Chemically mediated tritrophic interactions: opposing effects of glucosinolates on a specialist herbivore and its predators


  • Rebecca Chaplin-Kramer,

    Corresponding author
    1. Department of Environmental Science Policy & Management, University of California, Berkeley, 130 Mulford Hall #3114, Berkeley, CA 94720, USA
      Correspondence author. California Institute for Energy & Environment, University of California, 2087 Addison Street – 2nd Floor, Berkeley, CA 94704, USA. E-mail:
    Search for more papers by this author
  • Daniel J. Kliebenstein,

    1. Department of Plant Sciences, University of California, Davis, One Shields Ave., Davis, CA 95616, USA
    Search for more papers by this author
  • Andrea Chiem,

    1. Department of Integrative Biology, University of California, Berkeley, 3060 Valley Life Sciences Bldg #3140, Berkeley, CA 94720, USA
    Search for more papers by this author
  • Elizabeth Morrill,

    1. Department of Environmental Science Policy & Management, University of California, Berkeley, 130 Mulford Hall #3114, Berkeley, CA 94720, USA
    Search for more papers by this author
  • Nicholas J. Mills,

    1. Department of Environmental Science Policy & Management, University of California, Berkeley, 130 Mulford Hall #3114, Berkeley, CA 94720, USA
    Search for more papers by this author
  • Claire Kremen

    1. Department of Environmental Science Policy & Management, University of California, Berkeley, 130 Mulford Hall #3114, Berkeley, CA 94720, USA
    Search for more papers by this author

Correspondence author. California Institute for Energy & Environment, University of California, 2087 Addison Street – 2nd Floor, Berkeley, CA 94704, USA. E-mail:


1. The occurrence of enemy-free space presents a challenge to the top-down control of agricultural pests by natural enemies, making bottom-up factors such as phytochemistry and plant distributions important considerations for successful pest management. Specialist herbivores like the cabbage aphid Brevicoryne brassicae co-opt the defence system of plants in the family Brassicaceae by sequestering glucosinolates to utilize in their own defence. The wild mustard Brassica nigra, an alternate host for cabbage aphids, contains more glucosinolates than cultivated Brassica oleracea, and these co-occur in agricultural landscapes. We examined trade-offs between aphid performance and predator impact on these two host plants to test for chemically mediated enemy-free space.

2. Glucosinolate content of broccoli B. oleracea and mustard B. nigra was measured in plant matter and in cabbage aphids feeding on each food source. Aphid development, aphid fecundity, predation and predator mortality, and field densities of aphids and their natural enemies were also tested for each food source.

3. Cabbage aphids growing on high glucosinolate plants like B. nigra contained more glucosinolates than aphids on lower glucosinolate B. oleracea. Aphids on B. nigra had shorter generation times and greater daily fecundity, while their predators (Diptera: Syrphidae) had lower feeding and higher mortality rates. Lower syrphid densities were found on B. nigra than on B. oleracea in the field.

4.Synthesis and applications. This study presents physiological and field evidence to suggest that weedy B. nigra may provide enemy-free space from an important predator. Habitat near crops should be examined for its potential to provide enemy-free space and compromise otherwise effective biological control. The issue of pest control must be considered from the bottom up as well as the top down.


Plants can provide an ecological refuge for herbivores by allowing them to chemically or physically escape their natural enemies, sometimes called an ‘enemy-free space’ (Jeffries & Lawton 1984). Although the best examples of enemy-free space have been documented in natural systems and through experimental host-plant shifts (Denno, Larsson & Olmstead 1990; Gratton & Welter 1999; Murphy 2004), the concept also has useful application in the context of biological control for agricultural systems. Our understanding of biological control is generally focused on the top-down pressures of predators on their prey, but the occurrence of enemy-free space reveals important subtleties in the relationship between these trophic groups. Successful implementation of top-down control by natural enemies may hinge on incorporating bottom-up factors such as the distribution of plants providing refuge to pests. Identifying which plants could provide enemy-free space for different crop pests could help guide management to reduce or eliminate an important potential source of pests to crop fields.

Many wild plants contain higher levels of defence compounds than their domesticated congeners (Baker 1972; Cole 1997a; Gols & Harvey 2009), and specialized herbivores can often utilize those toxins to compile their own chemical arsenal to escape enemies (Nishida 2002; Hopkins, van Dam & van Loon 2009). If this chemical refuge allows pests to escape their enemies more effectively than when feeding on crops, the occurrence of such plants around farmland could be supporting populations of pests that are then unregulated by their enemy community. The presence of refugia could allow pests to build to greater levels than if such refuges were absent from landscapes. The invasive mustard Brassica nigra (L.) Koch has much higher glucosinolate levels than the domesticated Brassica oleracea (L.) (Rodman, Kruckeberg & Alshehbaz 1981; Mithen, Raybould & Giamoustairs 1995), and weedy patches of B. nigra are commonly found in field margins or edges around B. oleracea fields (cole crops, such as broccoli, cauliflower, kale, cabbage). The co-occurrence of these two Brassica species provides an excellent opportunity to test whether an invasive weed can serve as enemy-free space for crop pests.

Plants in the family Brassicaceae are a good study system for chemically mediated enemy-free space, because of their sophisticated two-part defence system involving a glucosinolate compound and myrosinase protein complex that has been described as a ‘mustard-oil bomb’ (Ratzka et al. 2002). When plant tissue is damaged, glucosinolates come into contact with myrosinase enzymes, which remove glucose from the glucosinolates, leading to the formation of toxic hydrolysis products such as isothiocyanates. The cabbage aphid Brevicoryne brassicae (L.) is among a small group of insects that have found a strategy for exploiting Brassicaceae, a resource toxic to most herbivores. Generalist herbivores feeding on brassicaceous plants, including the lepidopterans Mamestra brassicae (L.), Spodoptera eridania (Cramer) and Trichoplusia ni (Huebner), tend to rely on direct metabolic detoxification of the glucosinolates, receiving little or no predator defence benefit (Li et al. 2000; Lambrix et al. 2001; van Leur et al. 2008). In contrast, specialist herbivores, such as the cabbage aphid, the sawfly Athalia rosae (L.), and the harlequin bug Murgantia histrionica (Hahn), co-opt toxicity from glucosinolates through sequestration, escaping many of their enemies as a result (Francis et al. 2001; Muller et al. 2001; Aliabadi, Renwick & Whitman 2002). What makes cabbage aphids relatively unusual among these specialists is that they both sequester the glucosinolate to avoid its toxic effects and produce their own myrosinase enzyme, arming themselves with their own mustard bomb, such that they not only deter but actually harm their predators (Bridges et al. 2002; Hopkins, van Dam & van Loon 2009). When the aphid body is damaged, the two compounds are mixed and the hydrolysis of glucosinolate by the myrosinase produces volatile toxic isothiocyanates just as it does in plants (Francis et al. 2001; Francis, Lognay & Haubruge 2004).

The degree to which a plant with high glucosinolate content like B. nigra serves as a viable predator refuge depends on the trade-off between enemy-free space and herbivore performance. Cabbage aphids are completely dependent on their food source for the acquisition of the mustard bomb. They cannot synthesize their own glucosinolates (Kazana et al. 2007) and therefore can only acquire defences through glucosinolate consumption (Francis, Haubruge & Gaspar 2000; Vanhaelen, Gaspar & Francis 2002; Olmez-Bayhan, Ulusoy & Bayhan 2007; Pratt et al. 2008). Thus, differences in the concentration of glucosinolates within the aphids’ diet may influence their ability to deter predation. However, any benefit of predator refuge derived from glucosinolates may be offset by slower aphid development or reproduction, because of potential energetic costs of sequestering these toxic compounds (Cole 1997a). While not framing their study specifically in terms of glucosinolate content, Ulusoy & Olmez-Bayhan (2006) found that aphids grown on another wild mustard, Sinapis arvensis (L.), had lower reproductive rates than those reared on domesticated B. oleracea cultivars. Aphids on S. arvensis also had shorter generation times, however, resulting in no net difference in intrinsic rate of increase for aphids on wild mustard vs. cole crops in the absence of predation. However, these herbivore development studies have not been conducted on the same plant species as the studies investigating effects on their predators, and therefore, no real conclusions regarding the potential trade-offs of enemy-free space can be drawn from the current literature.

It is possible that wild Brassica species could serve as sources of pests to nearby crops if they provide enemy-free space without compromising the pest’s own population growth. We simultaneously measured herbivore performance and predation risk on two host-plant species to assess the potential for chemically mediated trade-offs. Specifically, we quantified the glucosinolate content of two Brassica species (B. oleracea and B. nigra), measured the glucosinolate content in aphids reared on each of the two plant species, assessed aphid development and fecundity while growing on these two food sources, and evaluated syrphid development and predation of aphids raised on each host plant. Subsequently, we addressed the overall effects of these trade-offs by estimating the abundance of aphids, syrphids, and aphid parasitoids on the two Brassica species at a series of locations in the field. We tested the hypotheses that (i) high glucosinolate content places a physiological burden on herbivores utilizing these compounds for their own defence, (ii) high glucosinolate content confers an advantage to herbivores in the form of reduced predation and (iii) high glucosinolate content leads to lower densities of aphid natural enemies on B. nigra in the field. Whether aphid densities themselves are lower or higher on B. nigra would then depend on the balance of the bottom-up and top-down forces explored in the first two hypotheses. This research highlights the impact of phytochemicals on pests and their predators, with implications for the role a weed may play in inhibiting effective biological control in agricultural systems.

Materials and methods

Plants and Insects

Broccoli B. oleracea var. italica cv. Gypsy (acquired from Growers Transplanting, Salinas, CA, USA) and black mustard B. nigra (acquired from Reimer Seeds, Mount Holly, NC, USA) were grown in individual pots in potting soil with vermiculite at 18–24 °C and 16-h daylength in a greenhouse. All plants were at least 25–30 cm tall before being used for the aphid colonies. Cabbage aphids B. brassicae were reared on these host plants in dense colonies (>100 individuals per leaf) under the same greenhouse conditions for a period of 3 months prior to the start of the study.

Glucosinolate Profiles

Ten apterous aphids in their penultimate instar were selected from the colonies on six different plants in each treatment (B. oleracea and B. nigra) and preserved in 500 μL of 90% methanol for glucosinolate analysis. Fresh leaf matter totalling 150–250 mg (equivalent to a small portion of one leaf) was also collected from each of the 12 individual plants (six B. oleracea and six B. nigra) for analysis.

Glucosinolates were measured via high-performance liquid chromatography (HPLC) with diode-array detection (DAD) using an Agilent 1100 system (Kliebenstein, Gershenzon & Mitchell-Olds 2001). Glucosinolates in both plants and aphids were identified and quantified in relation to previously purified standards (Reichelt et al. 2002). All individual aliphatic glucosinolate values within a sample were summed and standardized to nmol per unit fresh weight for plant matter or per aphid to provide the total aliphatic glucosinolate content.

Impacts of Brassica nigra on Aphids

Small cages were clipped directly onto leaf surfaces of potted B. oleracea or B. nigra plants (10 of each species) in the greenhouse, at 18–24 °C and 16-h daylength. Each clip cage contained five adult aphids (denoted hereafter as the zero generation, or G0) originating from the greenhouse colonies on their respective host plants. Each aphid colony had been living on the same host-plant species for several (at least 8) generations. Aphids from B. nigra colonies were used only on experimental B. nigra plants and likewise for B. oleracea. The clip cages allowed individual aphids or offspring cohorts to be tracked on living plants rather than on detached leaves; this is important for simulating true field conditions, because plants will often increase the amount of glucosinolates in the leaf when they are attacked (Cui et al. 2002; Kim & Jander 2007).

Daily observations were made of the clip cages. On the first day that nymphs were observed on the leaf, all the G0 adults were removed and the cohort of between three and six nymphs (G1) remained in a cage together until their penultimate instar, at which point each of the G1 aphids was moved to its own cage on a separate leaf on the same plant and tracked individually. This experimental design was necessary to reduce aphid escape from the cages. Solitary nymphs will often search the plant for other aphids, but keeping a cohort of nymphs together in the same cage increased the likelihood that they would remain within the cage (R. Chaplin-Kramer, personal observation). Once the G1 adults were in their own individual cages, they continued to be checked on a daily basis for nymph production. The number of days until the first nymph in the next (G2) generation was produced was recorded for each G1 individual as a measure of their development time, from birth to reproductive maturity. The number of G2 nymphs produced per G1 aphid each day provided a measure of reproductive rate and total fecundity, as the G1 aphids were tracked until their death. The G2 nymphs were removed each day after counting to remove any artefact of crowding from the effect of the cage.

Forty G1 individuals were tracked for each treatment, but some aphids escaped from the clip cages during the experiment, when leaf surface irregularity prevented a perfect seal. Individuals that escaped before reproduction were excluded from the analysis. Individuals that escaped after reproduction were included in the development (time to first reproduction) analysis, but excluded from the analyses on total fecundity. This approach yielded 59 G1 aphids (28 in the B. oleracea treatment and 31 in the B. nigra treatment) for the development analysis and 32 G1 aphids (17 in the B. oleracea treatment and 15 in the B. nigra treatment) for the fecundity analysis.

Impacts of Brassica nigra on Syrphids

Syrphid larvae are the most common predators of cabbage aphids feeding on B. oleracea in the region where these cabbage aphids were obtained (Chaplin-Kramer 2010) and are thought to play an important role in their control for this crop (van Emden 1963; Nieto et al. 2006). Syrphid larvae were reared on aphids from the two different Brassica food sources and observed over the course of their larval development in a growth chamber at 18 °C and a 16-h daylength. Syrphid eggs were collected from the field on B. oleracea plants at the University of California Center for Sustainable Agriculture and Food Systems, in Santa Cruz, CA. Brassica nigra was also searched for syrphid eggs at the same site, but no eggs were found on these plants. The syrphid species found at this site included Allograpta obliqua (Say), Eupeodes americanus (Wiedemann), Eupeodes volucris (Orsten Sacken), Syrphus opinator (Orsten Sacken), Scaeva pyrastri (L.), Sphaerophoria sulphuripes (Thomson), Toxomerus occidentalis (Curran), and Platycheirus stegnus (Say). As species cannot be identified at the egg stage, it was not possible to ensure that an equal number of each species was assigned to each treatment, but the distribution of species was assumed to be random across the two treatments.

Twenty syrphids were tracked in each treatment from hatching until the larva died or pupated, and the emerging adults were saved for later species identification. The eggs were kept in a moist Petri dish and checked several times per day for hatching. Immediately after hatching, each syrphid larva was placed in a Petri dish on either a B. oleracea leaf or a B. nigra leaf, with aphids from colonies reared on the corresponding plants. Petri dishes with cut leaves were used in this case because clip cages could not contain the first-instar syrphid larvae. The senescence of the cut leaves in the Petri dishes may have reduced the glucosinolate content of the food source that the aphids were feeding on immediately prior to the introduction of the syrphid larva. However, this period represented a very small fraction of the aphids’ lifetime of feeding, and any reduction in aphid glucosinolate levels should diminish the difference between treatments, providing a conservative estimate of the impact of prey food source on predation.

Only prereproductive aphids were selected as prey to ensure that no new nymphs were born between feedings. When given a choice, syrphids tend to consume aphids in proportion to their size (younger, smaller syrphids preferentially attack younger, smaller aphids; R. Chaplin-Kramer, personal observation). However, to standardize feedings across the experiment, aphids were selected in their penultimate instar, the most easily identifiable. Each day, the number of aphids remaining was recorded, aphids were replenished to ensure a sufficient amount for each syrphid larva to reach satiation (ranging from 10 to 100 aphids, depending on the age of the syrphid), and the leaf was replaced to maintain freshness. Daily consumption rate and total lifetime consumption were used as measures of predation by syrphids on the two different aphid sources (fed on B. nigra or B. oleracea).

Field Observations

To determine whether physiological effects detected in laboratory trials have ecological consequences, insects were compared on the two different host-plant species in the field. Plant matter was collected from B. nigra and B. oleracea on six organic broccoli farms (spaced >1 km apart) in and around the Salinas Valley, CA. Each farm site contained a patch of B. nigra growing no more than 25 m from the field edge of the adjacent B. oleracea crop. Between 50 and 150 g of leaf material was gathered from 20 B. oleracea plants and 20 B. nigra plants on three dates (19 June, 1 July, and 17 July 2009) at each of the sites. Plants were selected at random every 0·5 m along 10-m transects. Leaf and stem materials were selected from the centre of the plants, as the younger plant tissues are more likely to be inhabited by aphids. The samples were individually bagged and returned to the laboratory for inspection. Each sample was washed over a sieve to count and identify all insects inhabiting the different plants. Aphid and syrphid densities were estimated as numbers per unit weight of leaf. Relative per cent parasitism by Diaeretiella rapae (McIntosh) was estimated as 100 × the ratio of the number of aphid mummies to the combined number of aphids and mummies.


All analyses were performed using the statistical program, r (version 2.9.1, Differences in syrphid mortality (proportion of larvae that did not survive to pupation) between B. oleracea and B. nigra treatments were compared using a chi-squared test. All other pairwise comparisons from the laboratory experiments were analysed using analysis of variance. In these analyses, B. oleracea and B. nigra treatments were the predictor variable for the each of the following response variables: glucosinolate content (nmol per 100 mg of leaf material or nmol per 10 aphids), aphid development (days from birth to first reproduction), aphid reproductive rate (mean nymphs produced per day per adult), aphid fecundity (total number of nymphs produced per adult), aphid longevity (days from birth to death), syrphid predation rate (mean aphids consumed per day per syrphid larva), and total predation (total number of aphids consumed per syrphid larva). Field data on aphid and syrphid densities (per unit weight of leaf material) and percentage parasitism were compared for B. nigra and B. oleracea using generalized linear mixed effects models with site and sampling date as random effects. Models for syrphid densities and parasitism included aphid densities as a covariate, as lower prey densities would be expected to result in lower enemy densities irrespective of treatment effects.


Glucosinolate Profiles

Brassica nigra contained several orders of magnitude higher total aliphatic glucosinolate concentrations (155 ± 46 nmol per 100 mg fresh plant matter, mean ± standard error) than B. oleracea (0·68 ± 0·15 nmol per 100 mg fresh plant matter). Mirroring the endogenous glucosinolate differences between the two food plants, aphids reared on B. nigra sequestered nearly ten times the aliphatic glucosinolates of aphids reared on B. oleracea (F = 11·56, d.f. = 1, 10, = 0·007, Table 1).

Table 1.   Glucosinolate content (in nmol per 10 aphids) of cabbage aphids reared on Brassica nigra (black mustard) vs. Brassica oleracea (broccoli). F- and P-values denote significance of anova tests comparing the two host plants, on 1 and 10 degrees of freedom
GlucosinolateB. nigraB. oleraceaFP
Allyl (sinigrin)16·480·0011·630·007
Total aliphatic17·991·8911·560·007
Total indolic0·820·790·030·842

Impacts of Brassica nigra on Aphids

Aphids reared on B. nigra reached reproductive maturity 14% faster on average than those reared on B. oleracea (Fig. 1a; = 18·84, d.f. = 1, 57, < 0·001). Aphids in the B. nigra treatment also reproduced at a faster rate, producing just over three nymphs per day, compared to two nymphs a day for aphids on B. oleracea (Fig. 1b; = 10·84, d.f. = 1, 28, = 0·003). However, aphids survived significantly longer on B. oleracea than B. nigra (Fig. 1c; = 7·46, d.f. = 1, 28, = 0·01), resulting in no significant differences between the two treatments in overall nymph production per individual (Fig. 1d; = 0·067, d.f. = 1, 28, = 0. 79).

Figure 1.

 Impacts of host plant (Brassica oleracea vs. Brassica nigra) on cabbage aphid physiology (means ± SE). (a) Aphid development time, in days from birth to first reproduction; (b) aphid reproductive rate, in average number of nymphs produced per adult per day; (c) aphid longevity, in number of days from birth until death; (d) total aphid fecundity, in total number of nymphs produced per aphid.

Impacts of Brassica nigra on Syrphids

Syrphid larval mortality in the B. nigra treatment was 95%, more than double that found in the B. oleracea treatment (Fig. 2a; χ2 = 9·64, d.f. = 1, = 0·002). Syrphids in the B. oleracea treatment consumed 65% more aphids per day than syrphids in the B. nigra treatment (Fig. 2b; = 4·25, d.f. = 1, 38, = 0·05). Because of these substantial impacts of B. nigra on syrphid larval development, total lifetime aphid consumption by syrphid larvae in the B. oleracea treatment was double that of the B. nigra treatment (Fig. 2c; = 4·18, d.f. = 1, 38, = 0·05). Of the 12 larvae that reached pupation (1 on B. nigra, 11 on B. oleracea), 17% did not emerge from their pupae, 41% were A. obliqua, 25% were S. sulphuripes and 17% were E. americanus.

Figure 2.

 Impacts of cabbage aphid host plant (Brassica oleracea vs. Brassica nigra) on syrphid larvae (means ± SE). (a) Syrphid mortality rate, in proportion of individuals that pupated (grey) vs. died (black); (b) predation rate, in average number of aphids consumed per syrphid per day; (c) total predation, in total number of aphids consumed per syrphid lifetime.

Field Observations

We found no significant differences between aphid densities in B. nigra and the adjacent B. oleracea crop (Fig. 3a; = 2·03, d.f. = 1, 25, = 0·16). Aphid densities were a significant factor in predicting syrphid densities (= 29·97, d.f. = 1, 24, < 0·001), but not percentage parasitism by D. rapae. Taking aphid densities into account, syrphid densities were orders of magnitude lower on B. nigra than in the crop (Fig. 3b; = 15·38, d.f. = 1, 24, < 0·001), and percentage parasitism was likewise several times lower on B. nigra (Fig. 3c; = 7·95, d.f. = 1, 24, = 0·01). All other insects were too rare to analyse.

Figure 3.

 Field observations of insects on Brassica oleracea and Brassica nigra (means ± SE). (a) Aphid densities per g of leaf; (b) syrphid densities per 100 g of leaf; and (c) percentage parasitism.


Laboratory Experiments

Cabbage aphids feeding on B. nigra contain more glucosinolates than aphids feeding on B. oleracea, supporting previously reported differences in glucosinolate sequestration on two other plant species (Francis et al. 2001). This suggests that the aphid’s ability to sequester glucosinolates is not saturated when reared upon B. oleracea in spite of the log-order difference in glucosinolate between the two plants. Thus, it appears that in this system, the food source can determine the aphid’s endogenous defence capacity. Aphids reared on B. oleracea still have detectable levels of glucosinolates, but these concentrations do not provide as much deterrence against aphid enemies as that generated by feeding on B. nigra. Similar results were found for the coccinellid Adalia bipunctata (L.), which was able to complete its larval development on aphids reared on the low levels of glucosinolates found in Brassica napus (L.), but not on aphids reared on the more concentrated glucosinolates of Sinapis alba (L.) (Francis, Haubruge & Gaspar 2000). Other natural enemies show slower growth rates or reduced reproduction rather than direct mortality with high levels of glucosinolates, but similarly seem much less impacted by low glucosinolate levels (Vanhaelen, Gaspar & Francis 2002; Olmez-Bayhan, Ulusoy & Bayhan 2007). While the number and variety of specific glucosinolate compounds prevent a direct comparison between this study and previous work, the general trend is congruent with the findings of this study: low glucosinolate plants can pass some of these compounds onto herbivores without providing the same protection from predators offered by high glucosinolate plants.

Surprisingly, we found no trade-off between the use of B. nigra chemicals as ammunition against their enemies and the overall performance of cabbage aphids. The sequestration of glucosinolates may not impart a significant energy cost upon the aphid as this system only requires a transporter to move the glucosinolate out of the digestive stream. Furthermore, the cabbage aphid may avoid the myrosinase cells in the brassicaceous plant via a behavioural change in stylet penetration, reducing or eliminating energetic cost of disabling the myrosinase (Kim & Jander 2007). While the results presented here may not apply to all cultivars of B. oleracea, the fact that aphids reared on B. nigra in our study reached reproductive maturity faster and had a greater daily reproductive rate suggests that the aphid population growth rate would be enhanced on B. nigra compared to broccoli, as these life-history traits are known to translate into faster population growth (Stearns 1992). Rather than a physiological cost of living on B. nigra, there appears to be a slight benefit. To understand the mechanisms behind this finding, it may be necessary to consider the specific glucosinolate profiles rather than the overall total content. Harvey et al. (2007) showed that larval development time of the lepidopteran Pieris rapae varied dramatically with plant strains differing in glucosinolate levels, but that total glucosinolate content was not predictive. Instead, larval growth was slowed by increased concentrations of indolic glucosinolate neoglucobrassicin, but showed no relationship to any other specific glucosinolate compound. Several of the indolic compounds measured in this study were higher in B. oleracea than B. nigra, potentially slowing aphid growth on B. oleracea relative to B. nigra without enhancing aphid defences, which may be related to other compounds. Cole (1997a) also found that an indolic glucosinolate suppressed the intrinsic rate of increase in the cabbage aphid, while two aliphatic glucosinolates enhanced it. Broccoli was not one of the cultivars tested by Cole, but our glucosinolate profile showed that it contained markedly less sinigrin, one of the aliphatic glucosinolates he found to be positively correlated with the intrinsic rate of increase in the cabbage aphid.

Our results further show that syrphids are far less effective predators of cabbage aphids on B. nigra than on B. oleracea. They consume fewer aphids per day and substantially fewer aphids on B. nigra in total. Moreover, the dramatic mortality of syrphid larvae observed on aphids from B. nigra suggests that there could be strong selection pressure against syrphid oviposition on B. nigra. The ability of aphids reared upon B. nigra to compromise the development and survival of their main predator reveals the potential for this weedy mustard to provide a refuge from predation.

Field Observations

Previous work has suggested that aphids have lower densities on higher glucosinolate food sources in the field, presumably because of the higher physiological burden of living on that food source (Newton, Bullock & Hodgson 2009). However, as our laboratory experiments showed that aphid development and reproductive rate were slightly greater on B. nigra relative to B. oleracea, and that predation by syrphids was much reduced on B. nigra, it may be expected that aphids would be found in higher densities on B. nigra. Contrary to this expectation, our results showed no difference between the two food sources. In fact, although high variation in aphid densities between sites resulted in a nonsignificant difference between the two plant species, the trend was towards higher not lower aphid densities on B. oleracea. This apparent discrepancy between laboratory-based expectation and field observations of aphid densities on the two plant species may be attributed to nutrient availability, as aphid populations on brassicas are known to be influenced by amino acid concentration as well as allelochemicals (Cole 1997a). For plants grown in the same soil in a greenhouse, aphid performance was greater on B. nigra. In contrast, aphids on cultivated B. oleracea plants grown as a commercial fertilized crop may well have benefitted from a greater amino acid concentration than those on the wild B. nigra plants growing in the field margins.

Our field data confirm the high syrphid mortality found in our laboratory trials; syrphids were found at far lower densities in B. nigra patches than in nearby crop. This is in contrast to Newton, Bullock & Hodgson (2009), who found no difference between syrphid larval abundance on wild B. oleracea phenotypes differing in the presence and absence of the glucosinolate sinigrin. As the sinigrin content of these B. oleracea phenotypes was not estimated, it is possible that the concentrations in these plants were not high enough to impact enemies to the extent that we saw in B. nigra. In general, cultivated B. oleracea has much lower levels of sinigrin than B. nigra (Cole 1997a), and as was previously mentioned, this was certainly the case for the variety of B. oleracea we studied (broccoli, var. italica). Thus, the much reduced densities of natural enemies that we saw on B. nigra compared to adjacent broccoli fields, beyond what could be expected from differences in aphid distributions, suggest that some aspect of the glucosinolate profile of B. nigra is far more toxic to syrphids than that of B. oleracea. Syrphid host-plant preference should be examined explicitly in the field, to test our hypothesis that syrphids avoid ovipositing on B. nigra owing to the perceived selection pressure resulting from this toxicity.

It is possible that other natural enemies, in particular parasitoids, can prey on cabbage aphids on B. nigra in the field. The evolutionary relationship between herbivores and parasitoids may better equip these specialist enemies to manage the higher glucosinolate concentrations than generalist predators (Gols et al. 2007; Gols & Harvey 2009). Olmez-Bayhan, Ulusoy & Bayhan (2007) have shown that parasitism rates of the cabbage aphid are as high in higher glucosinolate plants such as S. arvensis as on cultivated B. oleracea, indicating that specialist parasitoids continue to provide some measure of pest control in high glucosinolate systems. However, our field results indicate that parasitism by D. rapae was reduced on the higher glucosinolate B. nigra. This lower rate of parasitism is perhaps surprising, but the extremely low incidence of parasitism in this system (<5% on B. oleracea, <1% on B. nigra) suggests that parasitism does not play a major role in top-down control in any case. Whether specialist parasitoids or other natural enemies can compensate for lower syrphid predation on B. nigra patches probably depends upon geographic locations and cropping systems.

Future Directions

While our observations suggest that B. nigra can reduce aphid mortality (through its impact on syrphid predators) and enhance aphid production, further work is needed to fully investigate the implications of these observations. To investigate the potential for B. nigra to serve as a source of pests to crop fields, research should address host preference and short-distance dispersal patterns in the cabbage aphid. Despite no obvious physiological disadvantage of living on B. nigra compared to B. oleracea, aphids may prefer one food source to the other if given a choice, and their preference may determine whether B. nigra is a trap crop or an aphid exporter. Mark and recapture studies should be implemented to determine whether aphids are moving from B. nigra to the crop or vice versa. More frequent and longer-term field surveys would also help elucidate the relationship between wild B. nigra and B. oleracea crops, revealing whether aphids build up on one food source earlier in the season and then transfer to the other.

Other areas of exploration for this research include plant-level variables such as cultivar type and floral resources. We tested only one cultivar of B. oleracea (var. italica), and as previously mentioned, glucosinolate concentration as well as aphid performance can differ considerably on different cultivars (Cole 1997a,b). Therefore, the impact of B. nigra growing near brassicaceous crop fields may well depend on the type of cultivar planted, and this could be investigated through additional trials on other cultivars. Additionally, flowering B. nigra such as that found at our field sites could provide nectar and pollen to adult syrphid flies, and such resources have been shown to extend longevity and enhance reproduction in many natural enemies (Wäckers & Van Rijn 2005). These positive aspects of B. nigra may compensate to some extent for the negative aspect of providing a predator refuge to pests like the cabbage aphid, although this would be likely to depend on the availability of other floral resources. An energetic analysis of B. nigra pollen and nectar and field observations on the frequency of visits of syrphids to these plants could determine the relative importance of B. nigra in providing a resource to syrphids as well as a refuge to aphids.


These results provide a new perspective on enemy-free space and underscore the importance of its inclusion in biological control research. The lack of a significant reduction in aphid performance on B. nigra compared to B. oleracea indicates that there would be no long-term, population-level trade-offs of sequestering glucosinolates even with a near tenfold difference in glucosinolate concentration in the adult aphid. In fact, the shorter aphid generation time and reduced vulnerability to syrphid predation found on B. nigra suggests that this weedy mustard should be investigated in field studies for its potential to serve as source of pest colonization to surrounding crops. The potential existence of pest refuges from predators is an important consideration for pest control, and phytochemical ecology can be used to elucidate and potentially eliminate the occurrence of such refuges. Food web interactions in a heterogeneous environment such as the wild-crop interface are complex and often unpredictable, and enemy-free space presents an additional intricacy that must be acknowledged if we hope to better understand how to manage these systems.


We are grateful to Jeff Lozier and Kanthin Manivong for the serendipitous discovery of the mustard bomb operating in this system. Lora Morandin and Shalene Jha and three anonymous reviewers provided helpful comments on the manuscript. Financial support came from the Robert and Peggy van den Bosch Foundation.