Characterization of the natural enemy community attacking cotton aphid in the Bt cotton ecosystem in Northern China

Planting Bt cotton in China since 1997 has led to important changes in the natural enemy communities occurring in cotton, however their specific effect on suppressing the cotton aphids (being notorious in conventional cotton ecosystem) has not been fully documented yet. We observed strong evidence for top-down control of the aphid population, e.g. the control efficiency of natural enemies on cotton aphid increased significantly in open field cages compared to exclusion cages, accounted for 60.2, 87.2 and 76.7% in 2011, 2012 and 2013 season, respectively. The cotton aphid populations peaked in early June to late July (early and middle growth stages) in open field cotton survey from 2011 to 2013. The population densities of cotton aphids and natural enemies were highest on middle growth stage while lowest densities were recorded on late stage for aphids and on early plant stage for natural enemies. Aphid parasitoids (Trioxys spp., Aphidius gifuensis), coccinellids and spiders were key natural enemies of cotton aphid. Briefly, natural enemies can suppress aphid population increase from early to middle plant growth stages by providing biocontrol services in Chinese Bt cotton.

by open cotton field survey, and (iii) to identify the most abundant natural enemy species (especially parasitoids) of cotton aphid in Bt cotton fields of Northern China. The results of the present study will help to characterize changes in the natural enemy community in the current Bt cotton agro-ecosystem along the YRR of China.

Results
Field cage experiment. In 2011, cotton aphid densities did not differ significantly among blocks (P = 0.299) and cage types (P = 0.552). But as function of the dates (P < 0.001), and the two factors (cage type and sampling date) interact significantly (P < 0.001) ( Table 1). The lowest mean number of aphids (944 per 100 plants) was recorded in the open field cages, whereas the highest mean number of aphids (2,370 per 100 plants) was recorded in exclusion cages (Fig. 1). In 2012, cotton aphid densities recorded did not differ significantly among blocks (P = 0.719), but aphid densities differed significantly among cage types (P < 0.001), as function of the dates (P < 0.001), and the two factors (cage type and sampling date) also interacted significantly (P < 0.001) ( Table 1). In 2012, the lowest mean number of aphids (6,090 per 100 plants) was recorded in the open field cages, whereas the highest mean number of aphids (47,918 per 100 plants) was recorded in exclusion cages (Fig. 1). In 2013, cotton aphid densities did not differ significantly among blocks (P = 0.062), but aphid densities differed significantly among cage types (P < 0.001), as function of the dates (P < 0.001), and the two factors (cage type and sampling date) also interacted significantly (P < 0.001) ( Table 1). In 2013 season, the lowest mean number of aphids (2,574 per 100 plants) was recorded in the open field cages, whereas the highest mean number of aphids (11,731 per 100 plants) was recorded in exclusion cages (Fig. 1). Briefly, in 2011 and 2013, cotton aphid numbers were lower than 2012.

Open field survey.
In the open cotton field survey, the lowest (1,399 per 100 plants) and the highest (26,346 per 100 plants) mean number of aphids was recorded in 2011 and 2012 season, respectively. The most common groups of natural enemies recorded in the natural enemy guild were coccinellids (2011), aphid parasitoid (2012) and spiders (2013) (  Population dynamics of natural enemies. The most abundant natural enemy group differed according to the seasons and plant growth stages considered. Among predators complex, coccinellids and spiders were the most common species, followed by anthocorids and chrysopids from 2011-2013 (Table 2). Overall, natural enemy distribution at early, middle and late plant growth stages was recorded as 26, 38 and 36% in 2011 while 6, 70 and 24% in 2012 whereas 26, 50 and 24% in 2013, respectively (Table 3) Aphid parasitoids were largely attacking cotton aphid during 2011 when there was high population density of aphids at early seedlings stage. In 2012 and 2013, aphid parasitoids played a combined role with predators for aphid reduction throughout the season. Aphidiines were primarily observed as only 15 Aphelinidae mummies were seen during the whole study. Randomly collected mummies (stored in growth chamber) in 2012 and 2013 yielded parasitoids and hyperparasitoids. In early cotton stage, in 2012, 52% of collected mummies yielded hyperparasitoids (total collected = 69) while only hyperparasitoids were recorded in 2013 for that period. In middle cotton stage, in 2012, 46% of collected mummies yielded hyperparasitoids (total collected = 108) while 6%

Discussion
The differences in the population density of cotton aphid recorded in different exclusion cages showed a strong top-down impact of natural enemies over three growing seasons from 2011-2013. The numbers of aphid remained low from middle June to early July (early to middle plant growth stage of the crop). In open cotton field survey, population dynamics of aphids varied with the seasons and stages of plant growth where the highest population density of cotton aphid was recorded during 2012, and at the middle plant growth stage (July) over three growing season. More specifically, early and middle growth stages of cotton crop are the most critical for aphid infestation as a sudden increase in the density could be observed. Among the natural enemy guild, the most common species recorded were coccinellids, spiders, and aphidiine parasitoids in three sampling seasons where the highest numbers of natural enemy was recorded at middle plant growth stages of the cotton crop (see Table 2). Potential aphid parasitoids were identified as Trioxys spp. and Aphidius gifuensis A. during two sampling season (2012 and 2013). Overall, these results suggest that natural enemy populations should be conserved by avoiding insecticide application at early plant growth stage of cotton crop that would ultimately suppress the increasing numbers of cotton aphid populations at middle and late planting growth stages. Moreover, this series of experiments is very useful information to highlight and provide a direct assessment of the seasonal importance of different natural enemy groups attacking cotton aphid in the Bt cotton ecosystem along YRR of Northern China.   Exclusion cages experiments evaluated a great contribution of natural enemies against cotton aphid density over three growing seasons in transgenic Bt (Cry1Ac) cotton field in Northern China. Similar findings have been reported by Lin et al. 12 and Han et al. 11 in Bt cotton ecosystem of Northern and central China, respectively 12,11 . When there were no natural enemies (i.e. in exclusion cages), cotton aphid populations could increase up to maximum of 317-fold in 2011, 5703-fold in 2012 and 1223-fold in 2013 (from the aphid density at the initial release date). Several studies on population dynamics of cotton aphid and its natural enemies in cotton field without any insecticide use had been conducted in Northern China, and the results showed that natural enemies could not effectively suppress the aphid population 13,14 Fig. 1; Table 1), which might result from the low seasonal mean population density of aphids recorded during 2011, which was 7,500 aphids per 100 plants, 16.6 and 4.3 times less than aphid density recorded in 2012 and 2013, respectively. Safarzoda et al. 15 reported non-significant effect of exclusion cages during the low aphid density season 15 . These major differences in cotton aphid population dynamics indicated strong, but not systematic, top-down influence of natural enemies on cotton aphid in fields.
In our visual observations when predators were present in the open field cages, the parasitoid population density remained low during the whole season. Two possibilities could exist for this trend (i) either because of possible intra-guild predation (IGP) of parasitoid mummies by coccinellids 16,17,18,19,20 , and/or (ii) through resource competition of parasitoids with the generalist predators in cages {for more details, see Fig. 1 (2012 aphid population density)} 21,22 . In this case, the aphid parasitoids may help reducing aphid densities but primarily in mid-season (each season in July) as population dynamics change between years and is affected by many factors. The results of cages showed the impact of natural enemies by fluctuation in cotton aphid population using exclusion cages.  The open field survey confirmed the prevalence of parasitoids, coccinellids and spiders on the cotton aphid in Bt cotton field. Visual presence of C. septempunctata, H. axyridis, P. japonica and A. variegata matches with other studies which highlighted the role of C. septempunctata as major predator responsible for variation in population of cotton aphid in northern China 23,24 . P. japonica is reported as one of the most common predators of cotton aphid because of its life history phenology and features that proved its importance as useful biocontrol agent for the management of the aphid in cotton fields 25,26 . Coccinellids are important natural enemies of several aphid species 27,28 , while Lu et al. 3 reported the presence of the same coccinellid species playing an important role in the suppression of cotton aphids in cotton fields of Northern China 3 . More specifically, coccinellids as general predators can feed on various other sucking insects like thrips, spider mites, whiteflies and many other small prey 29,30,31,32 , these alternate hosts were also present during our study at middle to late growth stages of the crop but were not monitored. Harwood et al. 31 found that these prey can help the predators to establish in the early season when aphid density is low 31 . In short, coccinellids have great effect to reduce or delay the establishment of aphids and thereafter their subsequent population density in the early season of seedling stage 33 . In this way generalist predators are very useful as biocontrol agent for conservation biological control. Among Araneae (spiders), members from Linyphiidae and Thomisidae were visually observed attacking cotton aphid in Bt cotton fields of Northern China. Sheet-web weavers, Erigonidium graminicolum S., and Hunting spiders, Misumenopos tricuspidata F. and Pardosa t-insignita Boes. et Str. (Lycosidae) were reported as the most common species of spiders in Northern China 34 . For centuries, spiders have been used in Chinese field crops as a tool for management of rice pests 35 . Spiders were reported as good generalist predators due to their obligate predatory feeding strategies [36][37][38][39] . Spiders have potential to cause mortality of crop pests such as aphids 40 . A complex of spider species is more effective at controlling prey densities (including aphids) than the presence of a single species of spider 41 . In our study we found a complex of two above mentioned spider's families. Spiders do have the potential to be highly effective biological control agents as stressed in our study, notably during the last season (2013) where spiders were the most abundant natural enemies to suppress aphid population. However further studies including gut content analysis, would confirm it 42 . Parasitoids, Trioxys ( = Binodoxys) spp. and A. gifuensis proved to be major natural enemies (species identified from randomly collected aphid parasitoids) for suppressing cotton aphid populations density in Bt cotton fields of YRR of China. However, Aphidiinae alone could not be factor to limit totally aphid population build up as aphid density reached ~123,000 aphids per 100 plants by July 16 th in 2012 season. There might be a rapid aphid population growth in these restriction cages at early cotton growing season because predators were excluded (as general predators are known to prevent pest population build up early in the season) 43 .
Among the aphid parasitoids, Aphidius spp., are being successfully used in wide range of crops across the world 44 . In the past, various studies were conducted to report the diversity of natural enemy species in cotton in different regions of the China (e.g. in cotton fields along YRR of China near Beijing), and the dominance of Chrysoperla sinica T., P. japonica, various spiders and Orius minutus L. was reported 29 . In our findings, aphid parasitoids (A. gifuensis) were the most abundant species during the 2012 season, while coccinellids were the most abundant in 2011 but Men et al. 45 reported the decrease in diversity of natural enemies during three year studies (1999, 2000 and 2001) in Bt-cotton of Northern China 45 . In another study, A. gifuensis, P. japonica and C. septempuctata were recorded dominant in cotton 46 . There were similar findings in Hebei province of Northern China where the dominance of P. japonica was reported 47 . We found that aphelinid parasitoids were almost absent from the field (as reported in Brassicae crops) 48,49 . Trioxys spp. and A. gifuensis were the most common aphid parasitoids over two growing seasons (2012 and 2013). Several species from the Binodoxys genus ( = Trioxys) are known to efficiently attack cotton aphid [50][51] , and B. indicus or Trioxys indicus may be important natural enemies of this aphid pest in the YRR region and other cotton growing regions that have not yet been extensively surveyed.
Overall, the abundance of natural enemies especially the predators and aphid parasitoids, both in early and middle growth stages of the cotton crop presents a challenge to insect pests management researchers to develop sustainable biological control conservation techniques. If succesful in developping such optimized IPM, then it would help to manage outbreaks in populations of secondary pests in Bt cotton ecosystem along YRR in Northern China. Cotton was planted on May 5 th , 12 th , and 14 th in 2011, 2012, and 2013, respectively. The cotton was harvested in October in all growing seasons and then the field was plowed, fertilized, and irrigated before the sowing of cotton for the next year. Cotton seed was mechanically sown 5 cm deep at 20 kg/30,000 plants per ha at a plant, row and bed spacing of 40, 40 and 100 cm, respectively in all seasons. Cotton seedlings emerged 8-10 days after planting in all growing seasons. All agronomic practices of the cotton were followed according to local recommendations in which the experimental station is located. No pesticides were applied to the field.

Methods
To provide aphids for artificial infestation in exclusion cages experiments, naturally occurring cotton aphids were collected from this field in May for all the seasons. Aphids were cultured on the same cotton variety of 3-10 days old seedlings in plastic pots in a greenhouse (at 25 ± 1 °C, 60-70% RH and a photoperiod of 16: 8 (L:D) hour) at Langfang experimental station.
Scientific RepoRts | 6:24273 | DOI: 10.1038/srep24273 Experimental setup. Field cage experiment. Three levels of natural enemy exclusion cages were used: (i) exclusion cage with 1 × 1-mm mesh openings in which there was no entry of any predator or parasitoids and thus the aphids were fully protected, (ii) restriction cage with 2 × 2-mm mesh openings in which the activities of predators were restricted but allowed aphid parasitoids to enter the cages, (iii) open field cages with four bamboo wood sticks without using any mesh standing upright into the ground. This treatment allowed natural enemies complete access to the aphids. Similar cage type and mesh size were used before in cotton fields 12 .
Three different treatments were established on June27 th in 2011, June 18 th in 2012 and June 21 st in 2013 where all treatments were replicated six times with 60 healthy plants per replicate for each of three blocks following the completely randomized block design (see supplementary data, Fig. S1). Three blocks, each of 13 m × 15 m with 2 m buffer surrounding the blocks. Each cage/plot (1.8 × 2 × 2 m, length × width × height) was set over three rows of two planting beds. Ten healthy plants inside each cage/plot were selected and marked with plastic strips.
Exclusion and restriction cages were of polyester sacks 2 m in width, 1.8 m in length and by 2 m in height and supported on iron poles at each corner. There was 1.0 m and 0.80 m distance between cages (Fig. S1). The bottom edges of the mesh were buried in the soil up to a depth of 10 cm to prevent or exclude the ground-dwelling predators.
One day before artificially infesting the plants with aphids, the selected plants were cleaned for any resident arthropods manually by camel's hair brush. To infest plants, the aphids were placed on the highest central leaflet to the experimental plants by using a small and fine camel's-hair brush. In each replicate, the plants were infested artificially at the rate of 5 aphids per plant (adults) at June 20 th , 11 th and 14 th in 2011, 2012, and 2013 season, respectively. After the aphid infestation on plants, cages were closed by a zipper opening on one side, and aphids inside each treatment were left to reproduce for seven days. Samples of aphid density on each plant (as a whole) were visually examined and counted by weekly survey each year from Mid-June to early September.
Open field survey. The field was divided into three blocks every season and each block was 16 m × 15 m with 2 m buffer surrounding the blocks. Twenty plots were selected as fixed sampling sites in each block following the five plants method used in soybean field 52  Sampling during the open field survey was carried out as follow: Each plant (as a whole) was visually examined and counted for all stages (larvae, nymphs, adults) of the following arthropods; cotton aphid, coccinellids, chrysopids, anthocorids, spiders and aphid parasitoids. All the predators were identified to order. Samples were collected during three cotton growth stages; (i) early plant stage (May-June) at seedling and square formation, (ii) middle plant stage (July) at flowering and boll formation, and (iii) late plant stage (August-September) at boll formation and opening, and before harvesting. Aphid parasitoids were counted on the basis of their field appearance as tan (Aphidiinae) and black (Aphelinidae). Mummy samples were collected randomly in 2012 and 2013 from various open field plots (when parasitoid densities were at high levels) for further identification of parasitoids. The collected mummies (2012: n = 177, 2013: n = 150) were brought back to the laboratory and placed individually in gel caps in a climatic chamber (25 °C, 65% RH and 16:8 h/ L:D) for 10 days. The emerged parasitoids were identified using identification keys [53][54][55][56][57] . Data sheets are stored at IPP-CAAS, Beijing, P. R. China.

Statistical analyses.
Aphid densities in the field cage experiment were non-normally distributed and therefore were log transformed for analyses. Counts were converted into mean number ( ± SEM) per 100 plants in both experiments: 10 plots/cages for exclusion field cage experiments, and 20 plots in open field survey. For exclusion cage experiments, we tested the effects of block, cage type, sampling date, and interaction between sampling date and cages level on aphid density using PROC MIXED repeated measures ANOVA with SAS program, version 9.2 58 . Sample date was repeated within replicates and separate analyses were carried out for each year of the study. A probability level of P < 0.05 was considered as indicating statistical significance separately for each year of the study. For survey data, comparison of three cotton growth stages (early, middle and late stage) for all sampling parameters (cotton aphid and natural enemies) during three sampling years (2011, 2012 and 2013) were carried out using a One-way analysis of variance with the Student-Neuman-Keuls test (SAS program, version 9.2) 58 . A probability level of P < 0.05 was considered as indicating statistical significance separately inside each year of the study. GraphPad Prism version 6.00 was used for drawing all the graphs.