Aphid performance and population development on their host plants is affected by weed–crop interactions

Authors


Correspondence author. E-mail: Velemir.Ninkovic@slu.se

Summary

  1. Some farming practices, like organic farming, lead to greater numbers of weed plants in crop fields. These fields may give us some insights into any benefits that may be gained from biodiversity (e.g. improved pest control services) and allow us to understand the mechanisms behind crop–weed interactions.
  2. The influence of two common weeds, lamb's-quarters Chenopodium album and charlock Sinapis arvensis, on performance of the bird cherry-oat aphid Rhopalosiphum padi in spring-sown barley Hordeum vulgare is evaluated in three field experiments. Observations in field experiments indicated that the presence of S. arvensis reduced aphid population development in the barley crop significantly, but this effect was not observed in barley grown with C. album.
  3. Observed effects in the field were further studied in laboratory experiments with regard to aphid growth and reproductive performance. Above- and below-ground interactions of S. arvensis and C. album with barley were tested using twin-exposure cages. Aphid performance was negatively affected when barley plants had root contact with S. arvensis. The results of these laboratory experiments showed a difference in mode of action of the two weeds.
  4. Synthesis and applications. The results support the potential of associated resistance, mediated by neighbouring plants, in minimizing herbivore damage of focal plants and highlighted the mechanism by which herbivores might be affected. Since chemical exchange between plant neighbours can potentially occur in any plant community, increased understanding could be valuable for existing and new agroecosystems, invasion biology and sustainable crop production. To get a balance between herbicide and insecticide control, agricultural production systems need to focus on the thresholds of weed and insect tolerance, taking the associated resistance of biodiversity (here weeds) into account. Agricultural biodiversity may provide many long-term benefits over monoculture, from reducing pesticide pollution to preventing insecticide resistance. Our study is an important step forward in general understanding of the effects of vegetational diversity on herbivore population dynamics.

Introduction

Ever tighter regulation of the use of pesticides and the development of pesticide resistance make alternative control strategies against herbivores necessary. One option to promote reduced herbivore abundance is manipulation of the cropping system. It has been shown that increased within-field plant biodiversity offers advantages in terms of pest control (Risch 1983; Vandermeer 1989; Norris & Kogan 2005) and a meta-analysis of 21 studies (Tonhasca & Byrne 1994) supported the hypothesis that diversified crops cause a reduction in the populations of herbivorous insects. Reduced pest prevalence in polycultures may be the result of increased abundance of natural enemies, or decreased colonization and reproduction of pests (Elton 1958; Root 1973; Way & Cammell 1981; Andow 1991; Haddad et al. 2009). Nonhost plants might mask olfactory and visual cues of herbivores to find their host, confuse or repel these insects (Tahvanainen & Root 1972; Root 1973; Stanton 1983; Agrawal, Lau & Hambäck 2006; Randlkofer et al. 2010). However, there are some examples indicating that diversification had no effect, or increased herbivore densities. In a review of 150 studies on the effects of diverse agroecosystems on insect herbivores, Risch, Andow & Altieri (1983) found that 53% of the herbivore species were less abundant in diverse systems, 18% increased, 20% showed a varied response and 8% did not differ between the systems.

Associational resistance occurs when neighbouring plants decrease the amount of damage to a crop plant (Tahvanainen & Root 1972). These effects can arise by neighbours influencing the quality of the crop plant or by altering herbivore traits, such as feeding behaviour and host-plant search patterns. These complex interactions are illustrated in Fig. 1 to provide a visual frame of associational resistance given by neighbouring plants. Understanding the underlying mechanisms of observed changes helps to characterize and comprehend this phenomenon and will be important for a general theory of how diversification affects pest populations and explain further why exceptions to the theory occur. Most published studies on associational resistance are based on how insects (herbivores or their natural enemies) respond to vegetational diversity. Few studies have shown that crop plants grown together with heterogeneous neighbouring plants may change as a host for herbivores (Björkman et al. 2009), and the fact that naturally occurring weeds in fields may have similar implications on crop plants is almost overlooked. Furthermore, most ecological research on biodiversity is made outside agricultural systems, and there is therefore a need for basic research to understand the links between biodiversity, ecosystem function and sustainability (FAO 2011).

Figure 1.

The associational resistance mediated by neighbouring plants on herbivore damage of focal plants and the mechanisms by which herbivores might be affected. The influences represented by linking lines can be positive or negative.

Through their limited mobility, plants face varying types of stress in their environment and one of the most common challenges is coexistence with neighbouring plants. Plants perceive their competitive environment through different stimuli such as light quality (e.g. Franklin 2008), volatile compounds (Ninkovic 2003; Pierik et al. 2004) and root exudates (Bais et al. 2006; Hodge 2009), and react with a range of phenotypically plastic responses to improve resource capture (Callaway, Pennings & Richards 2003; Novoplansky 2009). It has been shown that barley plants exposed to volatiles from a different barley cultivar change their pattern of biomass allocation and leaf temperature (Ninkovic 2003, 2010), indicating that volatile interactions between plants may have considerable effects on physiology and morphology of receiving plants. Most herbivores are sensitive to slight changes in the quality and physiological status of their host plant (Pettersson, Tjallingii & Hardie 2007). Thus, barley plants treated with volatiles and root exudates from different weed species can be less attractive to aphids than untreated plants (Glinwood et al. 2003, 2004; Ninkovic, Glinwood & Dahlin 2009). It is still unknown whether these interactions may have implications on aphid performance and their population development under field conditions.

The aim of this study was to investigate whether the effects of botanical diversity on herbivores can be a consequence of induced changes in the host plant brought about by competitive interactions between plant species. As interactions between plants can take different forms, we investigated whether induced plant response occurs due to above- or below-ground stimuli. Another specific question addressed was whether inducing effects are common or depend on the identity of the inducing species (species specific). Our model system consisted of interactions between two of the most common weed species in Europe and barley Hordeum vulgare (L.). In addition, interactions between barley and one of its herbivores, the aphid Rhopalosiphum padi (L.), were investigated in our model system. Rhopalosiphum padi is a major pest of cereals in Sweden (Wiktelius, Weibull & Pettersson 1990) and an excellent model herbivore for detecting changes in plant status following interactions between plants (Ninkovic, Glinwood & Pettersson 2006). The model weed species were chosen on basis of a previous study which showed that Chenopodium album (L.) induced a response in barley plants and reduced aphid acceptance of them, whereas Sinapis arvensis (L.) had no effect (Ninkovic, Glinwood & Dahlin 2009). The present study highlights the impact of interspecific plant–plant interactions on herbivores in the field with implications for the role weeds can play in pest control in agricultural systems.

Materials and methods

Field Experiments

In 2006 and 2008, field experiments in spring barley were conducted at Sätuna (60o3′N, 17o4′E), and in 2007, at Fransåker (59o38′N, 17o49′E), Sweden. In accordance with common agronomic practice in the region, they were sown in the beginning of May. The plots were arranged in a conventional randomised block design in four blocks, which represent the replicates of each treatment. Three different treatments were compared in each block: barley in pure stand, barley grown with S. arvensis and barley grown with C. album. Plot size was 1·3 × 2 m in 2006, 2·4 × 4·8 m in 2007 and 1·4 × 2·5 m in 2008. Barley (cv. Scandium) was sown with a seed rate of 200 kg ha−1 and a row width of 12·5 cm. Seeds of the weeds S. arvensis and C. album were broadcast by hand a day after barley sowing with a seed rate of 38 g m−2 for C. album and 77 g m−2 for S. arvensis, calculated to reach a plant density of 500 plants m−2, which is the prevalent weed plant density in organic cultivation in Sweden (Lundkvist & Fogelfors 2004). No weeds were sown in the control plots, and any plants which appeared naturally were weeded out regularly by hand. Each plot was harvested, and grain yield (t ha−1) was calculated at a water content of 10–15%.

Estimates of aphid population development in the field

In 2006, the aphid population development was observed counting the number of aphids in a randomly chosen row of one metre length three times in each plot (Ninkovic et al. 2003). Observations were made weekly for four consecutive weeks, and the data collected were adjusted to number of aphids per plant. In 2007 and 2008, because of the dry weather conditions, the aphids colonized mostly the root necks of the plants which made it necessary to change the counting method. Thus, the estimates were done by destructive sampling by pulling 20 randomly chosen plants per plot and counting all the aphids on these plants. The mean value of three estimates of aphids per plant per 1-m row in 2006 and mean value from 20 plants per plot in 2007 and 2008 were pooled weekly and used as one experimental observation per plot to avoid pseudoreplication. The germination and appearance of barley and weeds in the field trial in 2007 were uneven and poor due to prevailing weather conditions. In addition, aphid immigration occurred 3–4 weeks earlier in 2007 as in the other years of field experiments. Since the amount of crop plants was insignificant and consequently inadequate for the establishment of aphids, the experiment could not be completed in 2007.

The mean values for aphid populations were expressed as log (aphid number + 1). Data on estimates of aphid occurrence consisted of repeated measurements made at intervals in the same plots during the experimental period. Therefore, mixed linear models were used for the analyses as suggested by Fitzmaurice, Laird & Ware (2004) and Littell et al. (2006). The dependence between observations over time was modelled using a spatial power covariance matrix. The models included the fixed effect of treatment, block, time and the treatment by time interaction. In the model, interactions between blocks and time points were included as an independent, normally distributed random effect. Least squares means were calculated and compared using Tukey's HSD test. Diagnostic plots were used to analyse the models for normality and homoscedasticity. The Mixed procedure of the SAS (2008) package was used for the analyses.

Differences in grain yields per plot per treatment were analysed by three-way anova with treatment (barley with weeds and without), block and year as factors. Using histogram and plots, data values and their residuals were found to be normally distributed and homoscedastic. Pairwise comparisons of means were achieved using Tukey's HSD test. For the analyses, the general linear model procedure of the SAS (2008) was used.

Laboratory Experiments

Plants and insects

All experimental plants were grown in a greenhouse with L16 : D8 h and 18–22 °C. Seeds for the spring barley (cv. Scandium) were obtained from Svenskafoder, SE. Seeds of the two weeds C. album and S. arvensis were obtained from Herbiseed, UK, and plants were raised in a separate greenhouse chamber for barley plants. For laboratory experiments, pre-germinated seeds of both weed species and barley were sown in different batches to maintain similar size between plants of the species.

A multiclonal population of Rhopalosiphum padi (L.) was reared on oats Avena sativa (L.) and grown in a greenhouse under similar conditions as plants. Aphid cultures were renewed yearly with migrants from the aphid's winter host, to maintain genetic diversity. Experiments were carried out in a glasshouse at a temperature of 18–22 °C, with minimum 16 h light (natural light supplemented by light from HQIE lamps).

Exposure of barley to volatiles from weeds

To test the volatile interaction between weeds and barley, twin chamber exposure cages existing of inducing and responding chambers were used (Ninkovic, Olsson & Pettersson 2002). Clear Perspex cages with two separate chambers (each 10 × 10 × 40 cm) connected by a 7-cm-diameter circular opening in the dividing wall were placed over the plants. Air passed over the single weed plant placed in the inducing chamber and then entered the single barley plant placed in the responding chamber. Weed plants at a height of 10–20 cm were used as inducers. The exposure period was 7 days. The control barley plants were not exposed to any plant volatiles by keeping the inducing chamber empty. Airflow through the cages was 1·3 l min−1. To prevent interaction between weed and barley plants by root exudates, plants were grown in separated plastic pots (10 × 10 × 7 cm) placed separately in Petri dishes. Each plant was supplied with water through an automatically drop system. Each treatment was contained 36 plants. For logistic reasons, the barley was exposed to volatiles from the two different weed species in two separate experiments.

Root interactions between barley plants and weeds

For the root interactions experiments, weed and barley plants were grown together in the same pot (26 × 19 × 7 cm) in potting soil, Special Hasselfors Garden AB, Hasselfors, Sweden. The treatments consisted of a single barley plant together with a S. arvensis plant and one barley plant together with a C. album plant. Pots containing one barley plant and no weeds were used for the control treatment. Barley was sown in the pots when the weeds plants reached a height of 10–20 cm. To prevent any volatile interaction between above-ground parts of the plants, clear Perspex cages (10 × 10 × 40 cm) were placed over each plant. Air entered the cage through an opening (7 cm diameter) and extracted through a tube (10 mm diameter). Each cage was connected with a vacuum tank to vent air out of the cages using an electrical fan. The plants were grown together for 14 days before the aphid test started. Twenty plants were used for each treatment and were randomly arranged on the greenhouse bench. The effects of root interaction of the two weed species on barley were tested in separated experiments.

Test of aphid relative growth rate

In order to study the effects on aphid relative growth rate on barley plants with root contact to weed plants, or exposed to weed plant volatiles, a cylindrical Perspex tube (2 cm diameter, 5 cm length) was placed around the base of the barley plants, and fine sand (0·1–3 mm) was placed on the soil (Ninkovic & Åhman 2009; Ninkovic, Glinwood & Dahlin 2009). Adult apterae were placed on barley seedlings to reproduce for a period of 24 h, yielding a cohort of first-instar nymphs. The nymphs were weighed on a microbalance and released onto the plants. One nymph was then placed into each tube around the base of barley plants and reweighed after 4 days. Mean relative growth rate was calculated by an equation originally used by Radford in Radford 1967:

display math

where W= weight at the first weighing, W2 = weight at the next weighing and t2t1 = the time (days) between first (t1) and second (t2) weighing.

Test of aphid development

A cage arrangement of the same design as used for the studies of relative growth rate was used. A single alate R. padi was placed on each barley plant previously used for the aphid growth rate test. On the next day, only one recently produced nymph was kept on the plant, while the alate and any other produced nymphs were removed. The time from birth to reproducing adult (development rate) was recorded for this remaining aphid.

Test of aphid intrinsic rate of increase

The intrinsic rate of increase (rm) was recorded during the following 5 days counting the number of produced nymphs on each plant. Each day all the newly produced nymphs were removed. The intrinsic rate of increase (rm), relating the fecundity of an individual aphid to its developmental time, was calculated (Wyatt & White 1977):

display math

where Md is the number of nymphs produced by the adult in the first d days of reproduction after the adult moult. The constant (c = 0·738) is an approximation of the proportion of the total fecundity produced in the first days of reproduction.

Whether aphid growth rate, their development rate and the intrinsic rate of increase on barley plants with weed root interaction or exposed to weed volatiles differed from that on control plants, was tested using a t-test for independent samples in the STATISTICA statistical package (SAS 2008).

Results

Field Experiments

The natural incidence of aphids in the field experiments was observed on a weekly basis for 3–6 weeks. In the 2006 field experiment, a statistically significant difference in the number of R. padi on barley plants was found between treatments (F3,12 = 13·76, = 0·0001). The interaction between treatments and time point was not significant (F4,18 = 0·83, = 0·53). Significantly lower numbers of aphids were observed on barley plants grown in plots with S. arvensis than in plots without weeds (= 0·001, Tukey's HSD test), while the number of aphids was not significantly different between plots with C. album and plots with only barley plants (= 0·97, Tukey's HSD test) (Fig. 2a).

Figure 2.

Aphid Rhopalosiphum padi incidence in barley plots with different weed species in field trials 2006 (a) and 2008 (b). Control plots consist of barley without any weeds. Data were transformed as natural logarithms. Error bars indicate ± SEM.

Analysis of field data from 2008 gave a similar pattern as in 2006. Statistically significant differences were found between treatments in the number of R. padi on barley (F2,6 = 6·16, = 0·035). The mean number of aphids on barley grown together with S. arvensis was significantly lower than on barley grown without weeds (= 0·04, Tukey's HSD test), while this reduction was not observed on barley plants in plots with C. album (= 0·90, Tukey's HSD test). The interaction between treatments and time point was not significant (F14,36 = 1·07, = 0·352) (Fig. 2b). These data indicate that the presence of S. arvensis reduced aphid population development in the barley crop, while this effect was not observed in barley grown with C. album.

Significant differences in the grain yield of barley grown with and without weeds were observed (F2,23 = 6·68, P < 0.0001). The difference between years was also significant (F1,23 = 47·57, P < 0.0001), whereas there were no significant differences between blocks (F3,23 = 0·838, = 0·119) and interaction between treatment (barley with weeds and without) and year (F2,23 = 0·71, = 0·086). Average grain yield of barley grown together with S. arvensis (mean 3·3 t ha−1) was significantly lower than the yield of barley grown in pure stands (mean 4·5 t ha−1) (< 0·0001, Tukey's HSD test), whereas this significant reduction was not observed for barley grown together with C. album (mean 4·1 t ha−1) (= 0·055, Tukey's HSD test).

Laboratory Experiments

Effects of plant interactions on aphid mean relative growth rate

The aphid growth rate was significantly lower on barley plants which had root contact with S. arvensis, than on barley plants grown alone (t = −2·14, d.f. = 24, = 0·04). No statistically significant differences in aphid growth rate were found on barley plants having root contact with C. album (t = −0·323, d.f. = 14, = 0·75). No statistically significant differences in aphid mean relative growth rate were found on barley plants exposed to S. arvensis volatiles and control barley plants (t = −0·178, d.f. = 15, = 0·86), as shown in Table 1. These data indicate that S. arvensis did induce associated resistance when having root contact with barley plants, but not via airborne interactions.

Table 1. Mean relative growth rate (μg μg−1 day−1) (±SEM) of Rhopalosiphum padi on barley plants which had root contact with S. arvensis, C. album or barley grown alone, and on barley plants exposed to weed volatiles or unexposed barley plants. Whether aphid growth rate on treated plants differed significantly from that on control plants was tested using a t-test for independent samples (< 0·05) ns = not significant
Type of interactionBarley with weeds (g)Barley alone (g)P-value
Sinapis arvensis
Root interaction0·18 ± 0·010·21 ± 0·0070·04
Volatile interaction0·12 ± 0·0080·13 ± 0·02ns
Chenopodium album
Root interaction0·13 ± 0·0050·13 ± 0·01ns

Effects of plant interactions on aphid development

The development rate of aphids on barley plants which had root contact with S. arvensis was significantly longer than on barley plants grown alone (t = −2·05, d.f. = 38, = 0·047). This effect on aphid development rate was not found on barley plants which had contact with C. album roots (t = −1·73, d.f. = 29, = 0·093). Exposure to volatiles of S. arvensis or C. album did not affect the aphid development rate on barley plants compared with unexposed barley plants (t = −1·251, d.f. = 47, = 0·22) and (t = −0·628, d.f. = 27, = 0·535), as shown in Table 2.

Table 2. Development rate (±SEM) of Rhopalosiphum padi on barley plants which had root contact with S. arvensis, C. album or barley grown alone, and on barley plants exposed to volatiles of these weeds or unexposed barley plants. Whether aphid development rate on treated plants differed significantly from that on control plants was tested using a t-test for independent samples (< 0·05) ns = not significant
Type of interactionBarley with weeds (days)Barley alone (days)P-value
Sinapis arvensis
Root interaction7·4 ± 0·26·9 ± 0·20·047
Volatile interaction6·4 ± 0·16·2 ± 0·1ns
Chenopodium album
Root interaction8·3 ± 0·257·7 ± 0·15ns
Volatile interaction7·9 ± 0·27·7 ± 0·25ns

Effects of plant interactions on the intrinsic rate of increase of aphids

The intrinsic rate of increase of aphids was significantly lower on barley plants which had root contact with S. arvensis than on barley plants grown alone (t = 4·26, d.f. = 38, = 0·0001). Similar results for the intrinsic rate of aphids were found on barley plants exposed to S. arvensis volatiles compared with aphids on unexposed barley (t = 2·38, d.f. = 48, = 0·02), as shown in Table 3. However, no significant differences found in the intrinsic rate of aphids reared on barley plants that were treated with C. album volatiles, neither on barley plants having root contact with C. album compared with their intrinsic rate on untreated barley plants (t = 0·525, d.f. = 26, = 0·604) and (t = −0·731, d.f. = 29, = 0·471), respectively, as shown in Table 3.

Table 3. Intrinsic rate of increase (±SEM) of Rhopalosiphum padi on barley plants which had root contact with S. arvensis, C. album or barley grown alone, and on barley plants exposed to weed volatiles or unexposed barley plants. Whether aphid intrinsic rate of release on treated plants differed significantly from that on control plants was tested using a t-test for independent samples (< 0·05) ns = not significant
Type of interactionBarley with weedsBarley aloneP-value
Sinapis arvensis
Root interaction0·50 ± 0·0090·55 ± 0·010·0001
Volatile interaction0·46 ± 0·020·55 ± 0·030·02
Chenopodium album
Root interaction0·50 ± 0·020·48 ± 0·02ns
Volatile interaction0·51 ± 0·030·53 ± 0·02ns

Discussion

Laboratory studies showed that the performance of aphids was negatively affected on barley plants having root contact with S. arvensis, reducing the growth and the intrinsic rate of aphids and prolonging their development rate significantly. This reduced aphid performance indicates that coexistence with other plants can induce considerable changes in their host plants which can be the main reason for a significant reduction in aphid population development as observed in our field experiments. In contrast, C. album induced responses in barley plants through exposure to volatiles of this weed (Ninkovic, Glinwood & Dahlin 2009). This induction caused lower mean relative growth rates of aphids feeding on exposed barley plants, but their development and fecundity were not affected. This is in line with observations from the field experiments which showed that the presence of C. album in barley plots did not affect aphid population development compared with the monoculture. Our results support the theory that the effects of biodiversity on herbivores depend on specific plant combinations. This confirms by the study of Schädler, Brandl & Haase (2007) where the effects on aphids were dependent upon the identity of the competing species (competition with Plantago lanceolata decreased aphid densities on Poa annua by 60–70% while the competition of Trifolium repens and P. annua caused an aphid decrease of 30%).

Plant semiochemicals can serve as different kind of messengers, evoking behavioural or physiological responses in neighbouring plants (Ninkovic 2003, 2010; Novoplansky 2009). The experimental units (twin-exposure chambers) used in this study were arranged to allow only communication via airborne signals, blocking root interactions or allowing only interaction by roots and not by airborne signals, and thus differentiate between these two mechanisms. Our results showed that focal plants can change as a host depending on responses to different signals from neighbouring plants. While the interaction between barley and S. arvensis seems to occur via root interaction, C. album seems to induce changes in barley plants via airborne volatiles. These airborne and below-ground interactions between plants had considerable effects on receiving plants and consequently changes in aphid growth and reproductive performance. Aphids are vulnerable to slight changes in desired plant qualities, since host plant quality is a key determinant of the fecundity of herbivores (Awmack & Leather 2002). Aphid growth and reproductive performance changed significantly when fed on barley having root contact with S. arvensis compared to barley grown without competition or contact with another plant, indicating that this type of plant interaction may change the quality of barley plants as a food source for aphids. Interspecific competition between plants can have negative effects on the growth and performance of herbivores by affecting crop plant quality (Hambäck & Beckerman 2003; Agrawal 2004), and it has been shown that growth and developmental rates of individual aphids are reliable indicators of future population growth rates (Leather & Dixon 1984). However, it seems that the altered growth rates of aphids feeding on with C. album exposed barley plants were not sufficient to reduce aphid population development in the field, showing that this associated resistance is not only dependent upon the identity of the inducing species, but also upon responses to different signals. Plant–plant interactions are dynamic and this can have different outcomes on particular plant–insect interactions.

The main aim of our study was to investigate whether crop–weed biodiversity has a potential role in aphid management. The experiments were designed with focus on the mechanistic aspect, and the high weed density in our study was chosen in order to test their potential impact. Our data show an average yield decrease of 26% in barley grown with S. arvensis in the 2 years of field trials, whereas no significant reduction was observed with C. album. We do not expect farmers to tolerate high levels of weed infestation. One could speculate that the reduction in yield caused by crop competition with weeds exceeds any gain resulting from reduced aphid infestation. However, taking into account that R. padi can cause yield losses of 50% in barley (e.g. Kieckhefer & Kantack 1986), further work is required to evaluate the weed density with optimal effects on aphids and an economically feasible harvest of the crop. Furthermore, for future studies, it would be good to include additional treatments that would distinguish the amount of yield loss due to the aphids from the yield loss due to the weed–crop competition.

The present results show clearly that diversified crops caused not only a reduction in herbivorous insect populations directly – through masking or repelling olfactory cues, or indirectly – due to increased abundance of natural enemies as alleged in earlier theories. The reduction in aphid population is even expressed through induced changes in host plants by above- and below-ground stimuli from neighbouring weed plants. Our findings contribute to the understanding of variations in herbivore responses to agricultural diversity. Interactions in an ecosystem that results in associational resistance are context dependent, and actions recorded for one ecosystem are not necessarily found in other ecosystems (Barbosa et al. 2009). The dependency of context implies that interactions are influenced not only by species and/or genotype but also by the state of resources and the physiological state and phenology of the responding plant (Barton & Koricheva 2010). Each interaction may modify another interaction in an ecosystem, and therefore, we cannot consider these individual interactions exclusively.

Agricultural biodiversity may provide many long-term benefits over monocultures, including reducing pesticide pollution and insecticide resistance problems. Weeds help increase the biodiversity of agricultural systems (Marshall et al. 2002), providing effective herbivore reduction through associated resistance in addition to their benefiting biodiversity directly. Nevertheless, most ecological research on biodiversity is made outside agricultural systems, and there is therefore a need for basic research to understand the links between biodiversity, ecosystem function and sustainability (FAO 2011). This study showed strong indications that agricultural biodiversity affects aphid response and performance through induced changes in their host plants, but effects are specific to composition of species and their mode of action. Now, the challenge shifts to understanding the physiological changes in plants responding to their neighbours. Further work is required to find more appropriate companion plants than S. arvensis, to evaluate the weed density that can give effects on aphids and still give an economically feasible harvest, and to find out what signals are involved in above- and below-ground interactions between plants. To get a balance between herbicide and insecticide control, agricultural production systems need to focus on the thresholds of weed and insect tolerance, taking the associated resistance of biodiversity (here weeds) into account. Improved understanding of the underlying ecology of interactions between plants will increase the ability to develop multidisciplinary integrated pest management, leading to development of crop management systems at higher levels of integration.

Acknowledgements

We thank Sätuna AB for the use of their fields and their collaboration. The study was financially supported by the Swedish Foundation for Strategic Environmental Research (MISTRA) through the PlantComMistra programme. We are indebted to Professor Ulf Olsson for statistical advice. We gratefully acknowledge Professor Jan Pettersson and Dr Richard Hopkins for their perceptive and constructive comments on this manuscript. Finally, we thank also two anonymous referees for helpful comments on the manuscript.

Ancillary