The effects of drought stress and type of fertiliser on generalist and specialist herbivores and their natural enemies

Abiotic stresses such as drought and nutrient availability can affect invertebrate herbivores feeding on plants, and potentially cascade up to impact their predators and parasitoids. Although these two factors separately been the subject of many studies, there are few tests of their combined effects in the context of pest species and their natural enemies on cultivated plants. Climate change models predict an increase in the frequency and severity of droughts, while the type and amount of fertiliser applied to crops is more under the control of growers. Understanding how these two abiotic factors may interact is key to utilising the potential of natural enemies to control pests under a future climate. To address this, a range of drought and fertiliser type treatments were applied to a model Brassica system in a factorial design, and the performance of two ubiquitous aphid species and their parasitoids was assessed. One aphid species was a specialist on Brassicas ( Brevicoryne brassicae , with parasitoid Diaeretiella rapae ) and the second a generalist aphid species ( Myzus persicae , with parasitoid Aphidius colemani ). The performance of both aphid species responded in a similar way to the treatments, and was maximised on plants growing in organic fertilisers under medium levels of drought stress. The strongest effects of drought and fertiliser cascaded up to affect parasitoids. Parasitoid performance responded in a broadly similar way to their aphid host performance in relation to fertiliser type. Some of the smaller effects of fertiliser treatments on aphid performance were not found for parasitoid performance. Aphid performance was greatest on plants under medium drought stress, but the parasitoids only responded consistently to the high drought stress treatment, on which their performance was reduced. Interactions between the drought and fertiliser did not have a large effect on aphid or parasitoid performance, compared with the strong main effects found for each treatment. These results are discussed in the context of previous and future research on the impacts of abiotic stresses on invertebrate herbivores and their natural enemies.


| INTRODUCTION
Abiotic stresses such as water availability , and the type and amount of nutrients supplied to a plant (Rowen, Tooker, & Blubaugh, 2019), have been shown to separately alter the performance and abundance of invertebrate herbivores feeding on plants, and can cascade up to affect their predators and parasitoids (Tariq, Wright, Bruce, & Staley, 2013a). In the context of pest species feeding on cultivated plants, the nutrients and water received by a host plant can affect populations of pests Tariq et al., 2012) and the behaviour and performance of their natural enemies (Staley et al., 2011a;Tariq, Wright, et al., 2013a), thereby altering the amount of damage a plant receives and potentially also the crop yield.
Under current climate change predictions, rainfall is forecast to become increasingly variable in many regions of the globe, resulting in both more frequent periods of drought and more extreme precipitation events (Field et al., 2014). Several hypotheses have addressed the relationship between drought and plant-herbivore interactions. The plant stress hypothesis predicts that herbivore performance and populations increase under a range of "stress conditions," including drought, because of an increase in the availability of foliar nitrogen (White, 1969(White, , 1984. This hypothesis was based largely on observational studies, in which phytophagous outbreaks were shown to coincide with dry, warm weather. Experimental tests of the plant stress hypothesis have produced varying results (Bultman & Faeth, 1987;Wheatley, Wightman, Williams, & Wheatley, 1989). Larson's modified plant stress hypothesis (the insect performance hypothesis) predicts that drought-stressed plants have a nonlinear relationship with herbivore performance, which is likely to be maximised under medium stress (Larsson, 1989). In the plant vigour hypothesis Price (1991) predicts that vigorous plants are more suitable hosts for herbivores than drought stressed plants because of the higher availability of nutrients (Cornelissen, Fernandes, & Vasconcellos-Neto, 2008;Price, 1991), lower abscission rates (Preszler & Price, 1995) or higher osmotic potential in vigorous shoots (Price, 2003). Mopper and Whitham (1992) suggested that the duration of stress application may influence a herbivore's response to host-plant stress. Huberty and Denno (2004) built on this idea to put forward a "pulsed stress hypothesis," proposing that herbivores (particularly phloem feeders) benefit from plants that are periodically drought stressed, but perform less well on constantly stressed plants, because the loss of turgor pressure restricts the ability of sap feeders to access any increased nitrogen concentrations in the phloem. Tariq et al. (2012) tested the insect performance and pulsed stress hypotheses for two aphid species feeding on Brassica plants under drought stress treatments differing in severity and duration. Their results supported the insect performance hypothesis, but not the pulsed stress hypothesis .
Fewer studies have assessed effects of drought on the performance or attack rates of parasitoids. Aslam, Johnson, and Karley (2013) found parasitism of the bird cherry-oat aphid, Rhopalosiphum padi L., was reduced on barley (Hordeum vulgare L.) plants under drought treatments. Weldegergis, Zhu, Poelman, and Dicke (2015) found parasitoids on Mamestra brassicae caterpillars feeding on Brussels sprout (Brassica oleracea L. var gemmifera) were not affected by drought treatments.
As sustainable farming practices become more widespread (Adnan, Nordin, Bahruddin, & Tareq, 2019), interest is increasing in the effects of the type of fertiliser applied to crops on populations of herbivores and their natural enemies. A recent review of the effects of animal manures on arthropod pests found evidence that the use of animal manures, as opposed to conventional mineral fertilisers, can reduce the performance and abundance of pest species (Rowen et al., 2019). The use of animal manures was found to alter both bottom up processes, through changes to concentrations of both macro and micro-nutrients in plants, and to alter top down processes, because of changes in plant volatile signals to natural enemies and the soil surface habitat used by predators (Rowen et al., 2019).
An understanding of how drought may interact with other abiotic factors, such as the type or amount of fertiliser supplied to crops, would help to inform growers of the increased risks from pests which may occur under future climate change, and potentially to plan adaptation such as increased irrigation or use of more drought resistant varieties, in combination with different fertiliser types. However, to date the majority of studies testing the responses of pest species and their natural enemies to drought and the type of fertiliser supplied to their host plants have looked at each abiotic factor separately (Cividanes, Silva, Martins, & Cividanes, 2020;Stafford et al., 2012;Staley et al., 2011b;Tariq, Rossiter, Wright, & Staley, 2013b). Assessments of whether the two factors may interact to alter pest and natural enemy performance or behaviour are rare, although Griffith and Grinath (2018) tested the interacting effects of drought and conventional fertiliser on a tobacco system.
The family Brassicaceae contains several important crop species, and includes cabbages (B. oleracea L. var. capitata, Brassicaceae: Brassicales) which are grown in many countries (Hopkins, van Dam, & van Loon, 2009  . In contrast, Khan, Ulrichs, and Mewis (2010) found M. persicae abundance was more affected by drought stress than the abundance of B. brevicoryne when feeding on kale, perhaps because of differences between the two crops or because only one level of drought was applied.  (Stafford et al., 2012). These responses may be because of effects of fertiliser on the host plant chemistry, as B. oleracea var capitata fertilised with animal manure produced more glucosinolates, a group of chemicals which the specialist B. brassicae is well adapted to, than those plants growing in conventional AN fertiliser (Kazana et al., 2007;Staley et al., 2010).
Natural enemies of these two aphid species respond to both drought stress and the type of fertiliser (Banfield-Zanin, Rossiter, Wright, . Tariq, Wright, et al. (2013a) found that two parasitoid species (Diaeretiella rapae McIntosh, Hymenoptera: Braconidae and Aphidius colemani Viereck, Hymenoptera: Aphidiidae) parasitized a smaller percentage of potential aphid hosts, had lower percentage emergence, produced more males and had reduced longevity when they developed in aphids feeding on highly drought stressed plants. In contrast, Romo and Tylianakis (2013)  Here, we test the combined effects of drought stress and fertiliser type on the performance of B. brassicae and M. persicae, and the performance of two parasitoid species (D. rapae and A. colemani respectively). Growers may apply more fertiliser to irrigated plants.
However, in some areas including arid and semi-arid countries, water is at times in short supply (Feng et al., 2020). As droughts increase in both severity and frequency, the potential to compensation for drought with increased irrigation may be limited. In this context, understanding how pest aphid species and their parasitoids respond to fertiliser type under drought is critical. To achieve our objective, B. oleracea var. capitata plants were grown under three drought treatments (two levels of drought and an unstressed control), and in three types of fertiliser (including those used in conventional and organic horticulture) and an unfertilized control. The two treatments were applied in full factorial combination to plants growing in pots, and the performance of aphids and parasitoids assessed under controlled environment conditions. Specifically, we tested the following hypotheses: 1. Performance of both aphid species will be increased when feeding on plants under medium drought stress compared to unstressed plants or those under high drought stress.
2. The two aphid species will respond differently to fertiliser type.
Brevicoryne brassicae performance will be increased on plants grown in organic fertilisers, while the performance of M. persicae will be reduced.
3. The combination of medium drought stress and organic fertiliser type will result in the optimum performance of both B. brassicae aphids and their parasitoids D. rapae. In addition, D. rapae performance will be reduced under high drought stress.
4. In contrast, performance of the generalist aphid M. persicae and its parasitoid A. colemani will be maximised on plants growing under medium drought in conventional fertiliser. Aphidius colemani performance will be reduced under high drought stress.
Myzus persicae and B. brassicae were collected from an infested field of Brassica juncea L. var. raya cv Anmol. Aphid cultures were maintained separately on B. oleracea stage 4 plants (BBCH-19; Andaloro, Rose, Shelton, Hoy, & Becker, 1983;Meier, 2001) grown in pots (15.5 cm diameter) with unfertilized compost containing peat, loam, sand and grit (3:3:2:1 ratio). Both aphid species were subcultured fortnightly and transferred to fresh plants. The parasitoid species A. colemani and D. rapae were collected from the same field as the aphids, and were reared on M. persicae and B. brassicae separately for at least two generations before being used in the main experiments, so that maternal host plant effects were avoided (Douloumpaka & van Emden, 2003). The insect cultures were maintained under controlled conditions (20 ± 2 C; 75% RH; LD 16:8 hr).

| Preliminary drought stress treatment trial
To assess the influence of different intensities of drought stress on leaf water content, four drought stress treatments were applied to B. oleracea plants, based on experience from previous studies on drought . Seedlings with first true leaves (BBCH-11; Meier, 2001) were transferred to pots with unfertilized compost. Four weeks after transplanting, four water regimes (100, 200, 300 and 400 ml/week) were applied to stage 2 plants with five true leaves (BBCH-15). Plants were randomised in plastic trays on benches in a controlled environment room (20 ± 2 C; 75% RH; LD 16:8 hr). At stage 4 (9-12 true leaves; BBCH-19; 4 weeks of drought treatments), the plants were tested for leaf water content using the method in Grace (1997). The results of this preliminary study were used to select drought stress treatments for the main experiment. All the plants were dead at 100 ml/week drought regime and thus this treatment was eliminated from the main experiment; plants were highly stressed at 200 ml/week, moderately stressed at 300 ml/week but no drought stress was observed at 400 ml/week water level.

| Experiment treatments
Treatments were allocated randomly to B. oleracea plants. Fertiliser treatments were applied first, through the addition of fertiliser to potting compost prior to transplanting B. oleracea seedlings into the plots.
Plants were left to grow in the allocated fertiliser treatment before drought treatments were applied (details below).
Fertiliser treatments were selected as examples of frequently used types of organic and conventional fertilisers (Hanafi et al., 2018;Pope et al., 2012;Staley et al., 2010). Three types of fertiliser were applied by adding one of the following to 1 L of potting compost: (a) 6.67 g of CM with 4.5% nitrogen, (b) 5.88 g of spent mushroom compost (SMAS) with 7% nitrogen, (c) 0.86 g of AN with 34.5% nitrogen and (d) for the unfertilized control treatment nothing was added to the potting compost. After the fertiliser treatments were added to the compost, a B. oleracea seedling with first true leaves (BBCH-11) was transplanted into each pot. After transplanting, plants in pots were placed in plastic trays on benches in a controlled environment room (20 ± 2 C; 75% RH; LD 16:8 hr).
Three drought regimes were established on stage 2 plants with five true leaves (BBCH-15) as described above for the preliminary trial. The quantity of water added per pot per week was 400 ml for unstressed control plants, 300 ml for medium drought stress and 200 ml for high drought stress. Each drought treatment was applied in factorial combination with the four fertiliser treatments giving 12 factorial combinations of the two treatments. Each factorial combination was replicated five times for each aphid species. After 4 weeks of drought stress treatments, the plants at stage 4 (9-12 true leaves; BBCH-19) were used for aphid or parasitoid performance experiments.

| Aphid performance
Experimental drought and fertiliser treatments were randomly allocated in factorial combinations to plants at the same growth stages as described in Section 2.3, in plastic trays on benches in a controlled environment room (conditions as above). The performance of each aphid species was assessed using 120 B. oleracea plants: 10 replicates of each factorial treatment combination. The 120 plants were evenly split into two spatial blocks, with five plants per factorial treatment combination in each block. Aphid performance was measured as described in Stafford et al. (2012) and Tariq et al. (2012).
Three clip cages were fitted to the underside of first, second and third fully developed leaves on each plant to give 15 clip cages per treatment per aphid species per block. As the performance of alate and apterae can differ (Leather, 1989;Tariq, Wright, & Staley, 2010), the same form of aphid (apterae) was used throughout both experiments. After 16 to 18 hr (overnight), all aphids were removed leaving only one nymph per cage. For each aphid, the prereproductive time (d) and the number of offspring produced in the subsequent days (Md) were recorded and removed daily from each cage.
Data from both aphid species were recorded for the following parameters: 1. the effective fecundity, by allowing the aphids to reproduce during a period equivalent to the prereproductive time and 2. the intrinsic rate of increase r m = 0.738 (lnMd)/d (Wyatt & White, 1977), where Md is the effective fecundity d is the prereproductive time.

| Parasitoid performance experiments
Experimental drought and fertiliser treatments were randomly allocated in factorial combinations to plants at the same growth stages as Five paired parasitoids (one pair of female and male parasitoids per 60 aphids; Jarošík & Lapchin, 2001;Tariq, Wright, et al., 2013a) were released per replicate under ventilated bell cloches. After 24 hr, the parasitoids were removed and the remaining aphids were allowed to develop for 10 to 14 days to allow mummy formation (Tariq, Wright, et al., 2013a;van Emden & Kifle, 2002). Mummified aphids were collected in individual gelatine capsules and percent parasitism, percent emergence, sex ratio (proportion of males) and female hind tibiae length (Blande, Pickett, & Poppy, 2004;Tariq, Wright, et al., 2013a) were recorded.

| Statistical analyses
Within each aphid and parasitoid species, the effect of drought stress and fertiliser treatments and their interaction were analysed using two way factorial ANOVA. Prior to analysis, data for aphid performance (effective fecundity and intrinsic rate of increase) and parasitoid performance (percentage emergence and female tibia length) were log root transformed. Least significant difference (LSD) tests were used to compare mean aphid and parasitoid performance between levels in a factor, to test the hypotheses above. These statistical analyses were carried out using SPSS statistical software version 16.  Figure 1). There was no significant interaction between the effects of drought stress and fertiliser on the fecundity of M. persicae.

| Intrinsic rate of increase
Intrinsic rate of increase for both aphid species followed a similar pattern to aphid fecundity. There were significant differences in intrinsic rate of increase for B. brassicae that depended upon both drought stress (F 2,108 = 294.94, p < .001) and fertiliser treatments (F 3,108 = 59.37, p < .001). Unlike for fecundity, the intrinsic rate of increase for B. brevicoryne feeding on plants growing in AN did not differ significantly from unfertilized control plants (Figure 2). The rate of increase for B. brassicae was also significantly affected by an interaction between drought stress and fertiliser treatments, but the interaction effect was small compared to the main effects (F 6,108 = 3.47, p < .01). Similar to B. brevicoryne, there were significant differences in intrinsic rate of increase for M. persicae of both the drought stress (F 2,108 = 308.99, p < .001) and fertiliser treatments (F 3,108 = 98.34, p < .001).
Intrinsic rate of increase for both aphid species was significantly greatest at medium drought stress, compared with plants in the

| Sex ratio
The main effects of drought stress (F 2,48 = 88.82, p < .001) and fertiliser types (F 3,48 = 11.87, p < .001) were significant for the sex ratio There was no interaction between the effects of drought stress and F I G U R E 2 Intrinsic rate of increase of Brevicoryne brassicae and Myzus persicae (mean ± LSD) feeding on Brassica oleracea L. var. capitata cv. Saint plants under different levels of drought stress (unstressed, medium and high) and fertiliser treatments (AN, ammonium nitrate; C, control without fertiliser; CM, chicken manure; SMAS, spent mushroom compost) fertiliser treatments on percentage emergence for either parasitoid species.

| Female tibia length
The impacts of drought stress (F 2,348 = 23.20, p < .001) and fertiliser treatment (F 3,348 = 8.70, p < .001) were also significant for the tibia length of female D. rapae. Female tibia length of D. rapae was longest at medium drought stress compared with the other two drought  Figure 4d). Female tibia length was also greatest on CM and SMAS, which differed significantly from shorter tibia lengths under both control and AN treatments (LSD, p < .05; Figure 4d). Female tibia length of A. colemani was significantly affected by an interaction between drought stress and fertiliser treatment (F 6,348 = 3.24, p < .001).

| DISCUSSION
The effects of drought stress and fertiliser were broadly similar for the two aphid species. Performance of individual aphids of both species, in terms of their fecundity and intrinsic rate of increase, was greatest at medium drought stress, intermediate at high drought stress and lowest on plants that were not subjected to drought stress, as found previously by Tariq et al. (2012). These results support our first hypothesis above, that performance of both aphid species will be increased when feeding on plants under medium drought stress. More broadly, our results also provide support for Larsson's (1989)  M. persicae, the intrinsic rate of increase also differed significantly between the two types of organic fertiliser (greatest on CM). This similar response shown by the two aphid species to the type of fertiliser is different to previous studies, in which M. persicae abundance  and performance (Stafford et al., 2012) were both greater on plants growing in AN than those in CM, but B. brassicae showed the reverse pattern. This difference may be because of varying composition of different sources of organic fertiliser, or variation in the response of B. brassicae populations from different countries to plant glucosinolate and nitrogen content. Given this previous work, the broadly similar response of the aphid species to fertiliser type was unexpected, and leads to the rejection of our second hypothesis which predicted the two aphid species would respond differently to fertiliser type.
Performance of the two parasitoid species also showed broadly similar responses to both the drought stress and fertiliser treatments.
Parasitoid performance in relation to drought treatments was reduced most under high drought, for all four variables tested. Some performance variables (% parasitism of D. rapae; sex ratio of A. colemani and female tibia length for both species) were greatest under medium drought stress, while the majority of parasitoid performance variables did not differ significantly between medium drought stress and unstressed control plants. While the effects of all three levels of drought treatment significantly affected aphid performance, only the high drought stress treatment cascaded up to consistently reduce parasitoid performance. Tariq, Wright, et al. (2013a) also found performance of these two parasitoid species was reduced under high drought stress compared with unstressed control plants, in line with the results here, but did not assess the effects of medium drought stress. Tariq et al. (2012) found glucosinolate concentrations were increased under drought stress and greatest under high drought stress, using the same plant and aphid species as the current study.
B. brassicae can sequester glucosinolates from their host plants and use them in defence against natural enemies (Pope et al., 2012), so it is possible D. rapae performance was reduced on high drought stress plants because of increased availability of glucosinolates for sequestration. However, M. persicae is not thought to sequester glucosinolates, so this cannot be the mechanism by which the performance of A. colemani is reduced under high drought stress.
Performance of both parasitoid species was better on plants growing in the two organic fertiliser types, compared with both the unfertilized plants and those growing in conventional (AN) fertiliser.
There were some minor additional results to this broad pattern, in that % emergence of A. colemani also differed significantly across all the four fertiliser types, and sex ratio was slightly greater on plants in CM than SMAS. As for the drought treatments, the largest effects of fertiliser type found to affect aphid performance (organic fertiliser vs. conventional or unfertilized) also altered the performance of their parasitoids, in a similar way for the two parasitoid species. However, the smaller fertiliser effects on aphid performance (between AN fertiliser and unfertilized controls) were not consistently found for performance of the two parasitoid species. This shows that while large effects of abiotic treatments on herbivore performance may alter their quality to affect parasitoid performance, not all the smaller effects cascade up to the next trophic level. Pope et al. (2012) found a lower rate of parasitism by D. rapae on B. brevicoryne feeding on Brassica plants grown in CM, when compared with those growing in a conventional John Innes fertiliser. This appears broadly in contrast with results in the current study, perhaps because of differences in fertiliser composition, although the two studies are not directly comparable as the current study did not include John Innes fertiliser.
The third hypothesis above predicted that performance of B. brevicoryne and its parasitoid D. rapae would be maximised on plants growing in organic fertiliser under medium drought stress, and that is supported by our results. However, as discussed in relation to aphid performance, for M. persicae we predicted performance would be greater on plants in conventional than organic fertiliser, and that this pattern would be similar for its parasitoid A. colemani (hypothesis 4). Our results show the two aphid species and two parasitoid species have broadly similar responses to both the fertiliser type and drought stress treatments, so the fourth hypothesis is rejected.
Significant interactions between fertiliser and drought stress treatments were found for most of the aphid performance variables, but the interaction effects were much smaller than the main effects, with interaction F values typically between 10 and 80 times smaller than those of the main treatments. Where interactions were found, generally they showed that differences between fertiliser type were not apparent for one of the three levels of drought stress treatment, rather than indicating opposing effects. Fewer interaction effects were found for parasitoid performance than for performance of their aphid hosts. Broadly, interactions between these two types of abiotic stress do not appear to be as important for the performance of these aphid species and their parasitoids as the separate effects of drought and fertiliser type.
To our knowledge, no previous study has addressed the effects of drought stress and fertiliser type on aphids and their parasitoids in a Brassica system. Griffith and Grinath (2018) assessed the effects of drought and fertiliser amount on an aphid and a caterpillar species on tobacco plants. Both a caterpillar and an aphid species were found to be more responsive to drought than to the amount of fertiliser, although results differed slightly with the tobacco variety (Griffith & Grinath, 2018). Further studies on the interaction between water and nutrient availability on pest species and their natural enemies are needed across a range of crop species, and in field trial conditions, to understand if the patterns found here may be more broadly applicable under a future climate.