Downstairs drivers - root herbivores shape communities of above-ground herbivores and natural enemies via changes in plant nutrients



  1. Terrestrial food webs are woven from complex interactions, often underpinned by plant-mediated interactions between herbivores and higher trophic groups. Below- and above-ground herbivores can influence one another via induced changes to a shared host plant, potentially shaping the wider community. However, empirical evidence linking laboratory observations to natural field populations has so far been elusive.
  2. This study investigated how root-feeding weevils (Otiorhynchus sulcatus) influence different feeding guilds of herbivore (phloem-feeding aphids, Cryptomyzus galeopsidis, and leaf-chewing sawflies, Nematus olfaciens) in both controlled and field conditions.
  3. We hypothesized that root herbivore-induced changes in plant nutrients (C, N, P and amino acids) and defensive compounds (phenolics) would underpin the interactions between root and foliar herbivores, and ultimately populations of natural enemies of the foliar herbivores in the field.
  4. Weevils increased field populations of aphids by ca. 700%, which was followed by an increase in the abundance of aphid natural enemies. Weevils increased the proportion of foliar essential amino acids, and this change was positively correlated with aphid abundance, which increased by 90% on plants with weevils in controlled experiments.
  5. In contrast, sawfly populations were 77% smaller during mid-June and adult emergence delayed by >14 days on plants with weevils. In controlled experiments, weevils impaired sawfly growth by 18%, which correlated with 35% reductions in leaf phosphorus caused by root herbivory, a previously unreported mechanism for above-ground–below-ground herbivore interactions.
  6. This represents a clear demonstration of root herbivores affecting foliar herbivore community composition and natural enemy abundance in the field via two distinct plant-mediated nutritional mechanisms. Aphid populations, in particular, were initially driven by bottom-up effects (i.e. plant-mediated effects of root herbivory), but consequent increases in natural enemies triggered top-down regulation.


Terrestrial food webs are woven from complex direct and indirect interactions between organisms, which often underpin community stability (May 1972; Polis 1998). Often, these involve plant-mediated interactions between herbivores and higher trophic groups (Oghushi, Craig & Price 2007). Within the last 20 years, it has been demonstrated that plant-mediated linkages between above-ground and below-ground systems are central to these processes (Bardgett & Wardle 2010). Interactions between root-feeding and shoot-feeding herbivores, mediated via changes to their shared host plant, have been prominent in the research literature from the outset (Masters, Brown & Gange 1993; Johnson, Bezemer & Jones 2008) and most recently the subject of meta-analysis (Johnson et al. 2012). Laboratory and glasshouse experiments suggest that these pairwise interactions can have cascading effects on other trophic groups, including parasitoids (Masters, Jones & Rogers 2001), hyperparasitoids (Soler et al. 2005) and pollinators (Poveda et al. 2003) and are therefore potentially important in shaping herbivore communities (Soler et al. 2012). However, the ecological significance of these interactions on herbivore communities has been difficult to establish because of the lack of evidence linking laboratory and field observations via induced changes in a range of plant chemical mechanisms (van Dam & Heil 2010; Vandegehuchte, de la Peña & Bonte 2010; Soler et al. 2012).

In terms of the plant-mediated mechanisms underpinning such interactions, two possible general mechanisms have emerged; the nutrient (stress) hypothesis and defence induction hypothesis (Johnson, Bezemer & Jones 2008). In the former hypothesis, Masters, Brown & Gange (1993) suggested that damage by root herbivores impaired the uptake of nutrients by plants, which resulted in stress related accumulation of nitrogen compounds that benefited shoot herbivore nutrition. The second hypothesis arose from subsequent studies that demonstrated how root herbivory caused systemic induction of defensive compounds that negatively affected shoot herbivore performance (van Dam et al. 2003; Bezemer & van Dam 2005) suggesting defensive chemistry underpinned such interactions. Specific examples supporting both nutritional and defensive mechanisms (reviewed by Johnson, Bezemer & Jones 2008) continue to appear in the literature, but few attempts have been made to simultaneously quantify the effects of a range of plant chemicals (nutritional and defensive) on different functional groups of herbivore. In particular, nutrients such as phosphorus are often overlooked, despite emerging evidence that this is a limiting nutrient in many herbivore diets (Elser et al. 2000), which could potentially mediate interactions between herbivores (Huberty & Denno 2006). While sap-feeding aphids are unlikely to be affected by phosphorus, chewing herbivores such as sawfly and weevil larvae may well be negatively affected if it is deficient (Elser et al. 2000).

Review articles consistently state that there is a deficit in empirical field evidence that validates herbivore interactions observed in the laboratory, which should be redressed if we are to assess realistically whether these interactions influence terrestrial community structure (van der Putten et al. 2009; van Dam & Heil 2010; Soler et al. 2012). Moreover, most studies focus on either primary or secondary metabolites as the underlying mechanism for such interactions, whereas herbivores may change multiple aspects of plant chemistry simultaneously. In this study, we set out to address these points by investigating the interactions between a root-feeding herbivore, the black vine weevil (Otiorhynchus sulcatus F.) and two shoot herbivores, the blackcurrant aphid (Cryptomyzus galeopsidis Kaltenbach) and the European blackcurrant sawfly (Nematus olfaciens Benson), on the perennial shrub blackcurrant (Ribes nigrum L.). The shoot herbivores represent phloem-feeding (C. galeopsidis) and leaf-chewing (N.olfaciens) feeding guilds. All three herbivores are commonly found feeding concurrently on R. nigrum (Mitchell et al. 2011) and therefore have the potential to interact.

We used a multisite field experiment to determine whether root-feeding weevils influenced the abundance of aphids and sawflies on plantations of blackcurrant. Using parallel controlled experiments, we examined a range of nutrients (minerals and amino acids) and defensive chemistry (phenolics) in an attempt to identify whether nutritional and/or defensive mechanisms affected different feeding guilds. We analysed phenolics as they are ubiquitous in terrestrial plants, they have a well-characterized role in plant defence, and their concentrations are known to respond to many environmental factors (Harborne 1994), including in blackcurrant (e.g. Johnson et al. 2011). We hypothesized that root herbivores would increase concentrations of amino acids and phenolics in leaves and cause a deficiency in foliar mineral nutrients (N and P) through impaired root function, as predicted by the nutrient stress and defence induction hypotheses. In consequence, we hypothesized that aphids would benefit from increased amino acid concentrations, but would be less affected by elevated phenolics as these are present in very low concentrations in the phloem (Larsson 1989; Johnson et al. 2003). We further hypothesized that larger aphid populations would have a cascading positive effect on populations of their natural enemies. In contrast, we hypothesized that sawflies would be negatively affected by elevated foliar phenolics and decreased foliar minerals.

Materials and methods

Field populations

The field study was conducted in 2009, having established four South facing field sites (50 × 50 m) 2 years prior to the start of the experiment (Fig. 1a). The sites were situated 0·5–1·2 km apart across a mixed arable 215 ha area in North East Scotland (56°N 27′, 3°W 4′). Each field site (Fig. 1b) comprised 250 blackcurrant bushes (4 years old) representing five different cultivars (50 bushes of each). Bushes were planted as shown in Fig. 1(b), with 1 m spacing between plants and 3 m between rows. The Ben Gairn cultivar, which is susceptible to aphids, was selected for this study (Mitchell et al. 2011). Six bushes at each site were selected at random for monitoring, three of which were inoculated with 50 weevil eggs on 28 April 2009. Eight canes on each bush were labelled (four upper canes and four lower canes; Fig. 1c) and examined every c. 14 days (13 May - 12 August 2009) to score non-destructively the occurrence of aphids and sawflies. The presence of any other invertebrates, especially potential natural enemies, was recorded. Soil cores were taken at the end of the experiment to verify that vine weevil larvae were present at higher densities on plants inoculated with eggs.

Figure 1.

Schematic of experimental design for monitoring field populations showing (a) relative position of the four sites over the mixed arable site, (b) configuration of bushes at each site with differently treated bushes (N = 3) and (c) canes sampled in upper (N = 4) and lower (N = 4) part of the plant.

Controlled experiment

Experiments were conducted inside tunnels approximating to 16:8 hours light/dark conditions at 20°C ± 5°C. Seventy-two blackcurrant (cv. Ben Gairn) plants were grown from stem cuttings (see Johnson et al. 2011 for details) and transferred to square plant pots (90 × 90 × 100 mm; W × L × H) containing 134 g of insecticide-free compost (peat-sand-perlite mix containing 17N:10P:15K; William Sinclair Horticulture Ltd, Lincoln, UK). To prevent insects moving between plants, they were secured individually on plinths (170 × 220 × 50 mm; W × L × H) and then placed in 600 × 600 mm water-filled trays (four per tray) so that plants sat above the waterline, effectively creating a ‘moat’ around each plant. After 3 weeks of growth, 36 plants were selected at random and inoculated with 25 vine weevil eggs from an established culture maintained at the James Hutton Institute (Johnson et al. 2011). Aphids and sawflies were obtained from cultures that were established from field collected specimens in June 2008 and maintained at 19°C (16:8 L:D). After 5 weeks, 12 of these 36 plants were selected at random and inoculated with eight teneral aphid adults and another 12 plants were inoculated with four-first instar larvae of sawflies that had been weighed prior to transfer. Of the remaining 36 plants without weevils, 24 plants were inoculated with either aphids or sawflies (selected at random) and 12 remained insect-free. The size of weevil egg inoculations was made on the basis of estimates of local field populations (Clark et al. 2012) relative to the size of field and pot grown plants used in this study. In summary, the experiment comprised 12 randomized replicates of each treatment: (i) insect-free controls, (ii) weevils, (iii) aphids, (iv) sawflies, (v) weevils and aphids and (vi) weevils and sawflies.

After 12 days, the number of aphids and sawflies per plant was counted in situ and insects were then removed from the plants. Roots were separated from the soil to recover weevil larvae. All sawflies and weevils, and a randomly selected sample of five aphid adults, were weighed on a microbalance. The average relative growth rate (RGR) of sawflies was calculated (final mass – initial mass/days on plant). Plant material was weighed and stored at −18°C prior to freeze-drying. The mass of dried samples was recorded before ball milling to a fine powder for chemical analysis.

Analysis of mineral nutrients

The carbon (C) and nitrogen (N) concentration of leaves and roots was determined using an Exeter Analytical CE440 Elemental Analyzer (EAI, Coventry, UK; see Johnson et al. (2011) for full details). For measurement of phosphorus (P), powdered samples (0·05 g) were digested for 20 min at 180°C in 3 mL of 15·8 M HNO3 (Aristar grade, VWR International, Poole, UK) and then oxidized for 20 min at 180°C with 1 mL of hydrogen peroxide in closed vessels using a MARS-Xpress microwave oven (CEM, Buckingham, UK). Digested samples were diluted to a final volume of 50 mL with de-ionized ultra-pure water. Total mineral concentrations were quantified by inductively coupled plasma-mass spectrometry (Elan DRC-e, Perkin-Elmer, Beaconsfield, Bucks, UK).

Analysis of foliar amino acids

Soluble amino acids were extracted in 80% methanol (HPLC-grade) from 0·05 g of powdered leaf material from controls and treatments with weevils or aphids and quantified as described by Johnson Hawes & Karley (2009). All protein amino acids except proline and cysteine could be detected with this method, with a detection limit of c. 0·5 pmol, including the nine essential amino acids that cannot be synthesized de novo by animals: histidine, isoleucine, leucine, lysine, methionine, phenylalanine, threonine, tryptophan and valine (Morris 1990). Although soluble foliar amino acids can differ compositionally from phloem amino acids, previous work in blackcurrant has demonstrated that foliar amino acid composition can indicate phloem composition of a number of amino acids (Graham et al. 2011).

Analysis of phenolic compounds

Analysis was carried out using the enzymatic method described by Johnson et al. (2011). Phenolics were extracted in a 10 : 1 ratio from 0·05 g freeze-dried leaf powder by incubating in 0·5 mL of 50% methanol at 80°C for 2·5 h. The aqueous phase was removed and cleared by centrifugation. A 1 mL enzymatic reaction was set up using 50 μL of the resultant supernatant mixed with 740 μL of 100 mM potassium phosphate buffer (pH 8·0), 100 μL of 30 mM 4-aminophenazone, 100 μL of 20 mM hydrogen peroxide and 1U horseradish peroxidase dissolved in 10 μL potassium phosphate buffer. The reaction was incubated at room temperature for 15 min and absorbance read at a wavelength of 500 nm. Absorbance data were converted to catechin equivalents using a standard curve produced by serial dilution (0–0·10 mg mL−1 catechin). All chemicals were obtained from Sigma-Aldrich (Dorset, UK).

Statistical analysis

Temporal variation in field populations of aphids, sawflies and natural enemies was analysed for each field survey time point using a generalized linear mixed model with Poisson error structure and log link function with an estimated dispersion parameter. Generalized linear mixed models were deemed the most appropriate for population counts that included extensive zero counts and dispersion parameter estimates fell within acceptable bounds. Plant number was nested within site number and both included as random terms in the model, with weevil presence and cane type (upper or lower) incorporated as fixed effects. The abundance of co-occurring species was included in the model to determine whether their distributions affected one another. Degrees of freedom were calculated for each survey using the method described by Schall (1991). Abundance and performance of insects in the controlled experiment was analysed using one-way analysis of variance (anova) with log+1 transformations applied. Survival of sawflies was analysed with a generalized linear model with binomial error structure and logit-link function. Plant biometric and chemical responses to all treatments (weevils, aphids and sawflies) were analysed with a one-way anova, using contrasts between different treatments (shown in Table 3). Amino acids were analysed in foliage from insect-free plants and those treated with aphids, weevils, or both and were therefore analysed with a two-way analysis of variance. Relationships between plant chemistry and insect performance (log+1 transformed) were assessed with generalized linear models (normal distribution and identity link function) that involved the stepwise removal of the least significant plant trait until all terms were statistically significant (< 0·05). Unless stated above or in table legends, analysis was conducted on untransformed data. Where appropriate, transformations were chosen to give residual diagnostic plots which best fitted a normal distribution and showed least heteroscedasticity. All analysis was conducted in Genstat (version 12, VSN International, UK).


Field populations

Examination of soil cores indicated that all of the bushes treated with weevil eggs had larvae on the roots (most likely a second generation), whereas no larvae were recovered from bushes that were not inoculated with weevil eggs. In field observations, aphid abundance was significantly increased by the presence of weevils in five of eight surveys (Fig. 2a; Table 1). The complex of natural enemies associated with blackcurrant comprised Adalia bipunctata (two-spot ladybird), Coccinella septempunctata (seven-spot ladybird), Forficula auricularia (common earwigs), Chrysoperla spp. (lacewings) and Aphidius spp. (brachonid wasps). All were identified as potential predators or parasitoids of blackcurrant aphids (Mitchell et al. 2011) and considered collectively in the statistical analysis because numbers were too low to reliably analyse each taxon separately. Natural enemy abundance increased on plants treated with weevil eggs, causing a subsequent decline in numbers of aphids, to the extent that aphid densities were similar on plants with and without weevils during the fifth and sixth survey (Fig. 2b). Natural enemies of aphids became significantly more abundant on plants treated with weevils than on untreated plants (Fig. 2b; Table 1) during four surveys, but the difference became smaller (Fig. 2b) as aphid density decreased over time (Fig. 2a). There was a strong correlation between the abundance of aphids and natural enemies in the majority of surveys (Table 1). Aphid abundance increased on lower canes in the third survey, but then switched to upper canes for the fourth and fifth surveys conducted in July (Fig. 2a). However, there were significant statistical interactions between cane type and weevils (Table 1), suggesting that weevil presence caused aphids to move higher in the bush.

Table 1. Summary of generalized linear mixed model analysis of aphid and natural enemy abundance as affected by the presence of vine weevils, position on bush (cane type; upper or lower) and co-occurring invertebrates (corresponding to Fig. 1)
SurveyAphid abundanceNatural enemy abundance
Fixed effectsCo-variatesFixed effectsCo-variates
Weevils (W)Cane Type (CT)W x CTNatural enemies (NE)Sawflies (S)Weevils (W)Cane Type (CT)W x CTSawflies (S)Aphids (S)
  1. Statistically significant effects (< 0·05) indicated in bold. Degrees of freedom were calculated for each survey using the method described by Schall (1991). Where populations of natural enemies were low (at the beginning of the season) or aphids were low (end of season) natural enemy abundance was not included as a covariate in the statistical analysis to avoid statistical errors.

1 7·86 0·006 1·210·2730·720·396 6·21 0·014
2 21·95 <0·001 0·260·6130·230·630 65·64 <0·001 0·750·387 19·83 <0·001 0·030·8710·190·6630·450·50353·96 <0·001
3 5·24 0·04 0·090·762 18·38 <0·001 63·15 <0·001 2·460·118 6·66 0·019 0·610·4341·910·1692·340·128 73·4 <0·001
4 13·18 0·006 16·74 <0·001 1·220·271 53·45 <0·001 0·010·999 12·11 <0·001 24·59 <0·001 0·660·4170·540·464 66·61 <0·001
50·190·666 9·63 0·002 21·25 <0·001 131·94 <0·001 2·770·0980·160·70016·59<0·0011·260·2650·480·489 86·46 <0·001
7 8·02 0·009 2·040·1600·220·640 37·09 <0·001 0·340·559 8·12 0·005 0·010·9063·130·0790·190·66232·89 <0·001
Figure 2.

Abundance of (a) aphids and (b) their natural enemies in field plantations of blackcurrant with (●) and without (○) vine weevil larvae feeding on the roots. Mean abundance per cane ± SE shown, N = 96. Statistical significance indicated *< 0·05, **< 0·01, ***< 0·001 for effects of weevil larvae (W), cane type (CT; upper and lower), sawflies (S), natural enemies (NE) and aphids (A) – see Table 1.

There were two generations of sawflies recorded in 2009. During the population peak, sawfly abundance was significantly lower on plants infested with weevils (Fig. 3; Table 2). Sawfly adults were observed earlier (11 June) on plants without weevils, compared to those treated with weevil eggs (30 June). No natural enemies of sawfly larvae were observed on bushes, and the extensive defoliation suggested that their numbers were not significantly affected by higher trophic groups, which is consistent with past reports of blackcurrant sawflies rapidly colonizing and pupating before their natural enemies can become established (Mitchell et al. 2011). The highly correlated co-occurrence of natural enemies and aphids (Table 1) further supports the suggestion that these were enemies of aphids rather than sawflies (Table 2).

Table 2. Summary of generalized linear mixed model analysis of sawfly abundance as affected by the presence of vine weevils, position on bush (cane type; upper or lower) and co-occurring invertebrates (corresponding to Fig. 2)
SurveySawfly abundance
Fixed effectsCo-variates
Weevils (W)Cane Type (CT)W x CTNatural enemies (NE)Aphids (S)
  1. Statistically significant effects (< 0·05) indicated in bold. Where populations were low at the end of the season no formal statistical analysis was conducted to avoid type I and II statistical errors.

10·060·8130·300·5820·500·482 4·01 0·047
3 4·54 0·034 0·010·9290·070·7851·990·1601·840·177
Figure 3.

Abundance of (a) the sawflies in field plantations of blackcurrant with (●) and without (○) vine weevil larvae feeding on the roots. Mean abundance per cane ± SE shown, N = 96. Statistical significance indicated * < 0·05, for effects of weevil larvae (W) and aphids (A). The black and grey arrows indicate first observations of adults on plants with and without weevils, respectively.

Apart from the first survey (Tables 1 and 2), there was no relationship between the occurrence of aphids and sawflies. Further analysis of the first survey with correlation tests suggested there was no strong trend for co-occurrence between the two species however (= 0·068).

Herbivore interactions

In controlled conditions, the presence of weevils caused a significant increase in aphid abundance (Fig. 4a), although individual aphids remained similar in size (F1,21 = 0·01, = 0·918) when reared on plants with (0·32 ± 0·06 mg per insect) and without (0·31 ± 0·06 mg per insect) weevils (mean ± standard error shown, as below). Conversely, the presence of weevils significantly impaired growth rates of sawflies by 18% (Fig. 4b). Sawfly survival was equivalent on plants with (85%) and without (87%) weevil larvae (F1,22 = 0·09, = 0·77). Weevils were not affected by the presence of aphids compared to when they fed on plants alone, either in terms of mean numbers per plant (12·9 ± 1·8 and 11·3 ± 1·8, respectively; F1,22 = 0·04, = 0·839) or mean body mass (9·10 ± 1·22 and 10·98 ± 0·84 mg, respectively; F1,22 = 1·80, = 0·193). Similarly, sawflies had no effect on weevil numbers per plant (9·9 ± 1·4; F1,22 = 0·25, = 0·625) or mean body mass (9·84 ± 0·61 mg; F1,22 = 0·97, = 0·336).

Figure 4.

Statistically significant (*< 0·05) effects of vine weevil larvae feeding on blackcurrant on (a) aphid abundance (F1,22 = 4·63, = 0·043) and (b) sawfly relative growth rate (RGR) (F1,22 = 4·64, = 0·042). Plants with and without weevils indicated W and NO, respectively. Mean values ± SE shown, N = 12.

Effects of herbivores on plant growth and chemistry

The herbivores had contrasting effects on plant biometrics and minerals. In particular, weevils caused significant reductions in plant mass (Table 3), which seemed to be due to loss of roots (Fig. 5a) rather than leaf mass (Table 3). Concentrations of C in both leaves and roots were unaffected by herbivores (F1,66 = 0·16, = 0·693 and F1,66 = 0·39, = 0·533), whereas sawflies caused a small reduction in root N concentrations (Fig. 5b, Table 3). Weevils significantly reduced leaf P concentration (Fig. 5c, Table 3) and concentrations of phenolic compounds in the roots rose sharply when the plants were simultaneously challenged by aphids and weevils (Fig. 4d, Table 3). Aphids caused significant reductions in concentrations of both essential and non-essential amino acids (Table 4), although did not alter the relative composition of the two types of amino acids. In contrast, weevils caused a small, but significant, increase in the proportion of essential amino acids in the foliage (Table 4).

Table 3. Summary of comparative anova tests (corresponding to Fig. 4) for the effects of herbivory on plant mass and chemistry. A more conservative statistical confidence interval of < 0·01 was adopted (indicated in bold) to avoid type I errors arising through multiple testing
ParameterFig.TreatmentControl (NO) versus weevils (W)Control (NO) versus aphids (A)Control (NO) versus sawflies (S)Control (NO) versus weevils and aphids (W, W + A)Control (NO) versus weevils and sawflies (W, W + S)
F 5,66 P F 1,66 P F 1,66 P F 1,66 P F 1,66 P F 1,66 P
  1. *Arcsine square root transformation applied prior to analysis. †One and ‡two samples contaminated and removed from analysis, denominator d.f. = 65 and 64, respectively.

Leaf mass4a0·570·7210·680·4132·410·1250·520·4710·550·4601·270·265
Root mass4a 6·53 <0·001 4·860·0310·250·6161·260·266 9·54 0·003 10·03 0·002
Total dry mass4a2·610·0331·010·3180·180·6711·130·2922·580·1132·170·146
Leaf N*4b1·310·2710·930·3382·240·1390·720·4013·610·0622·470·121
Root N4b 4·07 0·003 2·660·1081·510·2235·130·0270·260·6130·030·857
Leaf P*4c2·660·030 7·17 0·009 0·580·4510·010·931 8·26 0·005 4·880·031
Root P*4c1·250·2951·350·2502·560·1151·500·2250·220·6382·010·161
Leaf phenolics*4d2·200·0650·010·9032·110·1510·650·4230·990·3240·050·824
Root phenolics4d 11·16 <0·001 0·820·3680·380·5410·020·900 7·63 0·007 0·260·609
Table 4. Concentrations of amino acids in foliage of plants subjected to no herbivory, herbivory by either vine weevil larvae (W), or aphids (A), or a combination of both
TreatmentTotal amino acidsa (μg g−1 dry mass)Essential amino acidsa (μmol g−1 dry mass)Non-essential amino acidsa (μmol g−1 dry mass)Proportion of essential amino acids (%)
  1. a

    Data log transformed prior to analysis.

  2. Mean values ± SE shown, N = 12. Statistically significant factors (< 0·05) indicated in bold.

No herbivory7·87 ± 0·681·60 ± 0·226·27 ± 0·5120 ± 1
Weevils6·81 ± 1·261·68 ± 0·415·13 ± 0·8623 ± 1
Aphids1·01 ± 0·131·01 ± 0·133·43 ± 0·3822 ± 1
Weevil and aphids1·02 ± 0·101·02 ± 0·103·37 ± 0·3324 ± 1
  A: F 1,44  = 18·70, <   0·001 A: F 1,44  = 8·27, <   0·001 A: F 1,44  = 22·05, <   0·001 A: F1,44 = 2·32, = 0·135
W: F1,44 = 1·06, = 0·309W: F1,44 = 0·01, = 0·908W: F1,44 = 1·17, = 0·190 W: F 1,44  = 4·23, =   0·046
W × A: F1,44 = 1·26, = 0·267W × A: F1,44 = 0·19, = 0·666W × A: F1,44 = 1·75, = 0·192W × A: F1,44 = 0·95, = 0·336
Figure 5.

Effects of herbivore exposure on shoot and root (a) dry mass, and concentrations of (b) nitrogen, (c) phosphorus and (d) phenolics. Grey colouration indicates vine weevil (W), diagonal lines indicate sawfly presence (S) and dotted bars represent aphid herbivory. Statistical significant results indicated **< 0·01, with results at < 0·05 noted (*). Mean values ± SE shown, N = 12. See Table 3 for summary of statistical analysis.

Plant chemistry and herbivore performance

In the controlled experiment, aphid abundance was positively related to the percentage of essential amino acids in the soluble fraction of leaves (Fig. 6a; F1,22 = 14·92, < 0·001). Aphid body size was not related to any plant traits. Sawfly RGR was significantly and positively related to concentrations of foliar P (Fig. 6b; F1,22 = 9·93, = 0·005), whereas survival was unrelated to any measured plant trait. Leaf consumption by sawflies was not related to any aspect of plant chemistry, nor affected by the presence of weevils (F1,20 = 0·90, = 0·355). Weevil numbers and body mass were not affected by either shoot herbivore, but were positively correlated with root C concentrations (F1,34 = 5·80, = 0·022 and F1,33 = 8·09, = 0·008, respectively, results not shown).

Figure 6.

Statistically significant positive correlations between (a) aphid abundance and % of essential amino acids in the foliage and (b) RGR of sawflies and leaf phosphorus concentrations in plants with (●) and without (○) vine weevils. N = 12. Regression line included.


This study reports the effects of root herbivory on above-ground herbivores and their natural enemies in field conditions and empirically links these to plant tissue chemistry in controlled experiments, a linkage which has so far proved elusive (see Vandegehuchte, de la Peña & Bonte 2010). While they did not specifically elucidate the chemical mechanisms or consider higher trophic groups, several other studies have found good field-based evidence for linkages between above-ground and below-ground herbivores (Kaplan, Sardanelli & Denno 2009; Soler et al. 2009). Moreover, González Megias & Mueller (2010) have also shown how detritvores, in particular, can affect herbivores and natural enemies living above-ground in the field through induced changes in glucosinolates. Together with the novel findings reported here, these studies support the contention that above-ground–below-ground interactions are important for influencing community processes in terrestrial ecosystems.

Another novel aspect of this study is the identification of tissue P concentration as a factor mediating interactions between herbivores. Plant N has been the overwhelming focus of previous research (Mattson 1980), but here we report how a root herbivore induced a P deficiency in the foliage, which demonstrably reduced sawfly development rates. Phosphorus is required for RNA synthesis, and therefore, protein synthesis, so deficiencies are likely to retard consumer growth rates (Elser et al. 1996) rather than cause lethal effects. This is consistent with our observations of reduced sawfly RGR and delayed emergence of adults in the field. Sawflies showed no evidence of compensatory feeding (Simpson & Simpson 1990) on low P hosts, which may have been energetically unfavourable given that blackcurrant leaves are relatively tough (Mitchell et al. 2011). The reduced occurrence of sawflies on shrubs with weevils may have arisen because larvae relocated from low P hosts, a form of ‘dietary self-selection’ (Schoonhoven, van Loon & Dicke 2005), dropped from plants because of malnutrition or, more likely maternal insects avoided ovipositing on low P plants (e.g. Skinner & Cohen 1994) which might be less nutritious for offspring. It seems likely that root herbivory either impaired P uptake in the plant or induced a host plant sink (sensu Kaplan et al. 2011), evidenced by the increased root P concentrations which might indicate remobilization of P from foliage to below-ground tissues, or both. We did not examine the arbuscular mycorrhizal fungi (the principal site of P uptake) associated with host plant roots, but it is known that vine weevil reduces mycorrhizal colonization in some species (A.E. Bennett, personal communication), and this too may have reduced P uptake. In any case, this study provides empirical evidence that P can mediate interactions between below-ground and above-ground herbivores.

In terms of the effects on phloem-feeding herbivores, our results suggest that it is not necessarily the quantity of amino acids that are important, but rather compositional increases in essential amino acids that arise from root herbivory that underpin the positive effects on aphids. Essential amino acids cannot be synthesized by insects de novo, and instead aphids rely on obligate endosymbiotic bacteria, Buchnera spp., for their supply. As a result, aphids frequently have lower assimilation rates for essential amino acids compared with non-essential amino acids (Douglas 2003). Because the relative proportion of essentials compared with non-essential amino acids is generally low in the phloem, it could be envisaged that even a small increase in this ratio might have beneficial effects on insect performance (Douglas 2003). Given that amino acid composition in the phloem can influence the composition of the endosymbiont community in aphids (Wilkinson, Koga & Fukatsu 2007), this raises the intriguing prospect that root herbivory could indirectly affect microbial symbiosis in aphids. Certainly, in a recent meta-analysis of above-ground–below-ground herbivore interactions, Johnson et al. (2012) demonstrated that sap-feeders are the most responsive feeding guild, particularly when the root herbivores are chewing beetle larvae. The fact that aphids moved higher in the bush on plants undergoing root herbivory on two occasions may have arisen because of compensatory plant growth, particularly at the tips and buds, which would be highly nutritious in terms of amino acids.

Despite higher levels of root phenolic compounds, indicative of defensive response (Schmelz et al. 1999), and lower concentrations of root N, the lack of effects of shoot herbivores on root herbivores, disagrees with the original conceptual model of Masters, Brown & Gange (1993). Their conceptual model hypothesized that resources are diverted away from roots following shoot attack, with negative consequences for root herbivores. One explanation why weevils were not adversely affected by lower concentration of N and higher concentrations of phenolics is that it has become adapted to feeding on nutritionally inferior woody root tissue and may have been able to offset the negative effects of phenolic compounds (as reported by Clark, Hartley & Johnson 2011; Johnson et al. 2011). Indeed, some above-ground herbivores can exploit phenolics for cuticle sclerotization to conserve amino N when feeding on woody plants, which are generally poorer in terms of N concentration (Bernays & Woodhead 1982; Bernays, Chamberlain & Woodhead 1983). The positive relationship between root C concentration and root herbivore performance (as reported here) does accord with the model of Masters, Brown & Gange (1993), but as this was not affected by shoot herbivores this study joins a growing number that suggest root herbivores are generally the stronger competitors in above-ground–below-ground herbivore interactions (Blossey & Hunt-Joshi 2003).

The role of induced defences in above-ground–below-ground interactions is supported by many studies (see Bezemer & van Dam 2005; van Dam & Heil 2010 and references therein). In the present study, we did not find a direct link between concentrations of secondary metabolites (phenolic compounds) and herbivore performance, but there was some evidence for induced responses. In particular, phenolic levels rose in the leaves following attack by aphids, but the biggest systemic increases occurred when the plant was challenged by aphids and root herbivores. It could be envisaged that aphids trigger a modest defensive response in the plant but previous attack below-ground caused the plant to mount a full defensive response following aphid colonization. As root herbivores were first on the plant, it is possible that they caused a priming effect (Kaplan et al. 2008) causing plants to mount a more rapid and intensive response once they were attacked by aphids.

Plant-mediated changes in nutrients appear to be mediating the effects of root herbivores on aphids and sawflies in this system, lending general support for the nutrient stress hypothesis (Masters, Brown & Gange 1993). However, there was clear evidence for induced plant defences when plants were challenged by both weevils and aphids, which could have negatively affected a third herbivore sharing the plant. To date, investigating pairwise interactions between a single root- and shoot herbivore has been the norm (Johnson et al. 2012), but this probably reflects experimental constraints rather than natural situations. Investigating three-way interactions in the future might therefore reveal more complex competitive interactions than those predicted from simple pairwise herbivore experiments.

The initial positive effects of root herbivory on aphid abundance in the field, followed by the increase in natural enemies which then diminished this effect, suggests that bottom-up (i.e. plant-mediated effects of root herbivory) initially shaped aphid populations, but consequent increases in natural enemies triggered top-down regulation. We did not set out to determine whether natural enemies were recruited to plants via root herbivore-induced production of volatile compounds, but the higher incidence of predators (in addition to parasitoids), which do not typically use such cues, suggests that increases in herbivore density largely caused natural enemy recruitment (Soler et al. 2012). The strong positive correlation between aphid and natural enemy density also points to this as the underlying reason.

Had we measured aphid populations just in the middle of the season, we would have incorrectly concluded that weevils had no effect on aphid populations. This illustrates how the outcomes of herbivore above-ground–below-ground interactions are not static or fixed, but change over time, and crucially trigger other drivers in the system, in this case top-down effects. The interplay between top-down and bottom-up regulating forces in herbivore dynamics has long been debated but has focussed almost exclusively on above-ground herbivore interactions (Walker & Jones 2001); here, we present clear evidence that these patterns can be influenced by herbivores living beneath the soil surface too.


The authors are grateful to Sheena Lamond, Gill Banks, Steven Gellatly and Alison Vaughan for technical assistance with this research. This study was funded by the Scottish Government's Rural and Environment Research and Analysis Directorate Workpackage 1.3.