Authors:
Carlos Bustos-Segura1, Erik H. Poelman2, Michael Reichelt3, Jonathan Gershenzon3,
Rieta Gols2
1. Evolution, Ecology and Genetics Division, Research School of Biology, The Australian National University, Canberra, ACT 2601, Australia.
2. Laboratory of Entomology, Wageningen University, PO Box 16, 6700 AA Wageningen, The Netherlands
3. Max Planck Institute for Chemical Ecology, Dept of Biochemistry, Hans-Knöll-Str. 8, D-07745 Jena, Germany
124
ABSTRACT
Intraspecific plant genetic diversity can modify the properties of associated
arthropod communities and plant fitness. However, it is not well understood which traits are involved in explaining the ecological effects of intraspecific plant
diversity and how interactions between neighbouring plants influence species interactions at different trophic levels. We explored the effect of diversity in plant traits on the composition and abundance of the insect community and consequent damage levels using three wild cabbage (Brassica oleracea)populations that differ in foliar secondary chemistry, i.e. glucosinolate composition and concentration. In a common garden experiment, intraspecific diversity was manipulated to establish plots with plants originating from one, two or three plant population sources. We found that plants were larger, harboured more herbivores and were less damaged when intraspecific plant diversity increased, whereas the effect on carnivore community abundance was less clear. Since plant populations co-varied with size, the effects of these factors are confounded. Nevertheless, both glucosinolate composition (i.e. presence of specific glucosinolates) and increased variation in glucosinolate concentrations among plants in a plot correlated positively with herbivore diversity and negatively with plant damage levels in more diverse plots. These results show that variation in secondary chemistry among neighbouring plants of the same species can play an important role in determining the effects of plant genetic diversity on insect community composition and suggest that diversity in functional traits helps to explain the outcome of the interaction between plants and herbivores.
125
INTRODUCTION
It has been widely recognised that biological diversity plays an essential role in structuring communities and ecosystem processes (Haddad et al. 2011; Naeem & Li 1997; Tilman et al. 1996; Tilman et al. 2006; Snyder et al. 2006, Yachi & Loreau 1999). Both inter- and intraspecific genetic diversity serve as a reservoir for variation in traits involved in ecological interactions across environments. Heterogeneity in the biotic and abiotic environment, in turn, influences the
evolution of populations to maintain trait diversity (Crawford et al. 2007; Hughes et al. 2008; Lankau & Strauss 2008; Wimp et al. 2004). In terrestrial ecosystems, plants provide most of the resources for higher trophic levels and shape the composition of communities. Empirical studies have shown that intraspecific variation in plant genetic diversity, to a large extent, drives the assembly of invertebrate communities associated with a plant species and this can result in effects on plant fitness (Crutsinger et al. 2006; Ferrier et al. 2012; Johnson et al. 2006; Kotowska et al. 2010; Lamit et al. 2015). Remarkably, intraspecific genotypic diversity can be as important for the associated community as interspecific plant diversity (Cook-Paton et al. 2011).
Usually, in experiments manipulating intraspecific plant genetic diversity, the general trend is that plants in more diverse plots tend to have a more diverse associated community and a higher individual plant fitness (Hughes et al. 2008; Moreira et al. 2016) . Genotypic variation among neighbouring plants accounts for some of the effects of plant diversity on the associated community. For example, it may influence the distribution and damage levels of herbivores on focal plants through processes referred to as associational resistance or susceptibility (Barbosa et al. 2009). Here, focal plants gain protection by growing next to less palatable plants (also referred to as the repellent-plant hypothesis) or become more susceptible to herbivory by a spill over effect of growing in the vicinity of highly attractive or palatable plants. Alternatively, plants growing close to more palatable plants might receive lower levels of herbivory (attractant-decoy
hypothesis; Barbosa et al. 2009; Plath et al. 2012; Ruttan & Lortie 2015; Zakir et al. 2013). Therefore, variation in chemical and physical traits may act directly upon the arthropods or indirectly through plant-plant interaction (e.g. resource
126
on the associated insect community. Also, neighbouring plants can influence the expression of a focal plant phenotype through plant-mediated changes in the soil microbial community (Eisenhauer et al. 2010; Kostenko et al. 2012).
Diversity in neighbouring plants with different genetic backgrounds has been shown to affect the associated community at different trophic levels and functional groups (herbivores: e.g Ferrier et al. 2012; Johnson et al. 2006; carnivores of herbivores: e.g. Crutsinger et al. 2006; Johnson et al. 2006, and pollinators: e.g. Genung et al. 2012). The community changes due to increasing genetic diversity can be understood as a result of the indirect genetic effects which can shape the structure and function of communities and can explain the non- additive effects observed in diversity experiments (Bailey et al. 2014). However, despite the knowledge gained from diversity experiments, we are still far from understanding the functional characteristics of the ecological effects produced by genetic diversity and how genetic variation in the expression of certain traits among neighbour plants determines those effects at the population and community level (Hughes et al. 2008; Moreira et al. 2016).
Different plant traits may be responsible for the diversity effects. For example, plant secondary metabolites (PSM) are often studied for their role in defense against insect herbivores and other plant antagonists (Iason et al. 2012; Fraenkel 1959; Schoonhoven et al. 2005), but they have been less well studied as factors influencing community structure at larger scales. Variation in PSMs can directly affect the abundance and composition of the associated herbivore community and the consequent amount of damage by influencing the ability of herbivores to find and or colonize specific plant phenotypes (Finch et al., 2003). The effects of PSMs are not restricted to the ecological interactions with a focal plant, but they can also indirectly affect the interactions with neighbouring plants through the emission of plant volatiles (Heil & Karban 2010; Jactel et al. 2011; Schuman et al. 2015; Zakir et al. 2013) and changes in the induction of defensive PSM (Kostenko et al. 2012) resulting in associational resistance or susceptibility. In addition to the effects that individual PSM have on herbivores, PSM diversity is also important in determining interactions with herbivores (Lankau & Strauss, 2007; Moore et al. 2014; Poelman et al. 2009; Richards et al. 2015). However, although the influence of PSM has been explored in the context of genetic diversity
127
effects (Parker et al. 2010), the PSM diversity among neighbouring plants has received less attention.
The aim of this study was to investigate to what extent plant chemical diversity among neighbouring plants accounts for changes in the associated invertebrate community, both herbivores and their natural enemies, and whether this has consequences for traits considered important for plant fitness such as size and herbivore damage. Plant - insect interactions have been extensively studied in the wild cabbage, Brassica oleracea (Moyes et al. 2000; Gols et al. 2008b; Newton et al. 2009) and its derived crops such as cabbage, broccoli and cauliflower
(Broekgaarden et al. 2010; Bukovinszky et al. 2008; Hambäck et al. 2009; Poelman et al. 2009). Like other plants in the Brassicaceae family, B. oleracea contains glucosinolates that interact with plant myrosinase upon tissue damage producing a range of compounds (Wittstock & Halkier 2002; Halkier & Gershenzon 2006) that are toxic to many generalist herbivores and directly or indirectly also affect growth and development of their natural enemies (Agrawal & Kurashige 2003; Gols & Harvey 2009; Hopkins et al. 2009; Winde & Wittstock 2011). Natural populations of B. oleracea growing along the south coast of England exhibit substantial
quantitative and qualitative genetic variation in foliar glucosinolates (Mithen et al. 1995). This variation has been shown to affect the presence and abundance of a range of herbivores on individual plants and among populations in the field (Moyes et al. 2000; Moyes & Reybould 2001; Newton et al. 2009, 2010).
In a common garden experiment, we established plots with plants from the same (monocultures) and different populations (dicultures and tricultures) using three natural B. oleracea populations that differ dramatically in constitutive and herbivore-inducible glucosinolate concentrations (Gols et al. 2008b; Harvey et al. 2011). We monitored the associated invertebrate community throughout the growing season and characterised glucosinolate profiles of individual plants. Then, we statistically analyzed how glucosinolates in focal plants, glucosinolate variation among neighbouring plants and plant size explained the composition of the
associated invertebrate community and the amount of plant damage. We discuss the implications of chemical diversity among neighbouring plants on the
interactions with herbivores and how the herbivore community responds to those effects.
128
MATERIAL AND METHODS
Biological system
Brassica oleracea is a biennial cruciferous plant with natural populations distributed in coastal Europe, with many populations located along the south coastline of UK. Several cultivars have been domesticated from this species by selecting different phenology traits, including cabbage, cauliflower, broccoli,
Brussels sprouts and kale. The associated herbivores to the wild populations range from specialist caterpillars like Pieris rapae and P.brassicae and aphids like
Brevicoryne brassicae, to several generalist herbivores like Mamestra brassicae and
Myzus persicae. The herbivore community has been well described in several studies (Root 1973; Moyes et al, 2000, Newton et al. 2009; Poelman et al. 2009). We selected three B. oleracea populations from Dorset, United Kingdom located within a range of 10 km. The locations and their coordinates were: Kimmeridge (50° 60' N, 2° 13' W), Winspit (50° 59' N, 2° 03' W), and Old Harry (50° 64' N, 1° 92' W). Seeds from a bulk random sample of the three plant populations were sown in seedling trays on April 14th of 2014 and grown under glasshouse
conditions at Wageningen University. By 19th of May, the seedlings were
transplanted into the experimental field located next to the Wageningen University campus (51° 59' 22" N, 5° 39' 59" W). The experimental design of the field
consisted of 63 plots (4×4 m each), separated by rows (width 3 m) of grass (Lolium
and Poa species). Each plot contained 25 plants in a square array of five by five plants placed 75 cm apart.
Common garden experiment
For exploring how plant diversity and chemical complexity affected the associated invertebrate community, we manipulated the frequency of the three plant
populations within the plots. Twenty-seven plots contained only one plant population (9 plots for each of the three plant populations), twenty-seven plots consisted of two plant populations (9 plots for each of the three possible
combinations), and 9 plots contained plants from each of the three populations. Hereafter, these diversity treatments are named monocultures, dicultures and
129
tricultures, respectively. The diversity treatments were spatially arranged according to a fully randomized design.
In the dicultures, plants from the same population were planted in alternate diagonals of the 5×5 plant matrix. Thus, each plant individual inside the plot core of nine plants had four neighbouring plants of another population at each side. The same procedure was applied to the tricultures, except that each plant individual in the plot core of nine plants had two plants of each of the two other populations as neighbouring plants.
Invertebrate community monitoring
The invertebrate community was monitored on the nine central plants of each plot, however, early in the season some plants died (thirty-eight plants from the plots core had died; no difference in mortality was found between plant population, diversity treatment or their interaction; GLM: all P>0.05). These plants were substituted by neighbor plants of the same population but outside the plot core. The upper and lower surfaces of each leaf of these plants were visually inspected and the number of individuals of each invertebrate species observed per plant was recorded. Given that the herbivore community has been well studied, identification at the species level was possible through visual examination in most of the cases (Table 1). Identification of carnivore species was done at higher taxonomic levels (Table 2). The monitoring was performed six times during the whole season from June to August, in week 23, 24, 25, 26, 28 and 32 of the year 2014. The diversity treatment and population origin of the nine focal plants in each plot was unknown to the person who did the monitoring. For each plant we obtained the abundance per herbivore and carnivore species and we calculated the Shannon's diversity index (H) for the herbivore community. We did not obtain the Shannon's index for the carnivore community because many carnivores were not identified at the species level. Given that Brevicoryne brassicae and its parasitoids (mainly Diaeretiella rapae) were the dominant species; their numbers were vastly higher than the other species. Including a log transformation on the abundance would still
130
for excluding these estimates from herbivore and carnivore abundances and analyzed them in separate tests.
Table 1. Herbivore taxa identified during the monitoring period and their feeding type and host specialization.
Table 2. Carnivore taxa included in the monitoring period, both parasitoids and predators were included.
Plant measurements
Plant size and damage
Leaves were counted and maximum height and width were measured for the plants in each plot during each monitoring. Plant size was calculated as the volume
Order Family Species Specialization Feeding type Lepidoptera Pieridae Pieris rapae Specialist Leaf chewer
Pieris brassicae Specialist Leaf chewer
Plutellidae Plutella xylostella Specialist Leaf chewer Pyralidae Evergestis forficalis Specialist Leaf chewer Noctuidae Mamestra brassicae Generalist Leaf chewer
Autographa gamma Generalist Leaf chewer
Lacanobia suasa Generalist Leaf chewer
Coleoptera Chrysomelidae Phaedon cochleariae Specialist Leaf chewer
Phyllotreta atra Specialist Leaf chewer
Phyllotreta undulata Specialist Leaf chewer
Coccinellidae Subcoccinella Generalist Leaf chewer Hemiptera Aphididae Brevicoryne brassicae Specialist Phloem feeder
Myzus persicae Generalist Phloem feeder
Aleyrodidae Alyrodes proletella Specialist Phloem feeder Thysanoptera Thripidae Thrips tabaci Generalist Cell content feeder
Diptera Leaf miners Generalist Tissue feeder
Pulmonata Snails/Slugs Generalist Leaf chewer
Order Family Species Feeding type Host Hymenoptera Braconidae Diaeretiella rapae Parasitoid Aphid
Praon sp. Parasitoid Aphid
Cotesia rubecula Parasitoid Caterpillar
Cotesia glomerata Parasitoid Caterpillar
Microplitis mediator Parasitoid Caterpillar
Ichneumonidae Diadegma semiclausum Parasitoid Caterpillar Coleoptera Coccinellidae Coccinella spp. Predator
Diptera Syrphidae Hoverflies Predator Cecidomyiidae gall midges Predator Neuroptera Chrysopidae Lacewings Predator Araneae/Opiliones Spiders and opiliones Predator
131
of a cylinder that could fit each plant given its maximum height and diameter. Total damage was estimated visually as a proportion of the total leaf area and through comparisons with photographs of plants with a wide range of damage levels. The proportion of damage ranged from 0 to 1 with levels taken to the nearest first decimal, making up 11 levels in total. Both plant size and damage levels were measured during the six time points across the season.
Glucosinolate analysis
In the week 33 of 2014, when monitoring had finished, leaf samples were taken with a cork borer (diameter 2.5 cm) from the sampled plants only in 60 plots since three plots had been flooded previously. From each plant, 12-15 discs were taken randomly from 5 fully developed leaves, wrapped in tin foil and immediately flash frozen in liquid nitrogen. Samples were freeze-dried until constant weight and ground to a fine powder. Ten to fifteen mg of freeze-dried and pulverised material per plant was used for glucosinolate analysis. Glucosinolates were extracted with 1 ml of 80% methanol solution containing 0.05 mM intact 4-hydroxybenzyl
glucosinolate as internal standard and were desulfataed with arylsulfatase (Sulfatase from Helix pomatia, Sigma-Aldrich) on a DEAE Sephadex A 25 column. The eluted desulfoglucosinolates were separated using high performance liquid chromatography (Agilent 1100 HPLC system, Agilent Technologies, Waldbronn, Germany) on a reversed phase C-18 UPLC column (Zorbax Eclipse XDB-C18, 50 x 4.6 mm, 1.8um, Agilent Technologies) with a water-acetonitrile gradient (2-6.5% acetonitrile from 0-3 min, 6.5-24.5% acetonitrile from 3-9min, followed by a washing cycle; flow 1.1 ml min-1). Detection was performed with a photodiode array detector and peaks were integrated at 229 nm. We used the following response factors: aliphatic glucosinolate 2.0, indole glucosinolate 0.5 (Burow et al. 2006) for quantification of individual glucosinolates in micromoles per dry mass of leaves (μmol·g-1 DM). Glucosinolates were identified by comparing the retention
times and UV absorption spectra with those of known standards (Reichelt et al. 2002). The following glucosinolates (Gls) were detected in order of elution: 3- methylsulfinylpropyl Gls (glucoiberin), R-2-hydroxy-3-butenyl Gls (progoitrin), 4- methylsulfinylbutyl Gls (glucoraphanin), 2-propenyl Gls (sinigrin), 3-butenyl Gls (gluconapin), indol-3-ylmethyl Gls (glucobrassicin), 4-methoxy-indol-3-ylmethyl
132
Gls (4-methoxyglucobrassicin), 1-methoxy-indol-3-ylmethyl Gls (neoglucobrassicin).
Statistics
For analysing the relationships between variables we first explored with linear models to what extent population origin of focal plants and the number of
populations within plots (hereafter diversity treatment) explained the variation in invertebrate community properties (abundance and diversity) and plant traits (plant size, plant damage and foliar glucosinolates) at the plant level. For these analyses we used the values per plant for each time point (six time points in total), except for glucosinolate data that was measured only once at the end of the season. Then we investigated whether there were correlations between glucosinolates (concentrations and composition, respectively) and herbivore community characteristics, as well as plant damage levels, by using the values per plant averaged across the whole season. All statistical analyses were performed in R (R Core Group, v 3.2.2). For all models we applied an ANOVA type II using the function
Anova (car package), which allows the interpretation of the main effects and their interactions independently and is preferred over ANOVA type III for unbalanced designs (Langsrud 2003). This function performs F-tests for general linear models and Wald’s chi-square tests for mixed models. A Tukey post-hoc test was applied whenever there was a significant effect of a factor variable with more than two levels. The analyses are described in detail below.
Invertebrate community. The effects of the diversity treatment on the invertebrate community characteristics (abundance and diversity of herbivores and carnivore abundance, respectively) was assessed using linear mixed models (LMM, lmer
package) with a repeated-measures structure. In these analyses, plant population origin of the focal plant, diversity treatment, their interaction term, and time of sampling (in weeks) were entered as fixed factors and plant ID as a random factor. For the abundance models we used Generalized Linear Mixed Models (GLMM) with a Poisson error distribution. The models mentioned above were also analyzed including plant size (log-transformed) as a covariate. Using a similar GLMM, we
133
analyzed the data on abundance of the aphid B. brassicae separately from that of the other herbivore species
Plant traits. The effects of diversity treatment and plant population on plant size were determined with a linear model using the logarithm of plant size as the response variable and number of plant populations within plot, plant population and week as fixed factors. Plant ID was entered as a random factor in the model. For analysing variation in plant damage, we used a similar model but added the invertebrate community factors that could be involved in determining the levels of damage. We included the diversity of herbivores and the natural logarithm of the abundances of carnivores and leaf chewing herbivores (Table 1), and their
interactions with plant diversity treatment as fixed factors. The response variable was the logit transformation of plant damage, calculated as:
𝐷(𝑙𝑜𝑔𝑖𝑡) = log 𝐷 + 0.05
1 − 𝐷 + 0.05
where D is proportion of damage, instead of using the arc-sine transformation as suggested by Warton & Hui (2011).
To test if total glucosinolate concentration differed among plant populations and with the number of plant populations within a plot we used a general linear model. For analysing differences in the glucosinolate composition among plant populations we performed a Partial Least Squares regression with Discriminant Analysis (PLS-DA, mixOmics package, González et al. 2011) which reduces the