Effects of plant genetic diversity on multiple trophic groups
We found that plant genetic diversity (i.e. diversification of cropping or plant cultivation systems; see Methods and Supplementary Table 15) decreased the overall performance of plant antagonists (effect size = −0.539, t = −2.070, P = 0.039) and several of its components (i.e., herbivores (effect size = −0.606, t = −4.127, P < 0.001), weeds (effect size = −0.071, t = −0.167, P = 0.867), plant-feeding nematodes (effect size = −2.118, t = −1.313, P = 0.189) and plant diseases (effect size = −1.087, t = −5.826, P < 0.001)), while increasing the performance of plants (effect size= 0.344, t = 9.098, P < 0.001) and natural enemies of herbivores (effect size= 0.778, t = 4.220, P < 0.001; Fig. 1b). A similar pattern was found when trophic groups were divided into subgroups (e.g., plant performance into plant growth, plant quality and plant reproduction, weed performance into weed growth and weed diversity, and plant disease performance into disease spread and disease damage; see Supplementary Table 3). In the case of small sample sizes (Supplementary Table 3), the results for subgroups were inconclusive (e.g., for herbivore diversity (N = 9, effect size= 0.292, t = 0.991, P = 0.322) and parasitism (N = 16, effect size= 0.089, t = 0.169, P = 0.866)).
In a second step, we tested whether these effects differed among ecosystems, plant life forms, experiment types or climatic zones. In agroecosystems or grasslands, the overall pattern described above was consistently found (Fig. 2a, b; Supplementary Table 4). For the other ecosystems, the responses were variable (Fig. 2c–g; Supplementary Table 4). The overall pattern was also consistently found for both types of experimental studies (i.e., plot and pot experiments) (Fig. 3; Supplementary Table 5), and for both plant life-forms (i.e., herbaceous and woody plants) (Fig. 4; Supplementary Table 6). Across climatic zones, we found a stronger response in temperate than in tropical zones (Fig. 5; Supplementary Table 7), likely because of a smaller sample size in tropical systems (e.g., natural enemies: N = 19, effect size= 0.206, t = 0.388, P = 0.698; weeds: N = 5, effect size = −0.290, t = −0.745, P = 0.456).
a Agroecosystems (335 studies). b Grasslands (25 studies). c Forests (15 studies). d Old-field ecosystems (15 studies). e Marine ecosystems (11 studies). f Wetlands (10 studies). g Shrublands (6 studies). Horizontal lines indicate the 95% confidence intervals around the means. Numbers in brackets indicate the numbers of observations and studies. Black, red, green, blue, turquoise, purple and orange lines denote plant antagonists, invertebrate herbivores, natural enemies of herbivores, weeds, plant-feeding nematodes, plant diseases and plants, respectively.
a Plot experiments. b Pot experiments. Horizontal lines indicate the 95% confidence intervals around the means. Numbers in brackets indicate the numbers of observations and studies. Black, red, green, blue, turquoise, purple and orange lines denote plant antagonists, invertebrate herbivores, natural enemies of herbivores, weeds, plant-feeding nematodes, plant diseases and plants, respectively.
a Herbaceous plants. b Woody plants. Numbers in brackets indicate the numbers of observations and studies. Horizontal lines indicate the 95% confidence intervals around the means. Black, red, green, blue, turquoise, purple and orange lines denote plant antagonists, invertebrate herbivores, natural enemies of herbivores, weeds, plant-feeding nematodes, plant diseases and plants, respectively.
a In temperate zones. b In tropical zones. Horizontal lines indicate the 95% confidence intervals around the means. Numbers in brackets indicate the numbers of observations and studies. Black, red, green, blue, turquoise, purple and orange lines denote plant antagonists, invertebrate herbivores, natural enemies of herbivores, weeds, plant-feeding nematodes, plant diseases and plants, respectively. Data from greenhouse and other indoor experiments are not shown in this figure.
At the ecosystem level, we found that plant genetic diversity showed a positive effect on plant performance in agroecosystems (effect size= 0.362, t = 8.275, P < 0.001), grasslands (effect size= 0.353, t = 4.670, P < 0.001), old-field ecosystems (effect size= 0.681, t = 5.573, P < 0.001) and marine ecosystems (effect size= 0.778, t = 3.753, P < 0.001) (Supplementary Table 4). Two plausible mechanisms could explain the positive effect of plant genetic diversity on plant performance. Firstly, an increased complementarity (i.e., niche partitioning or facilitation) or decreased intensity of plant competition among different plant genotypes8,9. Secondly, an increased net positive interactions with higher trophic levels (i.e., increasing genotypic polycultures resulted in a decreased herbivore abundance) that might amplify plant performance9. However, the positive effect of plant genetic diversity on plant performance was not consistently found in forests, wetlands or shrublands (Supplementary Table 4). This may be due to one or more of the following potential explanations: (i) fewer studies have been conducted in these ecosystems, (ii) at higher genotypic richness, genotype-by-genotype interactions resulted in lower relative performance of each genotype relative to the monoculture yield (i.e., trait-dependent complementarity became more negative at higher genotypic richness treatments)35, or (iii) plant genetic diversity indirectly decreased plant growth by increasing the abundance and species richness of herbivores7, as individual genotypes varied in their resistance and susceptibility to herbivory36,37.
Interestingly, plant genetic diversity decreased herbivore performance in agroecosystems (effect size = −1.008, t = −5.419, P < 0.001), grasslands (effect size = −0.961, t = −2.322, P = 0.026), forests (effect size = −0.151, t = −0.679, P = 0.497) and marine ecosystems (effect size = −0.289, t = −1.765, P = 0.078) (Supplementary Table 4). This reduction might reflect resource heterogeneity effects on foraging behaviour of herbivores28, as well as on herbivore movement38. Decreased herbivore performance under higher genetic diversity could also be explained by associational resistance in genotypic mixtures4, as many studies focused on control of a single herbivore species by mixing crop or plant genotypes with known differences in resistance to this herbivore. Alternatively, the lower insect herbivore performance under higher genetic diversity may be explained by the resource concentration hypothesis (RCH). Although RCH has been generally tested considering plant species diversity, the concept may be extended to plant genetic diversity, as herbivores have been shown to be able to distinguish between plant genotypes19. In addition, an increase in plant volatiles or plant secondary metabolites from the wide range of genetic and chemical diversity within plant species39,40, might contribute to the release of defensive chemicals to control or repel herbivores41, and thus result in a decreased herbivore performance in plant genotypic mixtures.
On the other hand, plant genetic diversity was associated with an increase in herbivore performance in old-field (effect size=1.102, t = 1.224, P = 0.221) and shrub systems (effect size=0.178, t = 0.912, P = 0.362). In this case, complementarity in resource use among plant genotypes might have increased plant growth and quality, resulting in increased herbivore abundance9,42,43. Similarly, herbivores may be attracted by plant volatiles from plant genetic diversity42,43. Yet, the increased herbivore abundance could also have been driven by associational susceptibility in genotypically diverse plots where the ramets of otherwise resistant genotypes could be attacked by herbivores due to their close proximity to susceptible genotypes44,45.
We also found that plant genetic diversity reduced (i) weeds in agroecosystems (effect size = −0.582, t = −3.350, P = 0.001; Fig. 2a, b); (ii) diseases in agroecosystems (effect size = −1.085, t = −5.211, P < 0.001), grasslands (effect size = −1.161, t = −3.031, P = 0.002) and old-field ecosystems (effect size = −2.659, t = −3.176, P = 0.002) (Fig. 2a–d), and (iii) plant-feeding nematodes in agroecosystems (effect size = −2.118, t = −1.313, P = 0.189; Fig. 2a), indicating strong biocontrol services. Such biological control effects may be an indication of genetic heterogeneity: (i) inhibiting the growth of weeds through allelopathic effects, detrimental plant secondary metabolites and growth competition12, (ii) diluting the concentration of resources or disrupting the movement of pathogens between host plants1, or (iii) creating rhizosphere inhibition zones against nematodes46. Generally, physical barriers and variety resistance may account for the effects of plant genetic diversity on decreases in insect herbivores, nematodes or diseases (e.g., altered dispersal and transmission rates of air-borne pathogens, splash-borne propagules or soil-borne bacteria)47.
Plant genetic diversity also had a direct influence on higher trophic levels across ecosystems. Specifically, we found that more genetically diverse plant stands supported more natural enemies of herbivores, such as predators and parasitoids (Fig. 1b). Such effects can be direct15,32, mediated by increases in herbivore abundance31,42, driven by trait-mediated indirect effects48,49, or directly mediated by natural enemies of herbivores through top-down effects as indicated by the enemy hypothesis (EH)4,50. According to the EH hypothesis, increasing plant genetic diversity results in a greater ‘resource pool’ for natural enemies of herbivores, which favours a greater abundance and diversity of predators and parasitoids, and this ultimately leads to stronger enemy top-down effects on herbivore populations. In addition, an increased diversity or amount of plant volatiles at higher plant genetic diversity can also attract more natural enemies to plants51. Such top-down effects may partly explain a positive influence of plant genetic diversity on plant performance. However, such top-down effects were not significant in forests (effect size=0.134, t = 0.602, P = 0.548) where genetically rich communities often grew more slowly and suffered higher levels of herbivory than genetic monocultures52. In addition, we found that tree genetic diversity showed a weaker association with the performance of natural enemies of herbivores and related biocontrol services. As forests are more complex systems compared to some other systems (e.g. agroecosystems), it is possible that even low-diversity forests could provide niches for many predator and parasitoid species53.
To assess whether the level of plant genetic diversity (i.e., the number of added genotypes, relative to the monogenetic control; see Methods and Supplementary Table 15) correlated with effect sizes, we set up generalised least-squares models with the fitted effect sizes as the response variable, and the log-transformed number of added genotypes in the plant genetic diversity treatment over the control as an explanatory variable. The average effect sizes of number of added genotypes on plant antagonist performance were all less than zero (i.e., negative effect size) and meanwhile the effect sizes significantly increased with the number of added genotypes across all studies (d.f. = 1734, t = 6.657, P < 0.001; Supplementary Fig. 1a) and for agroecosystems (d.f.=1534, t = 3.472, P = 0.001) or grasslands (d.f. = 53, t = 2.168, P = 0.035) (Supplementary Figs. 2a, 3a), but significantly decreased for forests (d.f. = 87, t = −2.040, P = 0.044; Supplementary Fig. 4a). This implies that adding only one plant genotype or more than one plant genotype can potentially inhibit plant antagonists in agroecosystems, grasslands and forests (because the average effect sizes of plant antagonists were all negative, no matter whether the trends of the meta-regression lines for plant antagonists increased in agroecosystems and grasslands or decreased in forests). However, these inhibitory effects on plant antagonists were diminished with the number of added plant genotypes in agroecosystems and grasslands but were enhanced with the number of added plant genotypes in forests (because the trends of the meta-regression lines for plant antagonists increased in agroecosystems and grasslands but decreased in forests, although the average effect sizes of plant antagonists were all negative in these three ecosystems). Weakened inhibitory effects on plant antagonists in agroecosystems and grasslands might be due to the fact that an increased complementarity in plant resource use, as resulting from the increase in the number of added plant genotypes9,42,43, may benefit plant growth (Supplementary Figs. 1g, 4g) and thus gives rise to a decrease in inhibitory effect. For example, compared with monoculture controls, the abundance of the invertebrate herbivores was lower in treatments with multiple plant genotypes and meanwhile such herbivore abundance increased with the increase in plant genotypes.
In agroecosystems12,38, grasslands49 or forests30,31, for example, intercropping and cover vegetation are commonly applied and the number of added plant genotypes used is often smaller (2–4 in general). Thus, enhanced pest control can be realised by adding another one-to-three plant genotypes in agroecosystems, grasslands or forests. However, there were no significant differences between adding one and adding more than one plant genotypes in old-field ecosystems (d.f.=18, t = 0.774, P = 0.449), marine ecosystems (d.f.=11, t = 0.115, P = 0.289) and shrublands (d.f.=21, t = 0.985, P = 0.336) (Supplementary Figs. 5, 6 and 8), which might be an artefact of fewer studies documenting plant antagonist responses to increased plant diversity in old-field ecosystems (N = 20), marine ecosystems (N = 13) and shrublands (N = 23). We found significantly positive relationships between plant antagonists and the number of added genotypes in plot experiments (d.f.=1580, t = 5.265, P < 0.001), pot experiments (d.f.=152, t = 2.121, P = 0.036), herbaceous plants (d.f. = 1579, t = 6.133, P < 0.001), woody plants (d.f.=153, t = 2.558, P = 0.012) or temperate zones (d.f. = 1488, t = 5.512, P < 0.001) (Supplementary Figs. 9–13), but no significant relationships in tropical ecosystems (d.f. = 138, t = 1.565, P = 0.120) (Supplementary Fig. 14). Overall, there were variable relationships between the number of added genotypes and the performance of herbivores, their natural enemies, weeds, nematodes, plant diseases or plants for individual ecosystem types, experimental study types, plant life forms and climatic zone types (Supplementary Figs. 1–14; Supplementary Table 16).
Effects of plant genetic diversity on trophic interactions
We obtained 1606 estimates of interactions between pairs of trophic levels derived from 163 studies testing the effects of plant genetic diversity across multiple trophic levels. First, we tested the effect of plant genetic diversity on bi-trophic interactions between plants and plant antagonists (1484 estimates derived from 139 studies) using multilevel piecewise structural equation models. In these models, different plant antagonists (i.e., invertebrate herbivores, weeds, plant-feeding nematodes or plant diseases) were considered together. We found that plant genetic diversity affected plant performance directly and indirectly by reducing the performance of plant antagonists (Fig. 6). The same pattern was also consistently found in agroecosystems, but not in other ecosystems (Supplementary Fig. 19). In these cases, plant genetic diversity often showed a direct effect on plant or plant antagonist performance, but no indirect effects on plant performance mediated by effects on plant antagonists. Furthermore, we found no evidence of an indirect effect of plant genetic diversity on plant performance through a reduction of plant antagonist pressure when we separately tested herbivore, weed or nematode performance (Supplementary Fig. 18a–c). However, we consistently found a direct effect of plant genetic diversity on the performance of plants and their antagonists. It is likely that further studies will be needed to validate such models. Indeed, when more data were available, as for the case of disease performance (N = 969), such mediating effects were evident (Supplementary Fig. 18d). However, when we tested the effects of number of added plant genotypes on the bi-trophic interactions, we found that the direct and indirect effects of number of added plant genotypes on plants, weeds and plant diseases were different from those of plant genetic diversity (Supplementary Fig. 15b–d), and that the direct and indirect effects of number of added plant genotypes on plants and plant antagonists were also different from those of plant genetic diversity across different ecosystems (Supplementary Fig. 16), experiment types, plant life forms and climatic zones (Supplementary Fig. 17), respectively.
The effects of plant genetic diversity (measured as standardised mean difference betweem genetically diverse and genetically monogenotypic plant stands) on bi-trophic interactions of invertebrate herbivore and plant performance, weed and plant performance, plant-feeding nematode and plant performance, and plant disease and plant performance are presented in Supplementary Fig. 18a–d, respectively. Plant genetic diversity is shown in dusty blue. Plant antagonist performance including herbivore performance (abundance, damage and diversity of herbivores), weed performance (growth and diversity of weeds), plant-feeding nematode performance (nematode abundance) and plant disease performance (disease spread and damage) is shown in beige circles. Plant performance (growth, quality and reproduction of plants) is shown in teal colour. Blue and red arrows denote positive and negative relationships, respectively; numbers next to each arrow are the estimated coefficients from piecewise structural equation models, and line width is proportional to the magnitude of the coefficients (Supplementary Tables 10, 11).
For a subset of studies (N = 91), where the effect of plant genetic diversity was investigated for all three trophic levels (plant-herbivore-natural enemy interactions), we tested effects of plant genetic diversity on tri-trophic interactions. Structural equation modelling showed a significant direct influence of plant genetic diversity on herbivore (estimate = −0.865, P = 0.016) and plant performance (estimate = 0.884, P = 0.010), but no significant indirect effects mediated via trophic cascades (plant genetic diversity vs. natural enemies: estimate= 0.709, P = 0.228; natural enemies vs. hebivores: estimate = −0.011, P = 0.906; herbivores vs. plants: estimate = −0.025, P = 0.647). Specifically, we found a positive effect of plant genetic diversity on herbivore natural enemy and plant performances and a negative effect on herbivore performance. These findings indicate a potential effect of plant genetic diversity on a tri-trophic cascade (Fig. 7). However, when we tested the effects of number of added plant genotypes on the tri-trophic interactions of plants, herbivores and their natural enemies, we found that the direct (natural enemies: estimate= 0.069, P = 0.657; herbivores: estimate = −0.035, P = 0.918; plants: estimate= 0.106, P = 0.440) and indirect (natural enemies vs. hebivores: estimate = −0.012, P = 0.905; herbivores vs. plants: estimate = −0.037, P = 0.511) effects were not significant (Supplementary Table 9; Supplementary Fig. 15a).
Plant genetic diversity is shown in dusty blue. Natural enemy performance (predator abundance, predator diversity, parasitoid abundance, parasitoid diversity and parasitism) is shown in pink. Herbivore performance (abundance, damage and diversity of herbivores) is shown in beige circles. Plant performance (growth, quality and reproduction of plants) is shown in teal colour. The blue and red arrows denote positive and negative relationships, respectively; numbers next each arrow are the estimated coefficients from piecewise structural equation models, and line width is proportional to the magnitude of the coefficients (Supplementary Tables 8, 9). The asterisks indicate the significance at 5% level.
Our synthesis comprehensively shows that plant genetic diversity directly increases plant performance in terrestrial and aquatic systems on Earth, which has been partially shown also for agroecosystems in which cultivar mixtures increased the yield of maize, legumes, wheat, oats, barley, soybean and sorghum18. Our results also indicate that plant genetic diversity promotes ecosystem services by strengthening trophic interactions: benefiting natural enemies of herbivores including invertebrate predators and parasitoids (in line with the conclusion that plant genetic diversity benefited predators from 162 estimates of effect size in 60 experimental studies21), in turn suppressing invertebrate herbivores, and enhancing plant performance; while also enhancing plant performance by suppressing weeds, plant-feeding nematodes and diseases. These findings contribute to explaining the mechanisms by which manipulation of plant genetic diversity may affect different trophic groups and their interactions. From an applied perspective, promoting plant genetic diversity may help promote both pest control in managed ecosystems and consumer control of plant production in natural ecosystems via strengthening top-down control, enhancing associated ecosystem functions and services. In conclusion, our results indicate that plant genetic diversity can help society, decision-makers and other stakeholders to take advantage of the important biocontrol services provided by plant genetic diversity on Earth.
Source: Ecology - nature.com
