Volatile Proﬁling of Fifteen Willow Species and Hybrids and Their Responses to Giant Willow Aphid Infestation

: The giant willow aphid ( Tuberolachnus salignus ) is a large stem-feeding insect which forms dense colonies on infested plants. Since T. salignus is a new invasive species in New Zealand, we have a poor understanding of the plant chemical responses to aphid infestation. This study aimed to characterize the volatile organic compounds (VOCs) emissions of ﬁfteen di ﬀ erent willow species and hybrids growing in New Zealand, and to evaluate changes in response to T. salignus attack in a ﬁeld trial. Volatiles were collected using a headspace sampling technique and analysed using gas chromatography-mass spectrometry (GC-MS). We found high variability in the volatile proﬁles of di ﬀ erent species and hybrids, with ( Z )-3-hexenyl acetate and ( E )- β -ocimene being the only common components to all blends. Taxonomically related plants showed an overlapping pattern of VOC emission, and there seemed to be a clear separation between shrub and tree willows. Responses to aphid infestation were variable, with only four species / hybrids showing changes in their total VOC emission, or that of at least one class of VOCs. A weak positive correlation between aphid population estimates and VOC emissions suggests that responses are species-speciﬁc and not infestation-dependent. These results reveal useful information about the interaction between T. salignus and its potential host plants for biological control and pest management purposes.


Introduction
Plants naturally release a wide array of volatile organic compounds (VOCs) into the environment to perform various ecological and physiological processes [1]. Chemically, VOCs are low molecular weight lipophilic molecules, consisting of terpenoids, benzenoids, green leaf volatiles (GLVs), fatty acid and amino acid derivatives [2]. The production and release of these VOCs are highly responsive to biotic and abiotic factors, making them an excellent source of information for surrounding organisms [3,4]. VOCs mediate multiple ecological interactions: they can repel herbivores [5,6], attract the natural enemies of herbivores (predators and parasitoids), and lure pollinators and seed dispersers, which are key elements to the plant's defense and reproduction [7][8][9]. However, VOC can also enhance the attractiveness of plants to some herbivores, harming the emitting plants [10]. Plants also release VOCs to protect themselves from environmental stress, such as heat or UV-radiation [11,12].
A variety of factors are known to influence both the quality and quantity of VOCs emitted by plants [13]. Among them, the plant species is a determining factor, since some biosynthetic pathways

Study Site and Plant Material
This study was conducted in a willow field trial at the Orchard Block, Plant Growth Unit, Massey University, NZ (40 • 22 41.70" S, 175 • 36 30.67" E). The field trial was set up to investigate the interactions of the giant willow aphid with its host plants and the environment. Fifteen willow species and hybrids, from different geographical origins [36], were grown in three blocks of paired rows, with the position of each species being random within each paired row. Twelve ramets (an individual of plant species, vegetatively reproduced from a single parent plant) of each species or hybrid were planted within each row. Within each block, one row was randomly selected as a control (aphid exclusion), while the adjacent row was aphid-infested. Information on the willow species and hybrids, and the field trial layout are provided in Table 1 and Table S1.
The willow field trial was planted using stem cuttings in June 2017, with 0.4 m spacing between cuttings within the rows, and 4 m spacing between rows. The willow plants in the control rows were inspected for colonising aphids on a weekly basis, and any aphids found were removed manually. Mavrik ® insecticide (Nelson, NZ) was applied on 28 February 2018 and 17 January 2019, when manual control was impractical due to high population densities of T. salignus.

Aphid Inoculation
Willow plants in the aphid-infested rows were inoculated with five adult aphids per plant on 25-27 January 2018 and 6-7 December 2019. Additional inoculations with ten adult aphids per plant were done on 13-14 February 2018 and 30 January 2019. The aphid population per ramet of each sampled plant was quantified immediately before VOC collection using a visual scale from zero to six, with 0 = less than five aphids, 1 = 2 to 20 aphids, 2 = 20 to 50 aphids, 3 = 50 to 100 aphids, 4 = 100 to 300 aphids, 5 = 300 to 600 aphids, and 6 = 600 aphids or more per plant, as described by Collins [47] with slight modification.

VOCs Sampling
VOCs from willow branches were collected using the push-pull headspace sampling method as described in Effah, et al. [48]. Among the twelve plants in the row plots of each species or hybrid, one of the middle ramets (plants 5, 6 or 7) per plot was chosen to ensure that the VOCs collected were released from that treatment, without receiving VOCs from different neighboring plants, for a total of six plants per species. Willow branches of a suitable size, without visible sign of damage by insects and pathogens, were selected and enclosed in oven cooking bags (Glad ® , Melbourne, Australia). One inlet and one outlet tube were fastened with cable binders at each end of the bag. The portable volatile assay system (PVAS22 pump, VAS Rensselaer NY) was used to circulate carbon-filtered air through the bag ( Figure 1). Incoming air was pumped at 1.70 L/min and outgoing air was pulled at 1.20 L/min creating a slight overpressure to avoid contaminants from entering the bag. The VOCs emitted from the willow foliage were trapped in Haysep-Q filters attached to the outlet (pull) tubes. The pump ran for two hours, and then the filters were removed and individually wrapped with labelled pieces of aluminum foil, and stored in a cooler box to prevent contamination and evaporation of the collected volatiles. The willow branches were cut just below the bags, and the oven-dried weight of the branches was measured after drying at 60 • C for 72 h; therefore, volatiles measured are presented in nanograms per dry weight (g) per hour (ng·g·DW −1 ·h −1 ). Negative controls were also included by taking air samples from empty bags to exclude potential contaminants.
Agronomy 2020, 10, x FOR PEER REVIEW 4 of 16 branches was measured after drying at 60 °C for 72 h; therefore, volatiles measured are presented in nanograms per dry weight (g) per hour (ng·g·DW −1 ·h −1 ). Negative controls were also included by taking air samples from empty bags to exclude potential contaminants.

Figure 1.
Volatile organic compound (VOC) sampling from willow foliage using a portable volatile assay system. Branches were enclosed in oven cooking bags, into which carbon-filtered air was circulated through a push-pull system.
The first VOC sampling was performed on 17-23 January 2018, to characterize the willow VOCs (n = 6 branches/species or hybrid, 90 in total) before giant willow aphid inoculation. To estimate the effect of T. salignus on the VOC emissions of the willow plants, second VOC sampling was done shortly after aphid inoculation from both the control and aphid-infested plants on [15][16][17] March 2019.

Gas Chromatography-Mass Spectrometry (GC-MS) Analysis
The volatile compounds in the filters were eluted using a solvent solution with an internal standard (200 µL hexane with 10 ng/mL of nonyl acetate) into gas chromatography-mass spectrometry (GC-MS) vials and then stored in a −80 °C freezer before analysis. The willow volatiles were separated and identified using the GC-2010 Plus Gas Chromatograph (Shimadzu, Japan) coupled to the AOC-20 I Auto-injector, QP2010 SE-gas chromatograph-mass spectrophotometer, and TG-5MS column (30 m × 250 µm × 0.25 µm). Helium (He) was used as a carrier gas with the flow rate of 0.5 mL/min into split mode (10:1). The injector port and detector were set up at 250 and 230 °C, respectively. The oven temperature was initially held at 50 °C for 3 min, then increased by 5°C/min to 95 °C, and then ramped to 240 °C at 15°C/min, where it was maintained for three minutes. VOCs were tentatively identified using the NIST (National Institute of Standards and Technology) Mass Spectral Library and confirmed by comparing their retention times with those of commercial standards whenever available. Post-run analyses were carried out using the Shimadzu Lab Solutions software (version 2.50). VOCs were quantified by dividing their peak area by that of the internal standard and expressed as nanogram per microliter per gram of dry weight of foliage per hour (ng·g −1 ·h −1 ). Contaminants (toluene, p-xylene, o-xylene, diethyl phthalate, etc.) that were consistently identified in negative controls (empty oven cooking bags) were excluded from further analyses. Volatile organic compound (VOC) sampling from willow foliage using a portable volatile assay system. Branches were enclosed in oven cooking bags, into which carbon-filtered air was circulated through a push-pull system.
The first VOC sampling was performed on 17-23 January 2018, to characterize the willow VOCs (n = 6 branches/species or hybrid, 90 in total) before giant willow aphid inoculation. To estimate the effect of T. salignus on the VOC emissions of the willow plants, second VOC sampling was done shortly after aphid inoculation from both the control and aphid-infested plants on [15][16][17] March 2019.

Gas Chromatography-Mass Spectrometry (GC-MS) Analysis
The volatile compounds in the filters were eluted using a solvent solution with an internal standard (200 µL hexane with 10 ng/mL of nonyl acetate) into gas chromatography-mass spectrometry (GC-MS) vials and then stored in a −80 • C freezer before analysis. The willow volatiles were separated and identified using the GC-2010 Plus Gas Chromatograph (Shimadzu, Japan) coupled to the AOC-20 I Auto-injector, QP2010 SE-gas chromatograph-mass spectrophotometer, and TG-5MS column (30 m × 250 µm × 0.25 µm). Helium (He) was used as a carrier gas with the flow rate of 0.5 mL/min into split mode (10:1). The injector port and detector were set up at 250 and 230 • C, respectively. The oven temperature was initially held at 50 • C for 3 min, then increased by 5 • C/min to 95 • C, and then ramped to 240 • C at 15 • C/min, where it was maintained for three minutes. VOCs were tentatively identified using the NIST (National Institute of Standards and Technology) Mass Spectral Library and confirmed by comparing their retention times with those of commercial standards whenever available. Post-run analyses were carried out using the Shimadzu Lab Solutions software (version 2.50). VOCs were quantified by dividing their peak area by that of the internal standard and expressed as nanogram per microliter per gram of dry weight of foliage per hour (ng·g −1 ·h −1 ). Contaminants (toluene, p-xylene, o-xylene, diethyl phthalate, etc.) that were consistently identified in negative controls (empty oven cooking bags) were excluded from further analyses.

Statistical Analysis
The R statistical software (Version 3.6.1) [49] was used for all analyses. For the VOC profiling of the fifteen willow species and hybrids, 19 VOCs were chosen, based on their consistent occurrence in the samples of at least one species or hybrid. One replicate of S. × fragilis that did not emit the selected compounds was dropped from analyses. The VOC data were square-root transformed to achieve normality [50], and to allow rare VOCs to have equal weight by reducing the overestimation of highly-occurring VOCs in the headspace samples [51]. We performed a permutational multivariate analysis of variance (PERMANOVA) with Bonferroni adjustment to differentiate the VOC blends of the fifteen willow species and hybrids. The analysis was done using the Adonis function with Bray-Curtis distance matrix and 999 permutations [52]. Non-metric Multi-Dimensional Scaling (NMDS) was performed to depict differences in the VOC profiles of the willow species and hybrids. The PERMANOVA and NMDS were performed using the vegan [53] package.
The relative proportion of each major VOC class (aldehydes, GLVs, monoterpenes and sesquiterpenes) was calculated by summing up the specific VOC concentrations for each group, and then dividing by the whole blend as described by Digilio, et al. [54]. We constructed linear mixed model (LMM) ANOVAs on the square root transformed relative VOC proportions using the lme4 package, to further differentiate the emission of VOC class within each species or hybrid. The fifteen willow species and hybrids were treated as a fixed factor, while the row number of the VOC sampling was considered as a random variable. The linear mixed model was fitted on a proxy-log scale and a multiple comparison was then performed using Tukey's HSD test in multcomp and lsmeans packages.
To test the response of the willows to aphid infestation, a Tweedie generalized linear model with gamma distribution and log-link function was used to compare the total VOC emissions, and that of each of the four major VOC groups (aldehydes, GLVs, monoterpenes and sesquiterpenes). The concentrations of specific VOCs were summed up to become the total concentrations of the major VOC classes. A one-way ANOVA, followed by Tukey's HSD test was used to compare aphid population levels on willow species, monitored just before VOC sampling. Finally, a Spearman's rank correlation was performed to correlate the aphid population level (proxy-log scale) and the total VOC emissions for the aphid-infested willows; the relationship was visualized using the ggpubr package.

Number of VOCs Emitted
The VOC profiles of the tested willow species and hybrids differed significantly (PERMANOVA; Pseudo-F 14,88 = 5.83, p < 0.001). Due to the high overlap of the VOC profiles, the NMDS algorithm yielded a low stress value (badness-of-fit; 0.20). However, some of the willow species and hybrids were distinguishable from others. For instance, there was clear differentiation between S. lasiandra and S. lasiolepis, S. schwerinii, S. matsudana and S. matsudana × S. alba (NZ 1040). Salix schwerinii had no overlap with S. eriocephala, S. × reichardtii, S. lasiolepis × S. viminalis, S. lasiandra and S. matsudana × S. alba (NZ 1040) ( Figure 2). As expected, the closely related species and hybrids had more similar VOC profiles, as observed for S. matsudana and its hybrids. Tree willows were clustered separately from the shrub willows (Figure 2).
The The VOC profiles of the tested willow species and hybrids differed significantly (PERMANOVA; Pseudo-F14,88 = 5.83, p < 0.001). Due to the high overlap of the VOC profiles, the NMDS algorithm yielded a low stress value (badness-of-fit; 0.20). However, some of the willow species and hybrids were distinguishable from others. For instance, there was clear differentiation between S. lasiandra and S. lasiolepis, S. schwerinii, S. matsudana and S. matsudana × S. alba (NZ 1040). Salix schwerinii had no overlap with S. eriocephala, S. × reichardtii, S. lasiolepis × S. viminalis, S. lasiandra and S. matsudana × S. alba (NZ 1040) ( Figure 2). As expected, the closely related species and hybrids had more similar VOC profiles, as observed for S. matsudana and its hybrids. Tree willows were clustered separately from the shrub willows ( Figure 2). The

VOC Response of Willow Species and Hybrids to T. salignus Infestation
Eighteen VOCs were identified in the headspace samples from aphid-infested plants (Table S2) and different willow species responded differently to T. salignus infestation. In most cases, the VOC profiles of aphid-infested willow plants did not differ significantly from those of control plants ( Figure 4). However, upon closer inspection, aphid infestation was found to significantly decrease total VOC emission in S. × reichardtii (Figure 4), GLV emission in S. matsudana × S. lasiandra, monoterpene emission in S. × reichardtii and S. candida, and sesquiterpene emission in S. × reichardtii, S. matsudana × S. alba (NZ 1184) and S. candida ( Figure 5).

VOC Response of Willow Species and Hybrids to T. salignus Infestation
Eighteen VOCs were identified in the headspace samples from aphid-infested plants (Table S2) and different willow species responded differently to T. salignus infestation. In most cases, the VOC profiles of aphid-infested willow plants did not differ significantly from those of control plants (Figure 4). However, upon closer inspection, aphid infestation was found to significantly decrease total VOC emission in S. × reichardtii (Figure 4), GLV emission in S. matsudana × S. lasiandra, monoterpene emission in S. × reichardtii and S. candida, and sesquiterpene emission in S. × reichardtii, S. matsudana × S. alba (NZ 1184) and S. candida ( Figure 5).      Table S3.  Table S3.

Correlation between Aphid Infestation and VOC Emissions
Our observation of aphid populations before VOC collection revealed different degrees of infestation among willow species (Figure 6), with S. eriocephala, S. matsudana, and S. lasiolepis × S. viminalis having very low infestation rates (most plants having 20 aphids or less) and S. viminalis having the highest infestation rates (most plants having 300 aphids or more), followed by S. candida (most plants having over a 100 aphids). Correlation analysis revealed only a weak relationship between aphid infestation and VOC emissions (n = 45, ρ = 0.34, p = 0.02).

Correlation between Aphid Infestation and VOC Emissions
Our observation of aphid populations before VOC collection revealed different degrees of infestation among willow species (Figure 6), with S. eriocephala, S. matsudana, and S. lasiolepis × S. viminalis having very low infestation rates (most plants having 20 aphids or less) and S. viminalis having the highest infestation rates (most plants having 300 aphids or more), followed by S. candida (most plants having over a 100 aphids). Correlation analysis revealed only a weak relationship between aphid infestation and VOC emissions (n = 45, ρ = 0.34, p = 0.02).

Discussion
Our results showed variation in the VOC profiles of the different willow species and hybrids, attributed to the diversity of Salix spp. used in current study ( Table 1). The VOC emissions were found to be more similar between closely related plants, as shown by close clustering between S. matsudana and its hybrids (S. matsudana × S. alba (NZ 1040), S. matsudana × S. alba (NZ 1184) and S. matsudana × S. lasiandra). VOC composition was distinct between shrub and tree willows, supporting the genetic clustering between these two growth forms as recently reported by Ngantcha [55].
Although species and hybrids varied in the number of compounds emitted, two compounds were common to all blends: GLVs (Z)-3-hexenyl acetate and monoterpene (E)-β-ocimene. Both compounds have been reported for S. eriocarpa [45], S. viminalis [43] and in related poplar trees (Populus nigra and P. trichocarpa) [9,56], suggesting that they are ubiquitous to plants species in the family Salicaceae. Studies suggest that these two compounds play important ecological roles. For example, (Z)-3-hexenyl acetate has been identified as a key compound associated with herbivore damage in different willow varieties and in poplar species, and is known to attract natural enemies [9,30,57]. In other willow and poplar species (E)-β-ocimene is involved in within-plant communication as a signal emitted by damaged plant parts to alert nearby undamaged parts of potential attack [58]. In our study, some species and hybrids reduced their emissions of these two compounds in response to T. salignus herbivory ( Figure S1, Table S4). The reasons behind this reduction are not yet known and deserve further investigation.
Few studies have explored willow VOC responses to herbivore attack. HIPV emission by willows can be highly specific, varying with the life stage of their attacker (larvae vs. adult), and informing natural enemies about the suitable stage of their prey [59]. The predatory ladybird

Discussion
Our results showed variation in the VOC profiles of the different willow species and hybrids, attributed to the diversity of Salix spp. used in current study ( Table 1). The VOC emissions were found to be more similar between closely related plants, as shown by close clustering between S. matsudana and its hybrids (S. matsudana × S. alba (NZ 1040), S. matsudana × S. alba (NZ 1184) and S. matsudana × S. lasiandra). VOC composition was distinct between shrub and tree willows, supporting the genetic clustering between these two growth forms as recently reported by Ngantcha [55].
Although species and hybrids varied in the number of compounds emitted, two compounds were common to all blends: GLVs (Z)-3-hexenyl acetate and monoterpene (E)-β-ocimene. Both compounds have been reported for S. eriocarpa [45], S. viminalis [43] and in related poplar trees (Populus nigra and P. trichocarpa) [9,56], suggesting that they are ubiquitous to plants species in the family Salicaceae. Studies suggest that these two compounds play important ecological roles. For example, (Z)-3-hexenyl acetate has been identified as a key compound associated with herbivore damage in different willow varieties and in poplar species, and is known to attract natural enemies [9,30,57]. In other willow and poplar species (E)-β-ocimene is involved in within-plant communication as a signal emitted by damaged plant parts to alert nearby undamaged parts of potential attack [58]. In our study, some species and hybrids reduced their emissions of these two compounds in response to T. salignus herbivory ( Figure S1, Table S4). The reasons behind this reduction are not yet known and deserve further investigation.
Few studies have explored willow VOC responses to herbivore attack. HIPV emission by willows can be highly specific, varying with the life stage of their attacker (larvae vs. adult), and informing natural enemies about the suitable stage of their prey [59]. The predatory ladybird (Aiolocaria hexaspilota) was more attracted to VOC blends induced by willow beetle (Plagiodera versicolora) larvae, containing higher amounts of the GLVs (Z)-3-hexenol and (Z)-3-hexenyl acetate, the monoterpenes (E)-β-ocimene, (Z)-β-ocimene, allo-ocimene and linalool, the sesquiterpene (E)-α-farnesene, and two oximes (nitrogenous compounds) [59]. These results show that HIPVs play a role in host selection by herbivores, and in indirect defense in willows.
In the present study, we observed that not all willow species and hybrids reacted to aphid damage in the same way. While some willows (S. × reichardtii, S. matsudana × S. lasiandra, S. matsudana × S. alba, and S. candida) responded by decreasing their VOC emissions, the majority of species and hybrids did not show a significant change. Other studies on phloem feeders show contrasting results, with some reporting increases in VOC emissions after attack [26,[60][61][62][63]. For instance, infestation by spiral gall aphid Pemphigus spyrothecae (Hemiptera: Aphididae) on leaf tissue can alter leaf's photosynthetic activity that in turn triggers jasmonate transportation from petiole to lamina and finally modifies VOC emission in poplar (Populus × petrovskiana) [63]. However, there are also reports showing reductions in VOC emissions or no response at all [27,61,64]. Furthermore, studies comparing chewers and phloem feeders typically indicate that the latter have a less pronounced effect on VOC emissions than chewing herbivores [27,64].
The lack of response in most species and hybrids may be due to the fact that the giant willow aphid does not directly damage the photosynthetically active tissue (leaves), nor causes severe mechanical damage (as chewing herbivores do), and thus may not trigger strong changes in VOC emissions [64]. However, it has been suggested that aphids actively suppress plant responses to escape their natural enemies [61]. This manipulation of plant responses is possibly mediated by microbial endosymbionts of aphids in order to protect their hosts [65]. Further studies are required to clarify the mechanism behind the observed responses (or lack of them).
The VOC emission reduction observed in some species and hybrids may also be due to a trade-off between indirect and direct defense (i.e., production of VOCs vs. non-volatile secondary metabolites) [66][67][68][69]. Both direct and indirect defenses have a metabolic cost to the plant, and plants typically favor one type of defense over another. For instance, a study investigating wild and cultivated accessions of lima bean (Phaseolus lunatus) found that plants producing high levels of cyanogenic compounds (direct defenses) released low amounts of VOCs (indirect defenses) and vice versa [69]. In this study, some species and hybrids hosted lower aphid populations and appear to be more naturally resistant to aphid attack than others ( Figure 6). This resistance is possibly associated with the presence of physical defense mechanisms (e.g., rough bark of resistant species), suggesting that defense trade-offs could exist in different types of willows.
Plant volatiles are known to play a role in deterring herbivores and attracting natural enemies in related tree species [6,9,59], but considering the costs involved in VOC production and emission [70,71], it would be disadvantageous for a plant to increase its emissions if there was an elevated fitness cost (e.g., higher appetency to generalist herbivores) with no net benefit (e.g., no attraction of natural enemies); such as in our case, where the invasive aphid lacks specialist natural enemies.
The degree of infestation ( Figure 6) could also have contributed to different outcomes. The emission of HIPVs can qualitatively and quantitively vary depending on the population density of the insect feeding on host plants [72,73], with studies typically reporting a positive correlation between herbivore population density and VOC emission [74][75][76]. However, we only found a weak correlation between aphid infestation and VOC emission, and responding species and hybrids (except S. candida) were not heavily infested. This shows that responses are host-specific and less dependent on the degree of infestation, although within the responding species and hybrids, changes in aphid density may affect VOC emissions. Further studies are required to test this hypothesis.
A study by Aradottir et al. [46] found that T. salignus was significantly attracted to certain willow varieties but not to others in laboratory olfactometry tests. Although the compounds involved were not identified, this evidence shows that the giant willow aphid uses plant volatiles to choose their host plants. Therefore, future research should explore the role of VOCs in T. salignus host selection and colonization, and the behavioral responses of potential natural enemies of T. salignus, such as the harlequin ladybird Harmonia axyridis [77] to willow VOCs. Our results suggest that some naturally resistant species and hybrids (S. lasiolepis × S. viminalis and S. eriocephala) are rich GLV emitters, which are known to repel herbivores and attract natural enemies in other systems [78]. Representative compounds from this group, such as (Z)-3-hexenyl acetate, are good candidates for further testing.
In this study, plants belonging to the same species or hybrid had the same sex, so we did not explore the influence of plant sex on VOC emissions or responses to herbivory; this is an aspect that requires further investigation. Being an exploratory study, our results were limited to a low number of replicates, and therefore, we encourage additional studies with higher replication to confirm these findings. Aphid infestation was unequal between plants, as we wanted to explore aphid behavior in nature, and in doing so, were able to identify some species and hybrids which appear to be naturally more resistant than others to aphid attack. We are following this lead towards the selection of resistant species and hybrids for sustainable willow growing.

Conclusions
To summarize, there was a high variation in VOC emissions by different willow species and hybrids, with clear clustering between tree and shrub species. Most species and hybrids did not show significant changes in their VOC emissions in response to T. salignus infestation, but in those that did, this response was typically a reduction in VOC emissions. Whether this occurs due to the lack of response by the plant, trade-offs between direct and indirect defenses, or the active suppression of plant defenses by the aphid, requires further testing. Our study provides the foundation to further explore the role of willow VOCs in host selection by T. salignus. This information will contribute to the selection of willow species and hybrids for future planting, to reduce the ecological and economic impacts of this emerging pest.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4395/10/9/1404/s1. Table S1: Split-plot experimental layout of the willow field trial, Table S2: GC-MS analysis of VOCs released from the foliage of 15 willow species and hybrids, for the control and aphid-infested treatments, Table S3: Mean emissions of total and major VOC classes as influenced by willow species/hybrid and aphid infestation, Table S4: Mean emissions of selected VOCs as influenced by willow species/hybrid and aphid infestation, Figure S1: Total emissions of the terpenoid (E)-β-ocimene and the GLV (Z)-3-hexenyl acetate. Asterisks indicate significant differences between the treatments within each species or hybrid, Tukey's HSD test, α = 0.05. Details can be seen in Table S4.