Elevated CO2 and O3 alter the feeding efficiency of Acyrthosiphon pisum and Aphis craccivora via changes in foliar secondary metabolites

Elevated CO2 and O3 can affect aphid performance via altering plant nutrients, however, little is known about the role of plant secondary metabolites in this process, especially for aphids feeding behaviors. We determined the effects of elevated CO2 and O3 on the growth and phenolics of alfalfa (Medicago sativa) and feeding behaviors of the pea aphids (Acyrthosiphon pisum) and cowpea aphids (Aphis craccivora). Elevated CO2 improved plant growth, but could not completely offset the negative effects of elevated O3. Elevated O3 increased foliar genistin content at the vegetative stage, increased ferulic acid at the reproductive stage, and elevated CO2 increased those at both stages. Simultaneously elevated CO2 and O3 increased foliar ferulic acid content at the reproductive stage and increased genistin content at both stages. For pea aphids, feeding efficiency was reduced under elevated CO2 at the reproductive stage and decreased under elevated O3 at the vegetative stage. For cowpea aphids, feeding efficiency was increased under elevated CO2 at the vegetative stage and decreased under elevated O3 at both stages. Simultaneously elevated CO2 and O3 decreased both aphids feeding efficiency. We concluded that CO2 and O3 independently or interactively had different effects on two aphids feeding behaviors through altering foliar ferulic acid and genistin contents.

glycoside in trembling aspen (Populus tremuloides) are higher in the younger leaves than in the older leaves 15,16 . The plant ontogenetic stage and climate changes may interactively influence plant secondary metabolites, for example, phenolic glycoside concentrations increase in the leaves of the younger trees but decrease in the older trees under elevated CO 2 , while elevated O 3 has the opposite effects 17 . These changes in plant primary and secondary metabolites under elevated CO 2 and O 3 in turn influence the performance of insect herbivores 6,7 . Many previous work has shown that elevated CO 2 or O 3 reduce the performance of most leaf-chewing insects due to decreased nitrogen concentrations in plants 7,18,19 . On the other hand, the increased concentration of plant secondary metabolites may partially explain the rduced performance of leaf-chewing insects under elevated CO 2 and O 3 , because the secondary compounds can induce or decline the detoxication activities of the digestion system, which thereby prolonging developmental time and reducing growth rates 20,21 . For the piercing-sucking insects e.g., aphids, many studies have shown that elevated CO 2 or O 3 can increase aphid performance via promoting plant nitrogen based nutrition [22][23][24][25] . However, Johnson et al. 26 find that the fecundity of the pea aphid (Acyrthosiphon pisum) responds differently to elevated CO 2 when fed on five different alfalfa (Medicago sativa) cultures. Furthermore, the population abundance of Rhopalosiphum padi is reduced under elevated CO 2 when fed on tall fescue (Schedonorus arundinaceus), but is increased when fed on barley (Hordeum vulgare) 27 . These results presumably indicate that the same aphid species that fed on different host plants exhibits different responses to climate changes. On the other hand, different aphid species or even genotypes that fed on the same host plants may also perform differentially under elevated CO 2 and O 3 28 . For example, the specialist Brevicoryne brassicae are larger and accumulate more fat, while no changes are found in the generalist Myzus persicae reared on Brassica oleracea under elevated CO 2 29 . Therefore, the responses of aphids to elevated CO 2 and O 3 seem to be species-specific, demonstrating decreased, increased, or unchanged population abundance, growth, and fecundity [30][31][32][33] , but more evidence is needed to explain the heterogeneous responses.
Plant secondary metabolites may also contribute to the idiosyncratic responses of aphids to climate changes, though few studies have focused on it 34,35 . For the aphids that feed exclusively on the phloem sap, plant secondary metabolites mainly affect aphid feeding behaivor rather than the digestion system because the secondary metabolites rarely distribute in the phloem sap 36,37 . Many plant secondary compounds, including alkaloid, luteolin, genistein, apigenin, and saponin etc., can negatively affect the penetration pathway stage of aphid feeding [38][39][40][41] . For example, caffeic, ferulic, and sinapic acids disfavor the grain aphid (Sitobion avenae) feeding by prolonging the early pathway phases of probing, increasing the number of probing, and reducing salivation into sieve elements and ingestion of phloem sap 42 . Furthermore, the same secondary metabolites seem to have different effects on feeding activities of different aphid species. For example, the total time of probing of Aphis fabae, Aphis craccivora, A. pisum, and M. persicae increases with a reduced alkaloid content in narrow-leafed lupins (Lupinus angustifolius), whereas the alkaloid content has no influence on Macrosiphum albifrons; furthermore, when fed on the alkaloid-rich cultivar ' Azuro' , a reduced occurrence of phloem phases is observed, especially for A. pisum and A. fabae, whereas M. albifrons shows the longest phloem phase 38 . However, it is also reported that secondary compounds may act as probing stimulants of aphids 43 . Therefore, the increases in plant secondary metabolites may increase or decrease epidermis and mesophyll resistance against aphids during pathway and probing feeding stages under elevated CO 2 or O 3 , and the differential feeding responses of aphid species to plant secondary metabolites might contribute to their idiosyncratic responses to elevated CO 2 or O 3 . However, the evidence about how climate changes, especially for elevated O 3 alter the aphid feeding activities via plant secondary metabolites is lacking.
The current study aimed to investigate how elevated CO 2 and O 3 , alone or in combination, alter plant secondary metabolites at different ontogenetic stages and produce cascading effects on aphid feeding behaviors. By using 12 field open-top chambers with four treatments (control, elevated CO 2 , elevated O 3 , and elevated CO 2 and O 3 ), we measured plant growth traits, secondary metabolites, and feeding behaviors of the pea aphid (Acyrthosiphon pisum Harris) and the cowpea aphid (Aphis craccivora Koch) on alfalfa (Medicago sativa). Our specific objectives were to determine: (1) the effects of elevated CO 2 and O 3 on plant growth traits and secondary metabolites phenolics of alfalfa, and (2) the feeding behaviors of the two species of aphids when fed on alfalfa grown under different treatments.

Results
Responses of plant growth traits to elevated CO 2 and O 3 . Elevated CO 2 , O 3 , plant developmental stage, and their interactions significantly influenced alfalfa growth (Table 1). Elevated CO 2 did not significantly alter the plant chlorophyll content but did increase the plant net photosynthetic rate at the vegetative stage, while decreased them at the reproductive stage (significant CO 2 × developmental stage interaction, Table 1, Fig. 1a,b). Elevated CO 2 increased plant aboveground biomass only at the vegetative stage (Fig. 1d) and increased plant belowground biomass only at the reproductive stage (Fig. 1e). Elevated CO 2 also increased the numbers of flowers and pods (Fig. 1g,h). Contrary to elevated CO 2 , O 3 fumigation decreased the plant chlorophyll content, net photosynthetic rate, and aboveground biomass at both developmental stages, and decreased plant belowground biomass at the vegetative stage, regardless of CO 2 levels ( Table 1, Fig. 1a,b,d and e). Furthermore, O 3 fumigation alone or in combination with CO 2 , also decreased the numbers of flowers and pods (Fig. 1g, Fig. 2c). Elevated O 3 alone increased foliar p-coumaric acid content at the vegetative stage, and increased it at the reproductive stage, regardless of CO 2 levels ( Table 2, Fig. 2d). CO 2 and O 3 fumigation increased foliar genistin content at the vegetative stage, but only CO 2 fumigation increased genistin content at the reproductive stage (Fig. 2a). CO 2 and O 3 fumigation decreased foliar apigenin content at both developmental stages, with higher values at the vegetative stage than those at the reproductive stage ( Table 2, Fig. 2b). Both gases fumigation decreased foliar rutin content only at the vegetative stage and decreased genistein content only at the reproductive stage (significant CO 2 × developmental stage interaction, significant O 3 × developmental stage interaction, Table 2, Fig. 2f,g). Overall, elevated CO 2 and O 3 had significant effects on alfalfa foliar secondary metabolites, and these effects varied between vegetative stage and reproductive stage.
Responses of aphid feeding behaviors to elevated CO 2 and O 3 . Elevated CO 2 had significant effects on the time the pea aphids spent on nonpenetration, phloem feeding, potential drops, time to first potential drops, and time to first phloem feeding. Elevated O 3 and developmental stage had effects similar to those for elevated CO 2 , except for phloem feeding (Table 3). Elevated CO 2 had no significant impacts on the time the pea aphids spent on phloem feeding at the vegetative stage but decreased it at the reproductive stage, whereas elevated O 3 had the opposite effects (Fig. 3a). Simultaneously elevated CO 2 and O 3 reduced the feeding efficiency of the pea aphids at both developmental stages (Fig. 3a), especially at the vegetative stage, simultaneously elevated CO 2 and O 3 increased the time to first potential drops and time to first phloem feeding (Fig. 3e,g).
Elevated CO 2 had little effect on the cowpea aphids feeding, but elevated O 3 and developmental stage had significantly altered their feeding activities (Table 3). Elevated CO 2 increased the time the cowpea aphids spent on phloem feeding at the vegetative stage, but had no effects at the reproductive stage (Fig. 3h). Elevated O 3 , alone or in combination with CO 2 , decreased the time the cowpea aphids spent on phloem feeding at both developmental stages (Fig. 3h). O 3 fumigation also increased the time to first potential drops and time to first phloem feeding of the cowpea aphids at both developmental stages, regardless of CO 2 levels ( Fig. 3l,n).

Relationships between plant secondary metabolites and aphid feeding activities.
Based on the aforementioned results, we calculated correlations between plant secondary metabolites that were significantly altered by CO 2 and O 3 , and aphids feeding time of phloem feeding and potential drops. The results from the correlation analysis showed that the foliar apigenin and p-coumaric acid contents were significantly negatively correlated with the duration time of potential drops of the pea aphids, whereas the foliar ferulic acid and genistin contents were negatively correlated with the duration time of phloem feeding of the pea aphids ( Table 4). The foliar p-coumaric acid content was significantly negatively correlated with the duration time of phloem feeding of the cowpea aphids, and ferulic acid content was positively correlated with the the duration time of potential drops of the cowpea aphids ( Table 4).

Effects of secondary compounds on aphid feeding behaviors.
The time the pea aphids spent on phloem feeding was significantly reduced on treated plants but not on +ferulic acid compared to controls (Fig. 4a). The time that the pea aphids spent on potential drops was significantly increased, and the total number of potential drops was significantly increased on the plants treated with +apigenin and +p-coumaric acid (Fig. 4b,f).
Supplemental genistin also decreased the time the cowpea aphids spent on phloem feeding, whereas +ferulic acid and +p-coumaric acid significantly increased it (Fig. 4h). In addition, +ferulic acid and +p-coumaric acid reduced the time the cowpea aphids spent on potential drops and the total number of potential drops (Fig. 4i,

Discussion
Elevated CO 2 and O 3 affect the performance of herbivorous insects mainly by altering host plant primary and secondary metabolites 6 . Phenolics are important secondary metabolites for plants in defence against herbivory 44 . In the current study, elevated CO 2 and O 3 altered alfalfa growth traits and concentrations of phenolics such as genistin, ferulic acid, p-coumaric acid, and apigenin, which were antifeedants or feeding stimuli to aphids according to our exogenous application bioassay. Thus, aphid feeding efficiency was differentially altered by these changes in secondary metabolites. Furthermore, the responses of plant secondary metabolites and corresponding aphid feeding activities to elevated CO 2 and O 3 varied between plant developmental stages. The positive effects of elevated CO 2 on the plant photosynthesis, growth, and seed yield in legume alfalfa are consistent with the results of a previous work that has studied soybean (Glycine max) 45 . However, elevated O 3 negatively affected alfalfa growth, and these reductions in plant photosynthesis and growth under elevated O 3 may be due to the generation of ROS damage to photosynthetic processes such as the synthesis of rubisco 46 . Our study also showed that elevated CO 2 did not offset the negative effects of elevated O 3 on the plant growth. This finding is in contrast to other studies in which elevated CO 2 has been shown to ameliorate the negative effects of elevated O 3 47,48 . These contradictory results may be due to the interactions between CO 2 and O 3 depending on plant species 49 and plant developmental stages 17 .
In addition to having effects on alfalfa growth traits, elevated CO 2 and O 3 altered the concentrations of the foliar secondary metabolites phenolics, such as rutin, ferulic acid, genistin, apigenin, and p-coumaric acid. For example, ferulic acid and genistin contents were significantly increased at both plant developmental stages under elevated CO 2 . Plant phenolics are formed from phenylalanine via the shikimic acid-phenylpropanoid pathway 50 , and the biosynthesis of phenylpropanoids requires the efficient flow of carbon into phenylalanine biosynthesis 51 . This process can be regulated by phytohormones, such as JA (jasmonic acid), SA (salicylic acid), and ET (ethylene), which can be affected by elevated CO 2 52,53 . For example, elevated CO 2 increases the concentration of SA-regulated phenolics, such as flavonoids (e.g., quercetin, kaempferol, and fisetin) in Malaysian young ginger (Zingiber officinale Roscoe) 54 , but reduces the concentration of JA-regulated isoflavonoids, such as genistein in soybean (G. max) 55 . Although our results are not absolutely consistent with these studies mentioned above, we all indicate that individual phenolic compounds differentially respond to elevated CO 2 . In addition, plant secondary metabolites, such as alkaloids and phenolics, often change dramatically as plants develop 56,57 . Furthermore, plant developmental stages interact with atmospheric changes to influence the secondary metabolism 6 . However, most studies have investigated only one developmental stage. For example, six of the nine sympatric British grassland species studied exhibits a significant increase in one or more secondary metabolites throughout seedling ontogeny 56 . Few studies have focused on multiple ontogenetic stages, such as the seedling stage, vegetative juvenile stage, and mature stage 58 . Our study included two plant developmental stages, the vegetative and reproductive stages. We found that the responses of foliar ferulic acid and genistin contents to elevated CO 2 and O 3 were influenced by the plant developmental stage. For example, genistin content was increased only at the vegetative stage, ferulic acid content was increased only at the reproductive stage under elevated O 3 , and the foliar apigenin, kaempferol, and genistein contents were much higher at the vegetative stage than those at the reproductive stage. The heterogeneous defence chemical composition between vegetative and reproductive stages might be explained by that the direction of changes in defensive compounds during the transition from juvenile to mature stage depends on the types of compounds in herbs 58 . An increase in phenolic content under O 3 fumigation is also commonly reported 6 , though elevated O 3 has been shown to damage plant photosynthesis 59 . These changes in phenolics under elevated O 3 may be due to the increase in the activities of phenylalanine-ammonium lyase (PAL) and chalcone synthase enzymes (CHS), which are key enzymes in the biosynthesis of phenolics 60 . The increased phenolics may also act as antioxidants against oxidative stress caused by O 3 61 .
Plant secondary metabolites can affect the behavior, growth, and development of herbivorous insects 62,63 . According to our bioassay, ferulic acid and p-coumaric acid acted as feeding stimulants of the cowpea aphids, while ferulic acid did not affect the pea aphids, and p-coumaric acid acted as an antifeedant of the pea aphids. These results suggest that a single compound can have different and even opposite effects on these two species of aphids. Similar results have been reported for p-coumaric acid and ferulic acid, which are phagostimulants for the stem borer (Chilo partellus Swinhoe) 64 , but feeding inhibitors for maize weevil (Sitophilus zeamais Motschulsky) 65 . However, studies have concluded that p-coumaric acid and ferulic acid have negative effects on the performance of the grain aphid (S. avenae) 66 . The discrepancy among these studies might be due to the feeding guilds or food habits of herbivores, such as specialist and generalist responding differently to plant secondary chemistry [67][68][69] , reflecting different coevolution between insects and plant defence. In this study, genistin acted as an antifeedant of both the pea aphids and cowpea aphids, which had been demonstrated to negatively affect stinkbugs (Nezara viridula and Piezodorus guildinii) and whitefly (Bemisia tabaci) 70,71 . They all seem to imply that genistin may be important secondary compounds conferring resistance to many insects that belong to different feeding guilds.
Our study also indicated that p-coumaric acid did not play a key role in responses of the cowpea aphids feeding to elevated O 3 , though p-coumaric acid acted as their feeding stimulant. This may be because many compounds coexist in plant leaves and interactions among them may alter the effects of a single chemical 62 .
In contrast to studies suggesting that elevated CO 2 favours aphid phloem feeding, such as pea aphid fed on Meidicago truncatula and green peach aphid fed on Nicotiana attenuata [72][73][74] , our research demonstrated that elevated CO 2 discouraged the pea aphids feeding at the reproductive stage, though favoured the cowpea aphids feeding at the vegetative stage. These results imply that the responses of aphids or even aphid-plant to elevated CO 2 are species-specific. The heterogeneous responses have been widely reported 75 , however, the underlying mechanism is very complex. Our study suggests that the differential responses of two species of aphids to various compounds (genistin, ferulic acid, and p-coumaric acid) may be one reason. Indeed, studies have also shown that the aphid species or genotypes differentially respond to the same secondary compounds, such as thymol, The performance of aphids can be decreased, increased, or unchanged under elevated O 3 , depending on the duration and concentration of O 3 exposure and the age of the exposed plants 32,77 . However, our results showed that elevated O 3 decreased the feeding efficiency of aphids at both plant developmental stages, except for no impact on phloem feeding of the pea aphids at the reproductive stage. Furthermore, the feeding efficiency of two species of aphids was inhibited under simultaneously elevated CO 2 and O 3 , indicating that elevated CO 2 did not completely offset the negative effects of elevated O 3 . This finding is in contrast to previous studies that elevated CO 2 ameliorates the negative impact of elevated O 3 on herbivores 19,78 . A possible explanation is that those studies use leaf-chewing herbivores and use trees as the host plants, while we use phloem-feeding insects and use herbage as the host plants. The SoyFACE experiments have demonstrated that elevated O 3 has no impacts on soybean aphid (Aphis glycine) numbers, and that the effects of simultaneously elevated CO 2 and O 3 on aphid are similar to that of elevated CO 2 alone 79 . Thus, the responses of plant growth forms and feeding guilds to atmospheric changes seem to be heterogeneous.
The decreased time the pea aphids and cowpea aphids spent on phloem feeding under simultaneously elevated CO 2 and O 3 may result in reduced direct damage to the plants. In addition, the aphid feeding activities can also be used to predict indirect damage caused by transmitting plant virus. The piercing-sucking mouthparts of aphids facilitate the delivery of virions into plant cells without causing irrevocable damage 80 . Among a series of aphid feeding activities, the potential drops (pd) and phloem feeding (E) are relevant to virus transmission 81 . Elevated CO 2 and O 3 significantly influenced the time aphids spent on potential drops (pd) and the total number of potential drops (pd), which in turn altered the transmission of stylet-borne viruses by aphids. However, more evidence is needed to demonstrate how the loss of plant productivity caused by virus transmission will be altered under future atmospheric changes.
In summary, elevated CO 2 and O 3 have the potential to affect aphid feeding behaviors via the alteration of plant secondary metabolites, and the responses of aphids to climate changes depend on aphid species and plant developmental stage. This study has generated several significant findings. First, it provides evidence that the heterogeneous responses of aphids to atmospheric changes may result from the differential responses of aphids to the chemicals. Second, direct damage and population outbreaks of aphids may be decreased under future atmospheric conditions due to the reduced efficiency of phloem ingestion under simultaneously elevated CO 2 and O 3 . Finally, and perhaps most importantly, the responses of aphids to elevated CO 2 and O 3 alone or in combination are different and vary with plant developmental stage, suggesting that multi-factor and long-term research is needed. More research is needed to further elucidate the mechanisms underlying the effects of elevated CO 2 and O 3 on herbivores and the role of plant secondary metabolites in the adaption of aphids to future atmospheric environments.  Table 3. Effects of CO 2 , O 3 , plant developmental stage, and their interactions on aphid feeding activities. F and P values from MANOVAs are shown. P values < 0.05 are bolded. np (nonpenetration), stylets are outside the plants; C (pathway), mostly intramural probing activities between mesophyll or parenchyma cells; E (phloem feeding), aphids are injecting watery saliva into the sieve element and ingesting the phloem sap; pd (potential drops), aphids briefly puncture cells during plant penetration. The alfalfa cultivar ' Algonquin' was purchased from the Chinese Academy of Agricultural Sciences. The ' Algonquin' seeds were sown in sterilized soil and watered every 4 days. After the seedlings had grown in sterilized soil for 2 weeks, they were transplanted into plastic pots (17 cm diameter and 24 cm height) and placed in the OTCs. Pot placement was re-randomized within each OTC once per week. No chemical fertilizers or insecticides were used. After the plants were fumigated for 3 weeks (vegetative stage) or 8 weeks (reproductive stage), they were used for the assays described in the following sections.
Plant growth traits. Six plants per OTC (18 plants for each treatment and 72 plants in total) were randomly selected for measurement of growth traits. The leaf chlorophyll content was determined with a Minolta SPAD-502 plus (Konica Minolta Sensing Inc., Osaka, Japan). The leaf net photosynthetic rate was determined with a Li-Cor 6400 gas exchange system (Li-Cor Inc., Lincoln, NE, USA) between 9:00 hours and 12:00 hours. These plants were observed every day for flowering and podding after they had grown for 2 months. The remaining 12 plants per OTC (six for the vegetative stage, six for the reproductive stage, and 144 plants in total) were harvested for measurement of the biomass. The shoots and roots of each plant were collected, oven-dried (50 °C) for 48 h, and weighed.

Aphid feeding behaviors. Seven plants per OTC (21 plants for each treatment and 84 plants for each aphid)
were randomly selected as host plants at the vegetative stage to evaluate aphid feeding behaviors using the electrical penetration graph (EPG) technique. Another 84 plants were also randomly selected at the reproductive stage. The principle of EPG was introduced by Tjallingii and Hogen-Esch 82 . Eight plants were placed in a Faraday cage to avoid noise and interference. Each plant was infested with one apterous adult aphid, and its feeding behavior was recorded for 8 h. The aphids were starved for 10 h before the test. Two eight-channel amplifiers simultaneously recorded 16 individual aphids on separate plants (four plants per treatment). Twenty-one biological replicates were included for each treatment. The feeding waveforms in this study were scored according to Tjallingii and Hogen-Esch 82 : nonpenetration (np), stylets are outside the plants; pathway (C), mostly intramural probing activities between mesophyll or parenchyma cells; phloem feeding (E), aphids are injecting watery saliva into the sieve element and ingesting the phloem sap; potential drops (pd), aphids briefly puncture cells during plant penetration.
Plant secondary metabolites. The chemical analysis was determined according to Oleszek and Stochmal 83 and Nour et al. 84 with some modification. Freeze-dried leaves were ground into a fine powder. For a typical extraction, approximately 50 mg samples were soaked with 1.5 mL of 70% aqueous MeOH for 1 h in a 60 °C water bath. The extract was centrifuged at 12,000 rpm for 15 min, and the supernatant was filtered with a 0.22 μm filter. The samples were stored in a −20 °C freezer until chemical analysis. Using high-performance liquid chromatography  Table 4. Relationships between secondary metabolite contents in the leaves of alfalfa and the time aphids spent on E (phloem feeding) and pd (potential drops). Correlation coefficient r and P values are shown. Values in bold indicate a significant correlation. E (phloem feeding) indicates that aphids are injecting watery saliva into the sieve element and ingesting the phloem sap; pd (potential drops) indicates that aphids briefly puncture cells during plant penetration.
(HPLC), we analysed 12 phenolic compounds: phenolic acids, including chlorogenic acid, caffeic acid, cinnamic acid, p-coumaric acid, and ferulic acid; flavonoids, including rutin, luteolin, apigenin, kaempferol, and quercetin; and isoflavones, including genistein and genistin. Among these compounds, chlorogenic acid, caffeic acid, and . The time pea aphids and cowpea aphids spent in various feeding activities on treated (+genistin, +apigenin, +ferulic acid, and +p-coumaric acid) and control plants. (a-g) Various feeding activities of pea aphids, (h-n) Various feeding activities of cowpea aphids. 'Phloem feeding' indicates that aphids are injecting watery saliva into the sieve element and ingesting the phloem sap; 'potential drops' indicates that aphids briefly puncture cells during plant penetration; 'nonpenetration' indicates that stylets are outside the plants; 'pathway' indicates that mostly intramural probing activities between mesophyll or parenchyma cells.
Values are the means (±SE) of 15-20 biological replicates. Significant differences: *there was significant difference between the treatment and control at P < 0.05.
SCIentIfIC REPORts | (2018) 8:9964 | DOI:10.1038/s41598-018-28020-w cinnamic acid were not detected in the alfalfa leaves. Determination of compounds was performed on a Waters system with a diode array detector. Chromatograms were registered and integrated at 280, 350, and 254 nm for phenolic acid, flavonoids, and isoflavone, respectively. The mobile phase consisted of 1% H 3 PO 4 -AcN (a linear gradient of 15-100% AcN) with a flow rate of 1 mL/min for 60 min. Compounds were identified by comparing retention times to those of authentic standards.
Bioassays with pure compounds. According to the chemical analysis, elevated CO 2 and O 3 had significant impacts on foliar rutin, genistin, ferulic acid, p-coumaric acid, genistein, and apigenin contents of alfalfa ( Table 2). As aphid feeding behaviors were altered under elevated CO 2 and O 3 , and as aphid feeding activities were significantly related to genistin, ferulic acid, p-coumaric acid, and apigenin (Table 4), we performed bioassays to test whether the four secondary compounds affected the aphid feeding. All test compounds were commercially available products. For the bioassay, differing from the in vitro detached leaves and artificial diets 73,85 , we applied the pure compounds to the living plant leaves to provide aphids with the most realistic feeding conditions. All the bioassays were performed on alfalfa plants in a greenhouse. Twenty plants were randomly selected for the experiment. The leaves in the same location were separately treated with 50 μL genistin, ferulic acid, p-coumaric acid, and apigenin solution. Five biological replicates were included for each treatment. Another five plants were selected as controls. The treated and adjacent systematic control leaves were collected to measure the concentrations of the four compounds at 24 h, 48 h, and 72 h after treatment. Contents of genistin, ferulic acid, p-coumaric acid, and apigenin were much higher in the treated leaves compared to the control leaves. The variation tendency remained similar at 24 h, 48 h, and 72 h (see Supplementary Table S1 online). Thus, we selected 24 h to assess the effects of compounds on aphid feeding behaviors using EPG as above.

Statistical analysis.
We analysed the univariate responses of the growth traits, secondary metabolite contents and aphid feeding activities with a split-plot design using the following model Y ijkl = b i + C j + O k + CO jk + e ijk + D l + CD jl + OD kl + COD jkl + ε ijkl . In this model, b represents block i = 3, C represents CO 2 level j = 2, O represents O 3 level k = 2, e ijk represents the whole-plot error, D represents developmental stage l = 2, and ε ijkl represents the sub-plot error. Y ijkl represents the average response of block i, CO 2 level j, O 3 level k and developmental stage l (SAS 9.2, USA). Effects were considered significant when P < 0.05. LSD multiple range tests were used to separate means when ANOVAs were significant.