGO treatment was the dominant factor. Both methods revealed that thrips infestation had only a marginal impact, with a Q² value of less than 0.30 ( Figure 4B ).

Further analysis of the GC-MS data using S-plots of OPLS-DA allowed for the identification of characteristic metabolites associated with each factor. RGO treated leaf samples were generally found to have higher levels of octanal (most likely a component of the RGO) and lower levels of eucalyptol, chrysanthenone, and eugenol than those in water sprayed leaves. In terms of harvesttime, samples collected 10 days post-treatment exhibited higher levels of compounds such as 4(10)-thujene, cis-4-thujanol, eucalyptol, and isocaryophyllene, whereas samples collected after 25 days showed elevated levels of o-Cymene and 3-Methyl-2-butenoic acid.

Additional OPLS-DA analysis of GC-MS data, conducted separately for water- and RGO-treated samples, aimed to explore the relationship between RGO treatment, thrips infestation, and harvesttime ( Figures 4D, E ; Supplementary Figure 11 ). Similar to the ¹H NMR findings, RGO-treated Chrysanthemum leaves exhibited reduced differences between control and thrips-infested samples. The Q² value for control and thrips-infested RGO-treated samples was 0.30, compared to 0.79 in water-treated samples. The lower Q² value in the RGO-treated samples may indicate that the separation between control and thrips-infested samples diminishes over time, particularly after 25 days. This suggests that the RGO treatment becomes more effective later, around 25 days post-treatment ( Figure 4D ).