Cascading effects of N input on tritrophic (plant–aphid–parasitoid) interactions

Abstract Because N is frequently the most limiting mineral macronutrient for plants in terrestrial ecosystems, modulating N input may have ecological consequences through trophic levels. Thus, in agro‐ecosystems, the success of natural enemies may depend not only from their herbivorous hosts but also from the host plant whose qualities may be modulated by N input. We manipulated foliar N concentrations by providing to Camelina sativa plants three different nitrogen rates (control, optimal, and excessive). We examined how the altered host‐plant nutritional quality influenced the performances of two aphid species, the generalist green peach aphid, Myzus persicae, and the specialist cabbage aphid, Brevicoryne brassicae, and their common parasitoid Diaeretiella rapae. Both N inputs led to increased N concentrations in the plants but induced contrasted concentrations within aphid bodies depending on the species. Compared to the control, plant biomass increased when receiving the optimal N treatment but decreased under the excessive treatment. Performances of M. persicae improved under the optimal treatment compared to the control and excessive treatments whereas B. brassicae parameters declined following the excessive N treatment. In no‐choice trials, emergence rates of D. rapae developing in M. persicae were higher on both optimum and excessive N treatments, whereas they remained stable whatever the treatment when developing in B. brassicae. Size of emerging D. rapae females was positively affected by the treatment only when it developed in M. persicae on the excessive N treatment. This work showed that contrary to an optimal N treatment, when N was delivered in excess, plant suitability was reduced and consequently affected negatively aphid parameters. Surprisingly, these negative effects resulted in no or positive consequences on parasitoid parameters, suggesting a buffered effect at the third trophic level. Host N content, host suitability, and dietary specialization appear to be major factors explaining the functioning of our studied system.


| INTRODUCTION
Trophic cascades, defined as mutual consumer-resource interactions that alter performances of trophic levels across more than one link in a food web (Knight, McCoy, Chase, McCoy, & Holt, 2005), represent not only indirect effects of a higher trophic level on nonadjacent lower trophic levels, that is top-down cascade, but also effects of lower trophic levels that can transmit upward to nonadjacent higher trophic levels, that is bottom-up cascade (Hunter & Price, 1992;Johnson, 2008).
Ecological stoichiometry studies these interactions in ecosystems by investigating how the balance of energy and elements affects and is affected by organisms (Moe et al., 2005;Sterner & Elser, 2002). Indeed, there are differences in the nitrogen (N) content among trophic levels indicating that consumer development would be limited by the amount of N in the resources (White, 1993). The amount of N is an important factor limiting survival and growth of several herbivores and predators (Denno & Fagan, 2003;Mattson, 1980). Therefore, N in host plants and its accumulation by herbivorous insects may influence the performance of natural enemies (Kagata & Ohgushi, 2006). One hypothesis for explaining trophic cascades is that a higher N level in host plants leads to an increase in the N content of herbivorous insects, which in turn leads to an increase in predator performance (Mayntz & Toft, 2001).
In terrestrial agro-ecosystems, nitrogen (N) is an important macronutrient for plants that may vary and be patchy due to soil properties (Keddy, 2007). N availability strongly defines patterns of plant growth and allocation. Generally, in response to enhanced N, plant foliar N concentration, photosynthetic rate, growth rate, and aerial biomass production increase (Hermans, Hammond, White, & Verbruggen, 2006;Leith, Hicks, Fowler, & Woodin, 1999;Magill et al., 2000).
However, excessive N input may induce symptoms indicating toxicity such as chlorosis of leaves, reduced growth, or yield depressions which are due to excessive ammonium (NH4+) in plants (Britto & Kronzucker, 2002). Because N is frequently the most limiting mineral macronutrient for plants in temperate terrestrial ecosystems (Vitousek & Howarth, 1991), modulating N input may have ecological consequences through the different trophic levels.
Generally, improved plant nutritional quality due to enhanced leaf N content (i.e., decreased C:N ratio) has been reported to induce better herbivore development (Awmack & Leather, 2002;Huberty & Denno, 2006;Stiling & Moon, 2005;Zehnder & Hunter, 2008) and higher herbivore population growth rate (Gratton & Denno, 2003). However, using an aphid model and by experimentally manipulating foliar N content, Zehnder and Hunter (2009) have shown that strong N foliar content due to excessive N input may limit herbivore population growth rates.
Within agro-ecosystems, nitrogen can exert a variety of bottom-up effects through the different trophic levels, from the plant to the natural enemies (predators/parasitoids). Development and life-history traits of parasitoids directly depend upon their host quality and, indirectly, upon the quality of the host plant (Schädler, Brandl, & Kempel, 2010;Turlings & Benrey, 1998). Plant nitrogen inputs were reported to directly improve host nutritional quality for parasitoid development through an increase of N concentrations in host bodies (Winter & Rostás, 2010).
Several studies have shown that N plant inputs generally enhance parasitism success (Aqueel et al., 2015;Ponti, Altieri, & Gutierrez, 2007), but to our knowledge, the impact of excessive N input on the third level has not been investigated yet. Nevertheless, N is known to cascade up to higher trophic levels and subsequently alter the abundance and/ or performance of natural enemies, thus altering top-down influences.
This work explores how different levels of N input (no input, optimum input, excessive input) impact host-plant quality (biomass and C:N ratio) and subsequently may alter tritrophic interactions. As aphids have proven to be useful and interesting models for studying important questions in ecology and evolution (Huang & Qiao, 2014) Increasing attention has been paid to camelina, both because of its phyletic proximity to Arabidopsis thaliana (Beilstein, Al-Shehbaz, Mathews, & Kellogg, 2008) and its agronomic characteristics (Bansal & Durrett, 2016). In terms of yield, camelina responded positively to nitrogen input up to 80-120 kgN/ha and negatively to higher N input (Solis, Vidal, Paulino, Johnson, & Berti, 2013). Camelina is generally reported to be tolerant and resistant to various pathogens and insects (Séguin-Swartz et al., 2009), although it has recently been shown to be a potential host for some aphid pests (Chesnais, Verzeaux, Couty, Le Roux, & Ameline, 2015). Among those aphids, B. brassicae L.
In our study, different N fertilization rates (optimum vs. excessive) were used to modulate C. sativa nitrogen availability to determine whether altered leaf N content would cascade up through trophic levels and then would influence the interaction between the two aphid species and their common aphid parasitoid, D. rapae. We expected that insect performances would depend on their host (either the host plant for the aphid or the host aphid for the parasitoid) N concentration. The "tritrophic interactions" (TTI) hypothesis predicts that generalist herbivores should be more sensitive to variations in host-plant quality than specialist herbivores, and thus that the subsequent effects on natural enemies should be more important when the generalist host/prey feeds on low-quality plants (Mooney, Pratt, & Singer, 2012). We therefore expected that the variation in host-plant quality would affect more the parasitoid D. rapae developing on M. persicae than when it developed on B. brassicae.

| Insects
The M. persicae (Sulzer) colony was established from one parthenogenetic female collected in 1999 in a potato field near Loos-en-Gohelle (France).
Diaeretiella rapae (M'Intosh, 1855) parasitoids (Hymenoptera: Aphidiidae) were obtained from IFTECH (Les Ponts de Cé, France) as mummies and were reared on the complex B. brassicae-B. napus. Upon reception, mummies were transferred to plastic tubes (75 × 13 mm) closed with a cotton plug. Once emerged, parasitoids were sexed and mating was allowed by grouping three to four males with six to seven females in the same tube. They were fed ad libitum with a 1:1 honey/ water (v/v) solution until used for the experiments. Parasitoids were maintained in a room at 20 ± 1°C, 60% ± 5% RH, and a 16L:8D photoperiod at 2.5 klux. Three-day-old standardized parasitoid females (mated, fed, and without oviposition experience) and males (mated and fed) were randomly selected for the experiments.
Plants were randomly assigned to one of the three following treatments: (1) no nitrogen input, consisting of the potting soil only (control referred to as "0N"); (2) potting soil supplemented with 0.065 gN/pot (i.e., 80 kgN/ha) of ammonium nitrate (referred to as "80N"), constituting the recommended input for camelina (Solis et al., 2013); or (3) potting soil supplemented with 0.195 gN/pot (i.e., 240 kgN/ha) of ammonium nitrate (referred to as "240N"). Application of the nitrogenous solutions was carried out over the entire soil surface of each pot. A first half-dose, diluted in 100 ml of water, was applied at the sowing date. A second half-dose was applied 1 week later. Four-week-old plants were used for the experiments. Camelina pots were randomly arranged in the growth chamber, independent of their N treatment.

| Nitrogen treatment and aphid performance
Pools of synchronized first-instar nymphs (less than 24 hr old) of each aphid species were obtained from parthenogenetic adult females placed on leaves of B. napus set in 1.5% agar in Petri dishes (90 mm diameter). For the nymph survival study, groups of five-first-instar nymphs were transferred onto camelina plants. Each group of aphid nymphs was enclosed in a clip cage on one leaf at mid-height of each camelina plantlet. Nymph survival was recorded daily. Twenty repetitions per aphid species × N treatment were made. For the adult performance study, for each treatment and each aphid species, 27-30 surviving adults were randomly selected and placed individually in a clip cage on the plant to be tested. Their survival and fecundity were recorded daily. The prereproductive period (i.e., the period of time from birth until the onset of reproduction) of adult aphids was also measured. Their fecundity was assessed for a duration equivalent to twice the prereproductive period duration (PRP) for each aphid species. Newly larviposited nymphs were removed daily (Hackett, Karley, & Bennett, 2013;Le Roux, Saguez, Vincent, & Giordanengo, 2004).

| Nitrogen treatment and aphid size
Thirty-three to thirty-five eight-day-old aphids, randomly selected from the nymph survival study, were measured under a stereomicroscope (LEICA M165C) from the tip of the head to the base of the cauda.

| Nitrogen treatment and aphid C:N ratio
Cohorts of eight-day-old aphids were obtained as described in the nymph survival study. For each aphid species, ten samples of 15-20 aphids per N treatment were frozen and then freeze dried for further C and N content analysis. The same protocol as the one used for leaf plant samples was used.

| Third trophic level-parasitoid
The experimental setup used was modified from Boquel, Delayen, Couty, Giordanengo, and Ameline (2012). It consisted of six ventilated Plexiglas ® chambers (360 × 240 × 110 mm) used simultaneously, inside which, three plants of C. sativa (1 plant per pot) belonging to one of the three treatments ("0N", "80N" or "240N") were placed. In each chamber, 180-200 three-day-old nymphs, which had been reared on camelina from the corresponding treatment, were deposited evenly on the three plants. Groups of five to six males and 10 females standardized D. rapae wasps were introduced with a small paintbrush in the middle of the experimental setup for 24 hr before being removed.
Aphids were observed daily until formation of a mummy. Mummies were transferred into plastic tubes (75 × 13 mm) closed with a cotton plug. Emerged parasitoids were sexed and then stored at −80°C for further measurements. Left hind tibia length was used as a proxy for parasitoid size. It was measured under a stereomicroscope (LEICA M165C) on 30-35 randomly selected parasitoid females for each N treatment × host aphid species combination.
The following parasitoid life-history parameters were then computed: (1) nymphal developmental time (from mummification to adult emergence) in days; (2) emergence rate expressed as a percentage ((no. of emerged parasitoids/no. of mummies) × 100); and (3) tibia length (in mm) of parasitoid females.

| Statistical analyses
Mean values are given with their standard error of the mean (SEM).
The effect of treatment on plant (biomass and C:N) and aphid (PRP, fecundity and C:N) data that were not normally distributed was analyzed using a Kruskal-Wallis one-way analysis of variance (H), followed by multiple comparison tests using the R package "nparcomp" (type: Tukey). Aphid survival was analyzed using a Cox-proportional hazards model. Aphid length data being normally distributed, the effect of N treatment on this parameter was analyzed using a parametric one-way ANOVA test followed by pairwise comparisons using permutational t tests (p-value adjustment method: fdr, 999 permutations, package "RVAideMemoire"). The combined effects of aphid host species and N treatment on parasitoid parameters were analyzed using GLM with binomial error (link: logit) for emergence rate, as the use of binomial distribution is appropriate to model binary data or percentages; GLM with Gaussian distribution (link: identity) for the hind tibia length and Poisson distribution (link: log) for the development time. All statistical analyses were carried out using the statistical program "R" (R 3.2.2-R Development Core Team, 2015).
Compared to the control "0N" treatment, the aerial biomass of Camelina sativa plants was significantly higher under the "80N" treatment and significantly lower under the "240N" treatment ( Fig. 2A).
Both N treatments ("80N" and "240N") induced a significantly lower foliar C:N ratio compared to the control (Fig. 2B).
Development time was not significantly affected by the plant treatment but was affected by the aphid species in which D. rapae females developed, with reduced development time inside B. brassicae (χ 2 = 11.515, df = 573, p = .0007) (Table 2 a and b). There was no interaction between the two factors.
Hind tibia length was significantly affected by the N treatment (χ 2 = 0.007, df = 190, p = .038) but not by the aphid species F I G U R E 3 Survival rates of aphid nymphs of Myzus persicae and Brevicoryne brassicae on camelina plants exposed to different nitrogen treatments ("0N", "80N," and "240N") T A B L E 1 Life-history traits (Mean ± SEM) of (a) Brevicoryne brassicae and (b) Myzus persicae aphids reared on Camelina sativa exposed to different nitrogen treatments (0, 80, and 240 kgN/ha)

| DISCUSSION
Nitrogen inputs differently affected N concentrations within the first two trophic levels and the subsequent bottom-up effects within our tritrophic system. Contrary to an optimal N application, when N was delivered in excess, the plant was stressed and its biomass decreased although N leaf concentration increased. Aphids parameters were negatively affected suggesting that the suitability of host plants receiving an excess of N was therefore reduced. Surprisingly, these negative effects resulted in no or positive effects on parasitoid parameters, suggesting a buffered effect at the third trophic level. Host N content, host suitability, and dietary specialization appear to be major factors explaining the functioning of our studied system.

| Nitrogen host content and performances of superior trophic levels
Nitrogen input affected severely the biomass and physiology of camelina. In our study, plants receiving the optimal "80N" treatment had an increased aerial biomass and a decreased leaf C:N ratio, due to increased leaf N concentration (data not shown). Our results are in accordance with previous works (Hermans et al., 2006;Leith et al., 1999;Magill et al., 2000). For the excessive "240N" treatment, even if the C:N ratio decreased, plant biomass decreased indicating a stressed status of the plants. In some cases of excessive N input, ammonium (NH4+) stress may induce symptoms such as reduced growth or yield depressions (Britto & Kronzucker, 2002) and this type of stress may then cascade up negatively on the 2nd trophic level such as observed in our study.
Previously published papers have shown positive relationships between high N concentration in the host plant and high performances of phytophagous insects in terms of population growth rate (Couture, Servi, & Lindroth, 2010), weight increase (Winter & Rostás, 2010), or both (Aqueel & Leather, 2011). Consistent with these previous works, in our study, M. persicae size and fecundity and B. brassicae size were positively impacted by the higher leaf N concentration obtained in the optimal "80N" treatment. However, increased leaf N concentration obtained in the excessive "240N" treatment did not affect positively M. persicae size and fecundity nor B. brassicae fecundity. A study from Zehnder and Hunter (2009) examining the impact of experimental increases of foliar N levels has documented a diminution in performance of an aphid species, suggesting that excessive nutrient levels can limit herbivore population growth rates. This may occur when nutrient additions lead to host plants presenting nutrient ratios that exceed herbivore threshold elemental ratios (Boersma & Elser, 2006). To maintain homeostasis, herbivores need therefore to excrete excess elements, and if the excretion is energetically costly, then they may exhibit reduced population growth rates (Boersma & Elser, 2006). Other physiological mechanisms (reviewed in Zehnder & Hunter, 2009) such as a decrease of feeding rate, an alteration of assimilation efficiency, or the impact of plant composition on aphid symbionts, may also explain how excessive N concentrations may result in decreased aphid performances. Fagan et al. (2002) have revealed an imbalance between herbivorous and parasitoid insect N concentrations, indicating that N limitation can influence higher trophic levels interactions (Denno & Fagan, 2003 Therefore, in accordance with Aqueel et al. (2015), our study suggests that the greater is the N concentration in the body of the aphid host, the better is the parasitoid emergence rate. Effects of N input on development rate of parasitoid seem to be variable as both positive (Garratt, Leather, & Wright, 2010) and no effects (Aqueel et al., 2015) have been previously observed. In our study, whatever the N host content, the developmental duration of parasitoids was not affected.
Interestingly, the size of female parasitoids was increased when developing in M. persicae in the excessive "240N" treatment, potentially enhancing the level of pest control they may provide as size is generally positively correlated to their fitness (e.g., egg load) (Jervis, Ellers, & Harvey, 2008).

| Variation of N input and the "tritrophic interaction" hypothesis
According to the "tritrophic interactions" (TTI) hypothesis developed by Mooney et al. (2012): (1) generalist herbivores should be more sensitive to variation in host-plant quality than specialist herbivores; and (2) a low-quality host plant should impact more the third trophic level when the herbivore host/prey is a generalist than when it is a specialist. We therefore expected that (1)

| Validation of prediction (1) depends on the level of the N input
At the optimal "80N" treatment, consistent with prediction (1), the generalist M. persicae was more affected than the specialist B. brassicae.
Fecundity and size of M. persicae were enhanced whereas in B. brassicae only the size was enhanced. However, unexpectedly, under the excessive "240N" treatment (i.e., stressed plant), prediction (1) was invalidated. M. persicae fecundity and size were not affected whereas, paradoxically, fecundity was reduced and, in the specialist aphid B. brassicae, size was enhanced. Both N treatments may have influenced host-plant resistance to the second trophic level through the alteration of primary and secondary metabolites production (Chen, Schmelz, Wäckers, & Ruberson, 2008;Lou & Baldwin, 2004;Slansky 1992). For both N treatments, a decrease in the C:N ratio compared to the control treatment may be related to a higher percentage of amino acids in the phloem sap (Crafts-Brandner, 2002). Extensive studies of aphid physiology have revealed the central role of sucrose concentration, amino acid concentration and composition, and sucrose:amino acid ratio in shaping aphid performances (Auclair, 1963;Douglas, 1998;Febvay et al., 1988).
The better performance of M. persicae may therefore possibly be attributed to an increase in the quantity and/or quality of amino acids in the camelina plants submitted to the optimal "80N" treatment.
Several works have emphasized the role of bacterial symbionts in situation of low-N availability. Indeed, aphids have bacterial symbionts that enable them to modify the quality of the ingested phloem sap by upgrading nonessential amino acids to essential amino acids (Douglas, 1998), enabling them to maintain high reproductive rates on diets with low concentrations of low-quality nitrogen. However, there is no indication of how these microbial symbionts would behave in case of excessive N input and how this might alter their relationship with the aphids (Zehnder & Hunter, 2009).
Besides, high rates of N application may increase plant demand for water and then increase the possibility of water stress (Scagel, Bi, Fuchigami, & Regan, 2011). Many phytophagous insects, especially sap-feeders, are adversely affected by continuous water stress (reviewed in Huberty & Denno, 2004). During periods of plant stress, reductions in turgor and water content may also interfere with the herbivore's ability to access or utilize nitrogen. This effect may also partly explain the limited fecundity of B. brassicae on the "240N" treatment.
Depending on the plant species, N treatment may also affect the synthesis and levels of constitutive and induced defensive secondary compounds (Chen et al., 2008;Dudt & Shure, 1994;Lou & Baldwin, 2004). For example, foliar glucosinolate content in Brassica crops is influenced by fertilizer treatments (Staley et al., 2010, 2011) and generally decreases in reaction to N input (Schonhof, Blankenburg, Müller, & Krumbein, 2007). The specialist aphid, B. brassicae, is dependent on glucosinolate compounds, which can act as phagostimulants (Cole, 1997a;Kim, Lee, Schroeder, & Jander, 2008) and may be or not affected by GLS in their host plants ( Le Guigo, Qu, & Le Corff, 2011). N input may have induced lower glucosinolate concentrations in camelina leaves and then negatively affected B. brassicae feeding rate, thus leading to a decreased fecundity, whereas M. persicae performances, which are less strongly related to glucosinolate concentrations (Cole, 1997a), were enhanced. This latter result can be related to the way M. persicae deals with GLS. As a generalist aphid, it uses a strategy of toleration, whereby glucosinolates and other plant secondary metabolites are flushed rapidly through the aphid gut and are present in the honeydew (Weber, Oswald, & Zoellner, 1986). Wilkens, Spoerke, and Stamp (1996) found that N fertilization altered levels of phenolic compounds that are associated with plant resistance to aphids (Cole, 1984). The decreased performance of B. brassicae on the "240N" treatment could therefore be attributed to a variation in such phenolic compounds.

| Validation of prediction (2)
Consistent with our prediction (2), whatever the N treatment, D. rapae performances were more affected, and positively so, when it developed on the generalist aphid M. persicae than when it developed on the specialist B. brassicae. It has been shown that plant allelochemicals may not only affect the development of herbivores but also that of their parasitoids (Hunter, 2003;Ode, 2006 Nevertheless, performances improvement (i.e., emergence rate and size of the tibia) of D. rapae females when developing inside M. persicae could be explained by lower glucosinolates levels in camelina leaves exposed to N treatments. The better performances of D. rapae on M. persicae on N-treated plants may be linked to the aphid host nutritional quality, expected to be enhanced on plant exposed to N treatment. Indeed, size of parasitoids is often related to nutritional characteristics of the host (Godfray, 1994) and adult life span and fecundity of the parasitoids are increased by the availability of nutrients in host insects (Azzouz, Giordanengo, Wäckers, & Kaiser, 2004;Kaneshiro & Johnson, 1996).

| CONCLUSION
The originality of our work lies in the fact that although an excessive N input induced the expected negative impacts on the performance on the first two trophic levels, this effect was buffered at the third trophic level. This also highlights the fact that a high host N content does not always warranty a good nutritional quality for the upper trophic level.
Our work shows that according to the "tritrophic interactions" hypothesis, N treatments cascaded up differentially to the third trophic level depending on the aphid host dietary specialization. Variation in host-plant quality had indeed stronger effects on dietary generalist aphids than on dietary specialist ones. Consequently, it appeared that bottom-up forces were relatively more important on the parasitoids when a generalist aphid species was involved than when it was a specialist. This study provides key insight on the complex interactions between bottom-up and top-down forces that structure plant-herbivores communities under the context of agricultural intensification. Indeed, N recommendations based on tests and projected crop N demand often result in over-application (Cui, Zhang, Chen, Dou, & Li, 2010) and it seems that this excessive input, apart from being detrimental in terms of yield for the plant, could also have a significant impact on higher trophic levels.