Shifts in Plant–Microbe Interactions Over Community Succession and Their Effects on Plant Resistance to Herbivores

Research

Shifts in plant–microbe interactions over community succession and their effects on plant resistance to herbivores

Mia M. Howard1 , Jenny Kao-Knifﬁn2 and Andre Kessler3 1Plant Biology Section, School of Integrative Plant Science, Cornell University, Ithaca, NY 14853, USA; 2Horticulture Section, School of Integrative Plant Science, Cornell University, Ithaca, NY 14853, USA; 3Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, NY 14853, USA

Summary Author for correspondence: Soil microorganisms can influence the development of complex plant phenotypes, including Andre Kessler resistance to herbivores. This microbiome-mediated plasticity may be particularly important Tel: +1 607 254 4219 for plant species that persist in environments with drastically changing herbivore pressure, for Email: [email protected] example over community succession. Received: 28 October 2019 We established a 15-yr gradient of old-field succession to examine the herbivore resistance Accepted: 6 January 2020 and rhizosphere microbial communities of Solidago altissima plants in a large-scale field experiment. To assess the functional effects of these successional microbial shifts, we inocu- New Phytologist (2020) lated S. altissima plants with microbiomes from the 2nd,6th and 15th successional years in a doi: 10.1111/nph.16430 glasshouse and compared their herbivore resistance. The resistance of S. altissima plants to herbivores changed over succession, with concomi- Key words: above–belowground tant shifts in the rhizosphere microbiome. Late succession microbiomes conferred the interactions, herbivory, microbiome, strongest herbivore resistance to S. altissima plants in a glasshouse experiment, paralleling the rhizosphere, soil legacy effects, Solidago low levels of herbivory observed in the oldest communities in the field. altissima, succession, Trirhabda virgata. While many factors change over succession and may contribute to the shifts in rhizosphere communities and herbivore resistance we observed, our results indicated that soil microbial shifts alone can alter plants’ interactions with herbivores. Our findings suggest that changes in soil microbial communities over succession can play an important role in enhancing plant resistance to herbivores.

2018) and southern United States (Hakes & Cronin, 2012). This Introduction shift in resistance has been attributed to both rapid microevolu- Over the course of ecological succession, the relative importance tion and phenotypic plasticity, with the latter associated with of defending aboveground tissues from herbivores typically esca- changes in soil conditions (Hakes & Cronin, 2012; Howard lates for plants. This trend can be observed through changes in et al., 2018). Thus, changes in the soil structure, chemistry and plant life-history strategies: opportunistic pioneer plants that particularly the microbiome, may play an important role in shift- colonise early successional habitats tend to be fast growing, but ing plant resistance to herbivores over succession. relatively unprotected from herbivores, and thus are gradually Soil conditioning and plant–soil feedbacks can affect both the replaced by slower-growing, but more strongly defended, species succession of plant communities and their resistance to herbi- as the community ages (Cates & Orians, 1975; Bazzaz, 1979; vores. Successional shifts in soil biota have been implicated in Reader & Southwood, 1981; Grime, 2001). While this trend has driving the succession of plants in old-field communities. For been widely observed on a community level, little information is example, late successional plant species have been observed to known about how herbivore resistance might change within plant have effects on soil biota that promote the growth of other late populations over succession or the underlying mechanisms of succession species, while inhibiting the growth of early-colonisers these phenotypic shifts. (Kardol et al., 2006). In terms of affecting plants’ interactions The tall goldenrod, Solidago altissima (Asteraceae), may be a with herbivores, prior herbivory can produce lasting impacts on useful model for studying intraspecific successional shifts in resis- the soil that affects the resistance of subsequent plants (Kostenko tance phenotypes because it is an early coloniser, particularly of et al., 2012). This soil conditioning is not only likely to affect abandoned agricultural fields (old fields), that can persist for herbivore performance through changes in plant nutritional qual- decades as a dominant community member, even as communities ity (Kos et al., 2015), but also through altering plant defences. become forested (Bazzaz, 1968). The herbivore resistance of For example, soil conditioning can effect the expression of S. altissima has been observed to increase with successional age in defence-related genes and the production of secondary metabo- independent populations in both the northern (Howard et al., lites, as was recently observed in Plantago lanceolata (Zhu et al.,

Ó 2020 The Authors New Phytologist (2020) 1 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 2 Research Phytologist

2018). Recent evidence indicates that the diversity of plant com- both seeds and clonally via rhizomes, although the establish- munities conditioning the soil may affect the diversity of sec- ment of new genets is thought to be minimal after c. 5 yr of ondary metabolites produced by subsequent plants (Ristok et al., succession (Hartnett & Bazzaz, 1985; Maddox et al., 1989). 2019) and that conditioning by early successional forb species Thus, rapid evolutionary shifts in the later successional popula- may negatively impact their herbivore resistance (Heinen et al., tions are likely to be due to clonal sorting. It interacts with a 2018). Thus, over succession, soil conditioning by the different diverse array of insect herbivores (Maddox & Root, 1987), but communities of plants may contribute to changes in plant–herbi- at our field site the primary herbivore is the specialist vore interactions. chrysomelid beetle Trirhabda virgata. These beetles are capable Likely to be underlying these soil-mediated shifts in plant phe- of dispersing over long distances (hundreds of metres) as adults notypes are changes in the soil microbiota. With rapid generation (Herzig, 1995) and move from plant to plant as larvae (Morrell times, soil microbes have the potential to adapt much more & Kessler, 2016). quickly to selection pressures than their botanical counterparts and can drive ecological changes over succession. Soil microbial Old-field succession manipulation field experiment communities have been observed to shift rapidly over succession, not only in taxonomic composition, but also life-history strate- To examine changes in S. altissima and its community over suc- gies and functional relationships (for example pathogenicity) with cession, we used a large-scale, old-field succession sequence exper- plants (Hannula et al., 2017). Microorganisms are known to iment at Dunlop Meadow in Brooktondale, NY, USA affect complex plant phenotypes, such as growth (Vessey, 2003; (42°23013″N, 76°24000″W) (Howard et al., 2018). The field Gaiero et al., 2013) and flowering time or fitness (Lau & Lennon, site, on old agricultural land, consisted of duplicated 30 9 30 m 2012; Wagner et al., 2014; Panke-Buisse et al., 2015), as well as plots representing each of the first 6 yr of an old-field succession secondary metabolism (Ludwig-Muller,€ 2015). Microbially sequence (fallow after 3 yr of conventional maize agriculture) mediated shifts in plant secondary metabolite production have (Fig. 1). After growing maize, the fields are ploughed and been observed to have functional effects on important plant eco- colonised naturally from the surrounding environment, which logical interactions, including allelopathy (Meiners et al., 2017) largely consists of old fields, forest and agricultural fields. After and herbivore resistance (Hol et al., 2010; Badri et al., 2013). In 6 yr of succession, the plots are planted again with maize to addition to improving plant growth, inoculating crop plants with restart the sequence, with the exception of two plots that have individual strains of beneficial rhizobacteria often negatively been fallow since 2002. We refer to successional age as the num- affects aboveground insect pests (Pineda et al., 2010). Thus, we ber of years of succession that the communities have undergone hypothesised that shifts in plant defence phenotypes that we have at the time of sampling. previously observed along a successional gradient (Howard et al., 2018) would be at least be partially mediated by successional Plant and soil samples To characterise differences in plant shifts in soil microbial communities and plant–microbe interac- interactions with herbivores and microorganisms over succes- tions. sion, we surveyed 10 randomly selected S. altissima plant stems In this study, we characterised changes in the herbivore resis- per successional stage (five per plot) on two dates 1 yr apart (9 tance of S. altissima and its rhizosphere microbial communities June 2016 and 7 June 2017). We selected stems that were over early old-field succession in a large-scale, manipulated growing at least 3 m away from each other to avoid sampling chronosequence field experiment. We then examined the func- clones (Cain, 1990). Cumulative precipitation from January to tional effects of these microbial shifts on S. altissima herbivore June differed by 216 mm between these 2 years, with 2016 as a resistance phenotypes in a microbiome transfer experiment notable dry year (28% below the 1981–2010 average), particu- (Howard et al., 2019) with plants grown in a glasshouse. We pre- larly in comparison with the 2017 season (22% above the aver- dicted that the herbivore resistance of S. altissima plants would age) (Northeastern Regional Climate Center records for Ithaca, increase over succession, along with shifts in the microbial com- NY, USA). For each plant, we recorded size (height and num- munities colonising S. altissima rhizospheres. Based on the distri- ber of leaves), number of leaves with herbivory, and number of bution of herbivory in the field, we predicted that S. altissima Trirhabda sp. beetles (identified only to genus, as T. virgata and plants grown in soils inoculated with later succession micro- T. borealis are difficult to differentiate at the larval stage) biomes would be more resistant to herbivores than those inocu- (Messina, 1982a). To characterise differences in the microbial lated with earlier succession microbiomes. communities that interact with the plants, we sampled rhizosphere microbial communities by scraping the soil that Materials and Methods remained attached to the roots when the plants were uprooted into a tube with a spatula. We stored these samples at 20°C until analysis. We collected bulk soil samples (top 10 cm of soil) Study system adjacent to each plant and analysed them for microbial biomass The tall goldenrod, Solidago altissima L. (syn. S. canadensis), is a and activity, soil nutrients, pH and organic matter content. We perennial forb in the Asteraceae family that dominates old-field did not survey the rhizosphere microbial communities for plants habitats in its native range of northeastern North America and in the 6th year of succession in 2017 because the plots had is widely invasive across Eurasia. It colonises old fields through cycled back into agricultural production.

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 3

30 m N

Block 1 Block 2

1st year 4th year 5th year 2nd year succession succession succession maize

6th year 2nd year Late 4th year succession succession succession succession

3rd year 1st year 3rd year 1st year Fig. 1 Aerial view of the old-field succession maize maize succession succession field experiment. We manipulated community successional age using experimental plots that are staggered chronologically to represent each stage of a 9-yr sequence: 3 yr of conventional maize agriculture followed by 6 yr of fallow old- 3rd year 2nd year 3rd year 1st year field succession. In late May/early June of succession maize maize maize each year, the plots advance to the next stage in the sequence and plots that have undergone 6 yr of old-field succession are ploughed, treated with herbicide and planted with maize. Each stage of the chronosequence is represented by one plot within each of two blocks. There are also two 5th year late 6th year 2nd year older plots, labelled as ‘late succession’, that succession succession succession succession have been fallow since 2002 (representing the 13th,14th or 15th yr of succession in this study). Each plot is 30 9 30 m with at least 10 m of mowed grassland between each plot. Photograph credit: Jere Salonen.

each plant. We excised three to five leaves (3rd to 8th youngest Herbivore resistance bioassays To compare the herbivore resis- fully expanded) from each plant to provide c. 40 cm2 of leaf tissue tance of S. altissima plants growing in different successional to individual larvae in a 9 cm diameter Petri dish with moist filter stages, we performed bioassays with plants collected from the paper to maintain humidity. After 90 h of feeding, we quantified field and Trirhabda sp. We collected stems from 20 individual the amount of leaf tissue consumed via image analysis with randomly selected plants from each successional year (1–6 and Adobe Photoshop and recorded larval weight gain. We calculated 14) on 7 June 2017 and stored them as cut stems in water the efficiency of biomass assimilation as mg weight gained per overnight at room temperature. We collected second instar cm2 of leaf eaten. We omitted one replicate from each of the 5th Trirhabda sp. larvae from a field adjacent to the experimental and 14th year of succession due to no feeding and missing final plots on 7 June 2017 and stored them at 4°C with their food area data, respectively. (field collected S. altissima leaves) removed 15 h before the bioas- To compare the preference of herbivores for plants from differ- says on 8 June 2017. ent successional stages, we performed choice tests with Trirhabda As a measure of herbivore resistance, we quantified the effi- sp. larvae in the laboratory. Results from Petri dish choice assays ciency of biomass assimilation of larvae feeding on leaves from

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 4 Research Phytologist

with Trirhabda sp. larvae (Uesugi et al., 2013) have previously sequences) and proportions (total sums scaling) (McKnight et al., corresponded with their behaviour, that is avoidance of herbi- 2019). For comparison, we additionally normalised the commu- vore-induced plants, in the field (Morrell & Kessler, 2016). In nity data using cumulative sums scaling via the package the choice test, we simultaneously presented larvae with leaf tissue METAGENOMESEQ (Paulson et al., 2013). from each of three successional years: 1 (early), 5 (mid), or 14 (late) in a three-way cafeteria-style feeding choice test. We ran- Microbiome transplant experiment in the glasshouse domly grouped plants into 20 early–mid–late combinations and tested each combination in triplicate with different individual lar- In order to assess the functional effects of successional shifts in vae, for 60 tests. We cut leaf discs (13 mm diameter) from the the soil microbiota on plant phenotypes, we generated soil micro- 2nd youngest fully expanded leaf of each plant and presented biome inoculants from three different successional stages (2, 6, these discs equidistantly to each larva on 9 cm diameter agar (21 g and 15 yr of succession, as well as a mock inoculant) and assessed l 1) plates. After 3 h of feeding, we measured the area of each disc their ability to modify plant host growth and herbivore resistance consumed using IMAGEJ software (Schneider et al., 2012). We traits. We utilised six plant genotypes, which were collected from omitted six tests with no feeding from analysis. Dunlop Meadow and propagated vegetatively under standardised glasshouse conditions for at least two reproductive cycles, in an Soil analysis To assess changes in soil quality and microbial attempt to remove previous acclimation. Each genotype 9 inocu- activity over succession, we analysed six soil samples from each lant treatment combination was performed with two or three successional stage per season for pH, content of organic matter replicates, resulting in 17 plants in each inoculant treatment and nutrients (K, P, Ca, Mg), and soil respiration (2016 samples group. only) by the Cornell Nutrient Analysis Laboratory (Ithaca, NY, USA) using standard methods (United States Department of Plant inoculation and growth We collected c. 4 l of soil for use Agriculture, 2014). A subset of the 2017 soil data (pH, organic as inoculants from the top 10 cm of one Dunlop Meadow plot in matter, K, P, Ca and Mg) had been published previously each of years 2, 6 and 15 of succession on 22 June 2017. We (Howard et al., 2018). We measured microbial biomass as micro- sieved these inoculants to 4 mm and stored them at 4°C for 4 d. bial C and N in 10 samples per successional stage from the 2016 To control for microorganisms in the growth environment and season using a chloroform fumigation method (Vance et al., soil properties of the inoculants, we also used a mock inoculant 1987). Briefly, we fumigated fresh soil samples (c. 7 g) with chlo- treatment derived from field soil comprised of an equivolumetric roform for 24 h to release C and N from microbial cells, and then mixture of all three successional inoculants that was triple auto- extracted them in 0.05 M K2SO4 for 2 h. We used a Shimadzu claved with 24 h rest periods between sterilisation cycles. We pre- TOC-L TNM-L ASI-L (Shimadzu Corp., Kyoto, Japan) to mea- pared the plant growth medium on 26 June 2017, which sure total organic C (TOC) and total N (TN) in these extracts. consisted of 75% (v/v) commercial sphagnum moss potting To calculate microbial C and N, we subtracted the TOC and TN medium (Lambert’s All Purpose, Quebec, Canada), 20% (v/v) of extracts from baseline (nonfumigated) samples. topsoil, and 5% (v/v) field soil inoculant, which is an inoculation rate previously determined to transfer a representative soil micro- Microbial community analysis We assessed the bacterial and bial community to potting medium (Howard et al., 2017). We fungal community composition of rhizosphere soil samples using autoclaved the medium (minus the inoculant) three times with high-throughput amplicon sequencing of the V3–V4 region of 24 h between sterilisations, before use. We watered the prepared the 16S rRNA gene (16S) and the ITS2 region of the fungal medium to a 10% (v/v) moisture level with filter-sterilised internal transcribed spacer (ITS) gene regions, respectively, using (0.1 lm pore size; Sawyer Products Inc., Safety Harbor, FL, a similar procedure to that documented in Howard et al. (2017), USA) deionised water and incubated for 24 h at ambient temper- with modifications described in Supporting Information Meth- ature (c. 27°C) in the glasshouse (Cornell University, Ithaca, NY, ods S1. We performed all sequence processing using the platform USA) in sterilised 11.5 cm diameter standard pots before plant- QIIME (Caporaso et al., 2010). After merging paired-end ing. We introduced the plants as surface-sterilised crown cuttings sequences, trimming primers and removing singletons using (rinsed in 70% ethanol for 30 s followed by sterilised water) on MOTHUR v.1.41.1 (Schloss et al., 2009), we used VSEARCH v.2.9.1 27 June 2017. We positioned the inoculated plants in a ran- (Rognes et al., 2016) to cluster the sequences into de novo opera- domised block design in a glasshouse with a 16 h photoperiod tional taxonomic units (OTUs) with 97% sequence similarity and irrigated ad libitum with filter-sterilised water. To evaluate and classified them taxonomically using the databases the effects of successional microbiomes on plant growth, we cal- GREENGENES v.13.8 and UNITE v.7 for 16S gene and ITS culated growth rates based on increases in height and leaf number sequences, respectively. After processing, we had 1 433 721 16S from measurements taken every 3 d from 18–30 d after planting. and 3 192 634 fungal ITS reads, representing 13 592 bacterial and 5431 fungal OTUs, respectively. We removed samples with Composition of microbiome inoculation We froze three sam- <1000 reads, which resulted in omitting two bacterial samples ples of each inoculant at the time of inoculation for microbial from the 2017 season (both from the 5th year of succession), and community analysis. To test that the different inoculants assem- normalised the samples using two different methods recom- bled different rhizosphere microbial communities, and to com- mended for community comparison: rarefying (to 1000 pare these communities to the inoculants, we collected

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 5 rhizosphere soil from each plant 49 d after planting by first constitutive defence phenotype, we compared the plants’ volatile removing loose soil adhering to roots, and then collecting soils secondary metabolite profiles (Methods S2). from roots shaken vigorously in a clean plastic bag. We consid- ered the soil that was most closely adhering to root the ‘rhizo- Statistical analysis sphere’ soil and stored it at 20°C in sterilised plastic tubes until DNA extraction. We extracted and analysed microbial DNA We performed all statistical analyses in R v.3.5.1 (R Core Team, from these samples using the same methods as described for the 2018). We treated successional ages as factors in all analyses, field survey samples (Methods S1), except that we decreased the except where noted, and visually assessed the residuals of each lin- mass of soil used to c. 0.15 g, due to the greater absorption of the ear model for normality and homogeneity of variance, transform- extraction buffer by the potting media compared to the field soil. ing the data if necessary. For all models, we assessed the We used dimethyl sulfoxide (DMSO) in all ITS amplifications significance of main effects using the function anova in lMERTEST and the samples were sequenced with the 2017 field samples. (Kuznetsova et al., 2017) and examined pairwise contrasts using After sequence processing, we had 1119 059 16S and 1959 450 the function emmeans in the package EMMEANS (Lenth, 2019) ITS reads, representing 9666 bacterial and 6356 fungal OTUs, with the Tukey method P-value adjustments for multiplicity. We respectively. We rarefied the samples to 1000 reads and one bac- used lmer in LME4 for all linear mixed-effect models (Bates et al., terial sample of the initial 6th year field soil inoculant was 2015). removed due to low reads. In the field survey, we analysed herbivory density and damage using linear mixed-effects models with successional age, field sea- Plant resistance to herbivory As a measure of insect resistance, son, and the successional age 9 field season interaction as main we assessed the preference of T. virgata for plants grown in soils effects and field plot (block) as a random effect. For the herbivore inoculated with the three different successional microbiomes in a performance, microbial biomass, soil respiration, and soil organic three-way cafeteria-style choice test, within each plant genotype, matter data we used linear mixed-effects models with successional 38 d after planting. At this stage, the plants were approximately age as a main effect and field plot as a random effect. We the size at which they start becoming damaged by Trirhabda spp. removed one sample from the 6th successional year, which had in the field. We replicated each test (three-way grouping of indi- 70% and 62% higher microbial C and N, respectively, than any vidual source plants from each inoculation treatment) in tripli- other sample, as an outlier. For the herbivore performance analy- cate with a different T. virgata individual, for 51 tests. We sis, we divided successional age into three categories (early, 1–3 collected adult T. virgata from a wild population near Beebe Lake yr; mid, 4–5 yr; and late, 14 yr). Due to the large proportion of (Ithaca, NY, USA) on 1 August 2017, fed them on fresh zeroes in the data for the area of leaf tissue eaten in the feeding S. altissima leaves from that area for 2 d in the laboratory, and preference test, we used a binomial generalised linear mixed starved them for 20 h before the choice test. We cut leaf discs model (using glmer in the package LME4) to model the binary (13 mm diameter) from the second youngest fully expanded leaf variable indicating whether any of the leaf disc from a succes- of each plant and presented them equidistantly to each larva on sional year was eaten or not, rather than comparing the amounts 9 cm diameter agar (21 g l 1) plates. After 1 h of feeding, we of leaf tissue consumed between successional years. We treated measured the area of each disc consumed. We omitted four trials successional age as a main effect, and individual plants and test with no feeding. replicates as random effects. We assessed pairwise contrasts between the different successional ages using the function emmeans in the package EMMEANS and calculated Wald and para- Secondary metabolite analysis metric bootstrap 95% confidence intervals (CIs) on the log odds To compare levels of constitutive secondary metabolite produc- ratios using emmeans and the confint function in LMER4, respec- tion, we collected the 2nd youngest fully expanded leaf of each tively. We visualised dissimilarity in microbial community com- plant for analysis of phenolics and diterpenes 38 d after planting. position between samples using nonmetric multidimensional We excised the leaf at the petiole and removed the midvein with scaling (NMDS) on Bray–Curtis distances and tested for effects a razor, and then immediately flash-froze the tissue in liquid N2 of successional age and field season with PERMANOVAs (999 and stored it at 80°C until extraction. We extracted the leaf permutations) using metaMDS and adonis in the package VEGAN samples in 90% methanol and analysed them on an Agilent 1100 (Oksanen et al., 2018). We analysed the bacterial and fungal series high-performance liquid chromatograph (HPLC) as community data separately. Before calculating the Bray–Curtis described in Howard et al. (2018). We identified peaks of 18 distances, we square-root transformed the fungal OTU relative phenolic compounds (chlorogenic and coumaric acid derivatives, abundance data. To confirm the robustness of our results in light and flavonoids) and 12 diterpene acids (pimaric or abietic acid of potential rarefaction noise, we performed the rarefaction pro- derivatives) (Table S1) to compound class using UV spectral cedures and PERMANOVAs 2000 times with nearly identical information and quantified their signal intensity at 320 nm (caf- results. To test for differences in the relative abundance of basid- feic and coumaric acid derivatives), 360 nm (flavonoids), 210 nm iomycetes over succession in the fungal data, we used a linear (pimaric acid derivatives), or 230.8 nm (abietic acid derivatives), mixed-effects model with successional age (as a continuous vari- and standardised the area of each peak by dividing it by the mass able) and field season as main effects and block as a random of leaf tissue in each extraction. As another measure of effect.

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 6 Research Phytologist

For the microbiome transplant experiment, we analysed from the 1st year of succession than plants from the late succes- height- and leaf-based growth over time using repeated-measures sion (Fig. 2c; Tables S4, S5). Larvae feeding on leaves from late analysis of variances (ANOVAs) with inoculant, measurement succession plants assimilated biomass less efficiently than those date, and genotype (and their interactions) as main effects and feeding on the mid-succession plants that they preferred to eat individual plant as a random effect. We analysed microbial com- (Fig. 2d; Table S6). munities of the inoculants and assembled rhizosphere using NMDS ordinations of Bray–Curtis distances (as described for Plant size There were significant effects of both successional age the field survey microbiome data) and assessed the effects of inoc- and field season on S. altissima stem heights, which were tallest at ulant successional age and plant genotype using PERMANOVAs. the latest successional stage, but slightly decreased in height from In the herbivore feeding preference assay, we assessed the effect of early to mid succession (Fig. S1a; Table S7). There were also sig- the inoculant on whether any tissue was eaten using the same nificant effects of successional age and field season on the total model design as for the field feeding choice test (except with inoc- number of leaves per plant, but while the pattern generally paral- ulant treatment as a main effect, and with plant genotype as an leled that of plant height, the differences were more subtle additional random effect) and assessed pairwise contrasts and CIs (Fig. S1b; Table S8). using the same methods. To assess the overall effect of inoculation on feeding preference, we used a likelihood ratio test com- Soil properties and microbial composition Soil nutrients and paring our model to a reduced model without the main effect of physical properties varied over succession (Table S9). Soil micro- inoculant. We analysed the overall leaf metabolite profiles using bial biomass in the soil increased significantly over succession in NMDS on Bray–Curtis distances and tested for effects of inocu- terms of both microbial N and microbial C extracted from soils lant and genotype using a PERMANOVA as described for the (Fig. 3a,b; Table S10). Soil respiration, a proxy for soil microbial microbial community analyses. We assessed the effect of inocu- activity, also increased over succession (Fig. 3c; Table S10), along lant on the concentrations of individual compounds with linear with soil organic matter (Fig. 3d; Table S10). The composition mixed-effects models with inoculant and plant genotype as main of both bacterial and fungal communities in the S. altissima rhi- effects and block (position in glasshouse) as random effect. We zosphere shifted significantly over succession and differed adjusted the P-values for the 30 compounds with the false discov- between the two field seasons, as shown in the NMDS ordina- ery rate correction via p.adjust in stats to correct for multiple com- tions with the successional ages of the plots indicated by increas- parisons. One replicate inoculated with a 6th year microbiome ing colour intensity (Fig. 4a,b; Tables 1, 2). The relative was omitted as an outlier from analysis of one diterpene com- abundance of different bacterial phyla represented in the rhizo- pound (#20), as well as the overall profile analyses, because its spheres were relatively similar over succession (Fig. 4c). By con- measured concentration was 134-fold greater than the mean. trast, the relative abundance of fungal basidiomycetes increased significantly over succession (Fig. 4d; Table S11), with a concomitant decrease in ascomycetes. The microbial community Data availability analyses based on data normalised via proportions (Fig. S2; The microbial sequence data are available in the NCBI Sequence Tables S12, 13) and cumulative sums scaling (Fig. S3; Tables Read Archive under the accession no. PRJNA533029. All other S14, S15), were similar to those based on the rarefied data shown data are available in the Cornell eCommons repository (https:// in Fig. 4. ecommons-cornell-edu.proxy.library.cornell.edu/) under the title of this publication. Microbiome transplant experiment in the glasshouse The inoculated rhizosphere communities were compositionally Results distinct from the initial bulk soil inoculants, yet differed by inoculant successional age (Figs S4, S5). In addition to a significant Field survey effect of the successional inoculant age, there was also a signifi- Herbivore damage and resistance Both herbivore density and cant effect of plant genotype on the composition of bacteria and damage levels on S. altissima plants differed over successional fungi in the S. altissima rhizosphere community (Tables S16, time, as well as by field season (Fig. 2a,b; Tables S2, S3). Par- S17). While the microbial community compositions varied most alleling patterns of herbivore density, we observed relatively drastically over succession between the bulk soils used as inocu- low levels of herbivory in the first couple of years after agricul- lants, they assembled relatively consistent, although still distinct, ture. However, after 3 or 4 yr of succession, herbivory peaked, communities in the S. altissima rhizosphere (Fig. S5), similar, at with damage on nearly every leaf of every plant, before drop- the phylum level, to the communities that were observed in ping significantly in late succession populations (13–14 yr) S. altissima rhizospheres in the field (Fig. 4c,d). The rhizospheres (Fig. 2a,b). of plants treated with the mock (autoclaved) inoculant differed Consistent with the distribution of herbivory, T. virgata larvae from those treated with the three successional inoculants (Figs were more likely to eat plants from the mid rather than late suc- S4, S5). The inoculation treatments did not significantly affect cessional stage in a three-way choice experiment, although they plant growth rates based on height or leaf number (Tables S18, were also significantly more likely to consume leaves of plants S19).

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 7

A (a) BC C (c) 5 Season A 2016 7.5 2017 0.6 1 14 Fig. 2 Levels of herbivory and herbivore resistance of Solidago altissima plants over old-ﬁeld succession. (a) Number of Trirhabda B 5.0 0.4 sp. larvae; and (b) overall herbivore damage A (% of leaves damaged) per plant over B succession based on surveys of a large-scale Larvae per plant A 2.5 B 0.2

ﬁeld experiment (see Fig. 1). (c) Feeding Probability of eating leaf preference of Trirhabda sp. larvae in a three- A way choice test between discs of leaf tissue AAA A A A 0.0 from S. altissima plants collected from years 0.0 1, 5 or 14 of succession. Values shown are 1234513/14 1 5 14 estimated marginal mean (SE) probabilities D D (b) (d) A of eating discs from each successional age E E EDE 15 AB and ages not connected by the same letter 100 B C indicate signiﬁcant differences based on C whether any leaf tissue was eaten from each 75 successional year during the feeding trial, 10 based on a binary logistic model. (d) eaten Performance (biomass gained per cm2 of leaf –2 B 50 A A eaten) of Trirhabda sp. larvae feeding on A A B B leaves from early (1–3 yr), mid (4–5 yr), or B C 5 late (14 yr) succession for 90 h. For box plots, C 25 boxes enclose the middle 50% of values, A mg gained cm Leaves damaged (%) with the 50th percentile indicated by the midline. The error bars span 1.5 times the 0 0 interquartile range in both directions, and 1−3 4−5 14 values outside of this range are indicated as 1234513/14 black dots. Successional age (yr)

However, there was a significant effect of microbiome succes- to herbivores, particularly in contrast to mid-succession plants, sional age on the likelihood that the adult T. virgata beetles on which herbivore densities were highest (Fig. 2). While herbi- would consume the inoculated plants in a feeding choice assay vore abundance and damage were relatively low during the first (Fig. 5). The beetles preferentially fed on plants with earlier suc- couple years after agriculture, this may be because the popula- cession microbiomes; specifically, they were significantly more tions of Trirhabda spp. (and other herbivores) had not yet estab- likely to consume plants inoculated with second year microbial lished, even though adult Trirhabda spp. beetles are capable of communities than with their 15th year-inoculated counterparts dispersing over distances larger than our field experiment (Fig. 5; Tables S20, S21). (Herzig, 1995). Our bioassay data indicate that – despite their While the overall leaf metabolite profiles, and the concentrations low levels of herbivory in the field – these very early succession of many individual compounds, varied significantly by plant geno- plants were not necessarily more resistant than their mid-succes- type, there was no overall effect of successional microbiome inocu- sion counterparts and were more susceptible to herbivores than lants (Tables S22, S23). Nevertheless, the concentrations of a few late succession S. altissima, at least based on herbivore preference individual compounds, particularly two diterpene acids (com- (Fig. 2c). This trend is evident even despite mid-succession plants pounds #25 and 28) appeared to be influenced by the inoculation experiencing higher levels of herbivory before the bioassay. As successional stage (Fig. S6), but after correcting for multiple com- such, these plants were more likely to be expressing herbivore-in- parisons (30 leaf compounds were assessed), there were no signifi- duced, as well as constitutive, plant defences, although these cant effects of inoculant treatment detected for any leaf compound induced defences may not have as strong effects on Trirhabda at the a = 0.05 significance level (Table S23). Likewise, the volatile spp. as other herbivores (Bode et al., 2013; Uesugi et al., 2013). organic compound (VOC) profiles of the plants varied significantly Furthermore, it is possible that abundances of natural enemies of by genotype, but not inoculant treatment (Table S24). Trirhabda spp. and other herbivores varied over succession and may affect levels of herbivory, but we did not assess this in our Discussion survey. Along with changes in herbivore resistance, we simultaneously observed shifts in the microbial communities that colonised the Plants’ interactions with herbivores and microbes shift over rhizospheres of these S. altissima plants over succession (Fig. 4). succession Previous studies have demonstrated that compositional shifts in We found that S. altissima plants in the latest successional stage soil microbial communities occur as fields are restored from agri- studied in our field experiment (13–14 yr) were the most resistant culture (Barber et al., 2017; Hannula et al., 2017; Morri€en et al.,

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 8 Research Phytologist

(a) 0.25 (c)

1.5 C 0.20 D D

dry soil B soil B –1 C D –1 B

0.15 g B B C A B 2 C A 1.0 B B B 0.10 B B

A mg CO 0.05 0.5 A mg microbial N g

012345613 012345613 Fig. 3 Soil microbial biomass and activity, and soil organic matter, over old-ﬁeld (b) 1.5 (d) C succession. Microbial biomass in topsoil 6 D based on microbial (a) nitrogen (N) and (b) B carbon (C) was measured using a chloroform C B C 5 fumigation method. (c) Microbial activity dry soil 1.0 C D D A B was measured as soil respiration. (d) Soil

–1 A A D E E B A organic matter was determined by loss on A B B 4 B ignition. All soil samples were collected from C B the top 10 cm of soil in a large-scale ﬁeld A A 0.5 experiment (see Fig. 1) in 2016. Years not A A 3 connected by the same letter indicate

Organic matter (%) signiﬁcant differences. Boxes enclose the middle 50% of values, with the 50th 2 mg microbial C g 0.0 percentile indicate by the midline. The error bars span 1.5 times the interquartile range in 012345613 012345613 both directions, and values outside of this are Successional age (yr) indicated as black dots.

2017), but did not focus on the rhizosphere of a single plant rhizosphere colonisation (Tables S16–S17). As our previous species. Most strikingly, we observed an increase in the relative study at the same field site (Howard et al., 2018) suggested that abundance of basidiomycetes, which functionally are thought to the increased herbivore resistance of S. altissima populations over be responsible for degrading recalcitrant plant compounds, such the first 6 yr of old-field succession was largely due to microevolu- as lignin, in the later stages of decomposition (Baldrian, 2008), tion, plant genotypic shifts may partially explain the changes in and thus are likely to have increased in abundance with the rhizosphere microbes. Previous work with S. altissima has also increase in woody plants in these later succession communities. indicated that strong variation in root metabolites, particularly Work by Morri€en et al. (2017) further suggested that these suc- polyacetylene compounds, can rapidly evolve (Uesugi & Kessler, cessional shifts in the fungal community are functional and may 2013, 2016), potentially providing an avenue to directly influ- result in increased carbon uptake in the soil. Notably absent from ence plant–microbe interactions in the rhizosphere. Host geno- our rhizosphere surveys were arbuscular mycorrhizal fungi (c. 0.3- type has been widely observed to affect microbiome assembly % relative abundance of Glomeromycota in samples) (Fig. 4). across the plant kingdom, ranging from annual herbs and crops However, as S. altissima is known to form symbioses with mycor- (Lundberg et al., 2013; Peiffer et al., 2013) to trees (Cregger rhizal fungi (Jastrow & Miller, 1993), this is likely to be an arte- et al., 2018). Likewise, communities of soil microbes can also fact of using the ITS2 rDNA region to characterise fungal taxa influence selection on plant populations (Lau & Lennon, 2011). (Stockinger et al., 2010). While we may be missing potential suc- Considering that natural selection is likely to act on plants as an cessional shifts in relationships with mycorrhizal fungi, our data integrated phenotype of both the plant and its associated micro- clearly indicated microbial shifts in the S. altissima rhizosphere bial community, it is important to understand the extent to over succession. We hypothesised that these belowground which microevolutionary changes in host populations affect their changes would mediate, at least partially, the changes in herbi- interactions with their microbes – and vice versa – as the mecha- vore resistance we observed, and the results of our microbiome nisms of these interactions will drive community structure and transplant experiment supported this hypothesis (Fig. 5). How- dynamics. ever, before discussing this focus of the study, there are several As S. altissima is a perennial, plant age and ontogeny may also potential explanations why we see this shift in the composition of contribute to the rhizosphere microbial shifts we observed. While the S. altissima rhizosphere over succession. S. altissima is likely to be continuously colonising via seed Changes in the composition of S. altissima populations over through the first 5 yr of succession (Hartnett & Bazzaz, 1985), succession may contribute to shifts in their rhizosphere composi- the average age of plants – particularly in the later successional tion over succession. We observed plant genotypic effects on communities (13–14 yr) – is likely to have increased over

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 9

(a) Successional age: Season: (c) 100 1142016 2017 80

0.2 60

0.0 20 Relative abundance (%) NMDS2 0 (d) 100 −0.2 80

60 −0.4 −0.2 0.0 0.2 0.4 NMDS1 (b) 0.4 40

20 Relative abundance (%) 0.2 0 Season: 2016 2017 2016 2017 2016 2017 2016 2017 2016 2017 2016 2016 2017 1 2 3 4 5 6 13 14 0.0 Successional age (yr) NMDS2 Bacterial phyla: Fungal phyla: Actinobacteria Ascomycota Proteobacteria Basidiomycota −0.2 Firmicutes Blastocladiomycota Acidobacteria Chytridiomycota Bacteroidetes Glomeromycota Verrucomicrobia Rozellomycota Planctomycetes Zygomycota −0.50 −0.25 0.00 0.25 Other Unclassified NMDS1 Fig. 4 Compositional shifts in Solidago altissima rhizosphere microbial communities in old-field succession experiment surveyed in two sequential field seasons. Nonmetric multidimensional scaling (NMDS) ordinations of (a) bacterial and (b) fungal communities based on Bray–Curtis distances. Communities were characterised based on 16S rRNA (bacterial) and internal transcribed spacer (ITS) (fungal) gene amplicon sequencing of soil collected from S. altissima rhizospheres over an experimental gradient of old-field succession (see Fig. 1) over two growing seasons (2016 and 2017). Phylum-level composition of (c) bacterial and (d) fungal communities based on mean relative abundance across samples. Community data were normalised via rarefication.

Table 1 Results of PERMANOVA showing the effects of successional age Table 2 Results of PERMANOVA showing the effects of successional age and field season on bacterial community composition of Solidago altissima and field season on fungal community composition of Solidago altissima rhizospheres in the field. rhizospheres in the field.

df Sum of squares F-value P-value df Sum of squares F-value P-value

Successional age 7 3.0170 2.1206 0.001 Successional age 7 5.084 2.7897 0.001 Field season 1 0.6505 3.2005 0.001 Field season 1 1.338 5.1381 0.001 Successional age 9 4 1.1376 1.3992 0.002 Successional age 9 4 1.512 1.4521 0.001 Field season Field season Residuals 115 23.3731 Residuals 117 30.457

Communities were surveyed over the first 14 yr of succession and two field Communities were surveyed over the first 14 yr of succession and two field seasons (2016 and 2017) in a large-scale old-field succession field seasons (2016 and 2017) in a large-scale old-field succession field experiment and characterised based on 16S rRNA gene sequences, experiment and characterised based on internal transcribed spacer (ITS) rarefied to 1000 reads. gene sequences, rarefied to 1000 reads.

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 10 Research Phytologist

A to not only alter the fungal communities of another aster, 6 Jacobaea vulgaris, but also created soil legacy effects that reduced 0.6 the resistance of the plant to a lepidopteran herbivore (Kostenko et al., 2012). However, herbivore induction might not only affect 2 15 the soil microbiome in the ﬁeld through changes in plant AB metabolism, but may also impact the rate of decomposition of 0.4 S. altissima leaves (Burghardt et al., 2018), thus presenting a mechanism through which herbivory in the current and previous B years of succession potentially alter microbial activity and soil nutrient cycling in close proximity to S. altissima plants. 0.2

Probability of eating leaf Soil microbial communities from different stages can affect herbivore preference

0.0 While many changes in S. altissima populations and environmen- 2 6 15 tal factors, such as soil nutrients, may be contributing to both the Inoculant successional age (yr) shifts in the rhizosphere community and herbivore resistance that Fig. 5 Herbivore feeding preference for Solidago altissima plants grown in we observed in the field experiment, the findings from the micro- soils experimentally inoculated with microbiomes from different stages of biome transplant experiment – in which these factors were stan- succession. Preference of adult Trirhabda virgata in a three-way choice dardised – indicated that these microbial shifts alone have the test between discs of leaf tissue from glasshouse-grown plants inoculated ability to modify plant–herbivore resistance. Our finding that late with soil microbiomes collected from old fields in the 2nd,6th or 15th year of succession. Values shown are the estimated marginal mean (SE) succession soil microbiomes conferred greater T. virgata resis- probabilities of eating discs from each inoculation treatment; treatments tance to S. altissima (Fig. 5), paralleling the preference and low not connected by the same letter indicate significant differences based on levels of Trirhabda spp. feeding in the field (Fig. 2), supported whether any leaf tissue was eaten from the different inoculation our hypothesis that microbial shifts in the soil play a major role treatments during the 1 h feeding trial, based on a binary logistic model in increasing the herbivore resistance of plants over succession. (overall effect of inoculant: v2 = 16.83, P=0.0002). Both the mechanism and importance of these soil microbial shifts in mediating the increase in herbivore resistance over succession succession. Plant age has been observed to explain variation in observed in the field relative to other factors, such as changes in bacterial root colonisation patterns in Boechera stricta, particularly S. altissima populations and soil nutrients, warrant further study. in the endosphere (Wagner et al., 2016). Likewise, plant develop- The results of our feeding preference assays with field-grown mental stage was also a major factor determining fungal colonisa- (Fig. 2c) and experimentally inoculated plants (Fig. 5) differed tion in the sorghum rhizosphere (Gao et al., 2019). Thus, the with regard to the relative attractiveness of the mid-succession differential growth of S. altissima over succession (Fig. S1), as well stage/treated plants, suggesting that these other factors are likely as their age, may contribute to the differences in rhizosphere to have important effects on resistance in the field, although it is communities we observed. also possible that the preferences of the larval and adult beetles Furthermore, the level of herbivory on plants may affect their differ (Messina, 1982b). microbial interactions. Herbivore damage varied drastically over While we found that S. altissima plants inoculated with late succession, with some plants in mid succession suffering damage succession soil microbiomes were less attractive to T. virgata than on every leaf, while some of their earlier succession counterparts their counterparts inoculated with earlier succession micro- were entirely damage free. Thus, our data are likely to have cap- biomes, the mechanism of this microbe-mediated resistance tured a wide range in the induction of plant defences and the remains to be elucidated. While we did not find an effect of the concomitant metabolic shifts associated with the herbivore microbiome treatment on the overall profiles of the leaf and defence response (Kessler & Baldwin, 2002). Solidago altissima is VOC secondary metabolites that we measured (Tables S22–S24), known to induce the production of secondary metabolites in the concentrations of two leaf diterpene compounds varied by their leaves, including protease inhibitors, phenolics and diterpe- inoculant (Fig. S6), indicating subtle but impacting changes in nes, as well as VOCs, upon feeding by T. virgata and other plant secondary metabolism. Many microorganisms can synthe- insects (Bode et al., 2013; Uesugi et al., 2013; Morrell & Kessler, sise terpenes (Yamada et al., 2015), so it is possible that these 2016). How these herbivore-induced metabolic shifts affect the metabolites originated directly from the plants’ microbiomes. plants’ interactions with their associated microbes remains rela- Microbes can alter plant secondary metabolite production both tively unknown, but several studies have indicated that artificial directly through metabolising plant compounds or synthesising induction via exogenous application of methyl jasmonate may precursors of plant metabolism (Ludwig-Muller,€ 2015), as well alter the plant microbiome (Carvalhais et al., 2013; Liu et al., as indirectly by inducing or priming plant defence responses 2017), suggesting that this differential herbivory may explain (Pieterse et al., 2014). some of the differences in the microbiomes that we observed over It is also important to note that we only examined one compo- succession. Moreover, aboveground herbivory has been observed nent of resistance in this experiment, feeding choice. We did not

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 11 assess the performance of the herbivores, for example growth Sustainable Future, Andrew W. Mellon student grants from Cor- rates or fitness, on plants inoculated with different microbial nell, a Schmittau-Novak small grant from the School of Integra- communities, or examine whether inoculation affected other eco- tive Plant Science at Cornell, and a grant from Sigma Xi to logically important behaviours such as oviposition preference. MMH, as well as grants from NIFA Multistate NE-1501 and Recent work has also found associations between rhizosphere New Phytologist to AK. MMH was supported in part by a Hor- community composition and rates of aphid parasitism (Blubaugh ton-Hallowell Graduate Fellowship from Wellesley College and a et al., 2018), indicating that examining the effects of microbial Sellew Family Fellowship from Cornell. shifts at the next trophic level may be important for understand- – ing plant herbivore dynamics. Author contributions MMH, JK-K and AK designed the research, interpreted the data Conclusion and contributed to the final manuscript; MMH performed the Overall, the results of our study suggested that successional pro- research, collected and analysed the data, and drafted the cesses belowground have the potential to affect aboveground manuscript. plant–herbivore interactions. Specifically, changes in the soil microbiome alone have the potential to mediate an increase in plants’ resistance to insect herbivores that was observed in late ORCID succession. While we focused only on a single plant species here, Mia M. Howard https://orcid.org/0000-0003-4940-0366 Solidago spp. are ecologically important plants that dominate in a Jenny Kao-Kniffin https://orcid.org/0000-0002-9469-8088 relatively wide range of successional stages and may even reduce Andre Kessler https://orcid.org/0000-0002-1797-7518 rates of succession (Wright & Fridley, 2010). It is not known if this pattern will hold more broadly across species of plants and in different successional environments and one recent study has indicated that plant communities respond differently to soil con- References ditioning in terms of their overall herbivore resistance (Heinen Agrawal AA. 2001. Phenotypic plasticity in the interactions and evolution of et al., 2018). In addition to assessing the importance of the soil species. Science 294: 321–326. microbiota in mediating plants’ interactions with herbivores in Badri DV, Zolla G, Bakker MG, Manter DK, Vivanco JM. 2013. Potential impact of soil microbiomes on the leaf metabolome and on herbivore feeding dynamic environments like these, understanding how broadly behavior. New Phytologist 198: 264–273. this microbe-mediated herbivore resistance phenotype affects Baldrian P. 2008. Enzymes of saprotrophic basidiomycetes. In: Boddy L, other species of plants is of interest from both a fundamental eco- Frankland JC, van West P, eds. Ecology of saprotrophic basidiomycetes. New logical and an applied agricultural perspective. Although the York, NY, USA: Academic Press, 19–41. specific mechanisms that link the microbial community to plant Barber N, Chantos-Davidson K, Peralta RA, Sherwood JP, Swingley WD. 2017. Soil microbial community composition in tallgrass prairie restorations resistance to herbivores and pathogens are multiple and still not converge with remnants across a 27-year chronosequence. Environmental well understood, the general plant population-wide effects Microbiology 19: 3118–3131. observed in this study pose some interesting questions. For exam- Bates D, M€achler M, Bolker BM, Walker SC. 2015. Fitting linear mixed-effects ple, can an environmentally acquired resistance, such as we have models using lme4. Journal of Statistical Software 67:1–48. seen here through the microbiota, affect the strength and direc- Bazzaz FA. 1968. Succession on abandoned fields in the Shawnee Hills, Southern Illinois. Ecology 49: 924–936. tion of natural selection on plant defensive traits (Agrawal, Bazzaz FA. 1979. The physiological ecology of plant succession. Annual Review of 2001)? Or, from a more applied perspective, can functional crop Ecology and Systematics 10: 351–371. rotation or intercropping facilitate crop plant resistance to herbi- Blubaugh CK, Carpenter-Boggs L, Reganold JP, Schaeffer RN, Snyder WE. vores and pathogens through a targeted alteration of the soil 2018. Bacteria and competing herbivores weaken top–down and bottom–up – microbial community (Mutyambai et al., 2019)? Answering aphid suppression. Frontiers in Plant Science 9:1 10. Bode RF, Halitschke R, Kessler A. 2013. Herbivore damage-induced questions like these requires a broader taxonomic survey of plants production and specific anti-digestive function of serine and cysteine protease being affected by microbial communities and a deeper under- inhibitors in tall goldenrod, Solidago altissima L. (Asteraceae ). Planta 237: standing of the mechanisms through which the microbial com- 1287–1296. munities affect plant metabolism and vice versa. Burghardt KT, Bradford MA, Schmitz OJ. 2018. Acceleration or deceleration of litter decomposition by herbivory depends on nutrient availability through intraspecific differences in induced plant resistance traits. Journal of Ecology Acknowledgements 106: 2380–2394. Cain ML. 1990. Models of clonal growth in Solidago altissima. Journal of Ecology We thank Lynn Johnson of the Cornell Statistical Consulting 78:27–46. Unit for statistical advice; Aino Kalske, Alex Chauta, and Alex Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello ~ et al Stotter for help with data collection; and Maria Gannett and EK, Fierer N, Pena AG, Goodrich JK, Gordon JI . 2010. QIIME allows analysis of high-throughput community sequencing data. Nature Methods 7: Liang Cheng for help with microbial DNA sequence processing; 335–336. and four anonymous reviewers for their constructive comments Carvalhais LC, Dennis PG, Badri DV, Tyson GW, Vivanco JM, Schenk PM. on the manuscript. This research was funded in part by a Sustain- 2013. Activation of the jasmonic acid plant defence pathway alters the able Biodiversity Fund grant from the Atkinson Center for a composition of rhizosphere bacterial communities. PLoS ONE 8:1–5.

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 12 Research Phytologist

Cates RG, Orians GH. 1975. Successional status and the palatability of plants to jasmonic acid signalling on the wheat microbiome differ between body sites. generalized herbivores. Ecology 56: 410–418. Scientific Reports 7: 41766. Cregger MA, Veach AM, Yang ZK, Crouch MJ, Vilgalys R, Tuskan GA, Schadt Ludwig-Muller€ J. 2015. Plants and endophytes: equal partners in secondary CW. 2018. The Populus holobiont: dissecting the effects of plant niches and metabolite production? Biotechnology Letters 37: 1325–1334. genotype on the microbiome. Microbiome 6: 31. Lundberg DS, Lebeis SL, Paredes SH, Yourstone S, Gehring J, Malfatti S, Gaiero JR, McCall CA, Thompson KA, Day NJ, Best AS, Dunfield KE. 2013. Tremblay J, Engelbrektson A, Kunin V, del Rio TG et al. 2013. Defining the Inside the root microbiome: bacterial root endophytes and plant growth core Arabidopsis thaliana root microbiome. Nature 488:86–90. promotion. American Journal of Botany 100: 1738–1750. Maddox GD, Cook RE, Wimberger PH, Gardescu S. 1989. Clone structure in Gao C, Montoya L, Xu L, Madera M, Hollingsworth J, Purdom E, Hutmacher four Solidago altissima (Asteraceae) populations: rhizome connections within RB, Dahlberg JA, Lemaux PG, Taylor JW. 2019. Strong succession in genotypes. American Journal of Botany 76: 318–326. arbuscular mycorrhizal fungal communities. ISME Journal 13: 214–226. Maddox GD, Root RB. 1987. Resistance to 16 diverse species of herbivorous Grime JP. 2001. Plant strategies, vegetation processes and ecosystem properties. insects within a population of goldenrod, Solidago altissima: genetic variation Chichester, UK: John Wiley & Sons Ltd. and heritability. Oecologia 72:8–14. Hakes AS, Cronin JT. 2012. Successional changes in plant resistance and McKnight DT, Huerlimann R, Bower DS, Schwarzkopf L, Alford RA, Zenger tolerance to herbivory. Ecology 93: 1059–1070. KR. 2019. Methods for normalizing microbiome data: an ecological Hannula SE, Morri€en E, de Hollander M, van der Putten WH, van Veen JA, de perspective. Methods in Ecology and Evolution 10: 389–400. Boer W. 2017. Shifts in rhizosphere fungal community during secondary Meiners SJ, Phipps KK, Pendergast TH, Canam T, Carson WP. 2017. Soil succession following abandonment from agriculture. ISME Journal 11: 2294– microbial communities alter leaf chemistry and influence allelopathic potential 2304. among coexisting plant species. Oecologia 184: 1155–1165. Hartnett DC, Bazzaz F. 1985. The genet and ramet population dynamics of Messina FJ. 1982a. Comparative biology of the goldenrod leaf beetles, Trirhabda Solidago canadensis in an abandoned field. Journal of Ecology 73: 407–413. virgata Leconte and T. borealis Blake (Coleoptera: Chrysomelidae). Coleopterists Heinen R, van der Sluijs M, Biere A, Harvey JA, Bezemer TM. 2018. Bulletin 36: 255–269. Plant community composition but not plant traits determine the outcome Messina FJ. 1982b. Food plant choices of two goldenrod beetles: relation to plant of soil legacy effects on plants and insects. Journal of Ecology 106: 1217– quality. Oecologia 55: 342–354. 1229. Morrell K, Kessler A. 2016. Plant communication in a widespread goldenrod: Herzig AL. 1995. Effects of population density on long-distance dispersal in the keeping herbivores on the move. Functional Ecology 31: 1049–1061. goldenrod beetle trirhabda virgata. Ecology 76: 2044–2054. Morri€en E, Hannula SE, Snoek LB, Helmsing NR, Zweers H, De Hollander M, Hol WHG, de Boer W, Termorshuizen AJ, Meyer KM, Schneider JHM, van Soto RL, Bouffaud ML, Buee M, Dimmers W et al. 2017. Soil networks Dam NM, van Veen JA, van der Putten WH. 2010. Reduction of rare soil become more connected and take up more carbon as nature restoration microbes modifies plant-herbivore interactions. Ecology Letters 13: 292–301. progresses. Nature Communications 8: 14349. Howard MM, Bell TH, Kao-Kniffin J. 2017. Soil microbiome transfer method Mutyambai DM, Bass E, Luttermoser T, Poveda K, Midega CAO, Khan ZR, affects microbiome composition, including dominant microorganisms, in a Kessler A. 2019. More than ‘push’ and ‘pull’? Plant-soil feedbacks of maize novel environment. FEMS Microbiology Letters 364:1–8. companion cropping increase chemical plant defenses against herbivores. Howard MM, Kalske A, Kessler A. 2018. Eco-evolutionary processes affecting Frontiers in Ecology and Evolution 7: 217. plant-herbivore interactions during early community succession. Oecologia 187: Oksanen J, Guillaume Blanchet F, Friendly M, Kindt R, Legendre P, McGlinn 547–559. D, Minchin PR, O’Hara RB, Simpson GL, Solymos P et al. 2018. vegan: Howard MM, Kaminsky LM, Kessler A, Bell TH. 2019. Merging microbial and community ecology package. R package v.2.5-2 [WWW document] URL https:// plant profiling to understand the impact of human-generated extreme CRAN.R-project.org/package=vegan. environments in natural and agricultural systems. In: Yergeau E, ed. Advanced Panke-Buisse K, Poole AC, Goodrich JK, Ley RE, Kao-Kniffin J. 2015. techniques for studying microorganisms in extreme environments. Berlin, Selection on soil microbiomes reveals reproducible impacts on plant function. Germany: de Gruyter, 57–91. ISME Journal 9: 980–989. Jastrow JD, Miller RM. 1993. Neighbor influences on root morphology and Paulson JN, Colin Stine O, Bravo HC, Pop M. 2013. Differential abundance mycorrhizal fungus colonization in tallgrass prairie plants. Ecology 74: 561– analysis for microbial marker-gene surveys. Nature Methods 10: 1200–1202. 569. Peiffer JA, Spor A, Koren O, Jin Z, Tringe SG, Dangl JL, Buckler ES, Ley RE. Kardol P, Martijn Bezemer T, Van Der Putten WH. 2006. Temporal variation 2013. Diversity and heritability of the maize rhizosphere microbiome under in plant-soil feedback controls succession. Ecology Letters 9: 1080–1088. field conditions. Proceedings of the National Academy of Sciences, USA 110: Kessler A, Baldwin IT. 2002. Plant responses to insect herbivory: the emerging 6548–6553. molecular analysis. Annual Review of Plant Biology 53: 299–328. Pieterse MJ, Zamioudis C, Berendsen RL, Weller DM, Van Wees SCM, Bakker Kos M, Tuijl MAB, De Roo J, Mulder PPJ, Bezemer TM. 2015. Species-specific PAHM. 2014. Induced systemic resistance by beneficial microbes. Annual plant–soil feedback effects on above-ground plant–insect interactions. Journal Review of Phytopathology 52: 347–375. of Ecology 103: 904–914. Pineda A, Zheng SJ, van Loon JJA, Pieterse CMJ, Dicke M. 2010. Helping Kostenko O, van de Voorde TFJ, Mulder PPJ, van der Putten WH, Martijn plants to deal with insects: the role of beneficial soil-borne microbes. Trends in Bezemer T. 2012. Legacy effects of aboveground-belowground interactions. Plant Science 15: 507–514. Ecology Letters 15: 813–821. R Core Team. 2018. R: a language and environment for statistical computing. Kuznetsova A, Brockhoff PB, Christensen RHB. 2017. lmerTest package: tests Vienna, Austria: R Foundation for Statistical Computing. [WWW document] in linear mixed effects models. Journal of Statistical Software 82:1–25. URL https://www.R-project.org/. Lau JA, Lennon JT. 2011. Evolutionary ecology of plant–microbe interactions: Reader PM, Southwood TRE. 1981. The relationship between palability to soil microbial structure alters selection on plant traits. New Phytologist 192: invertebrates and the successional status of a plant. Oecologia 41: 271–275. 215–224. Ristok C, Poeschl Y, Dudenhoffer€ J, Ebeling A, Eisenhauer N, Vergara F, Wagg Lau JA, Lennon JT. 2012. Rapid responses of soil microorganisms improve plant C, van Dam NM, Weinhold A. 2019. Plant species richness elicits changes in fitness in novel environments. Proceedings of the National Academy of Sciences, the metabolome of grassland species via soil biotic legacy. Journal of Ecology USA 109: 14058–14062. 107: 2240–2254. Lenth R. 2019. emmeans: Estimated marginal means, aka least-squares means.R Rognes T, Flouri T, Nichols B, Quince C, Mahe F. 2016. VSEARCH: a package v.1.3.4. [WWW document] URL https://CRAN.R-project.org/packa versatile open source tool for metagenomics. PeerJ 4: e2584. ge=emmeans. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB, Liu H, Carvalhais LC, Schenk PM, Dennis PG, Berendsen RL, Pieterse CM, Lesniewski RA, Oakley BB, Parks DH, Robinson CJ et al. 2009. Introducing Bakker PA, Glazebrook J, Carvalhais LC, MicheLet al. 2017. Effects of mothur: open-source, platform-independent, community-supported software

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust New Phytologist Research 13

for describing and comparing microbial communities. Applied and Fig. S6 Effects of successional inoculants on concentrations of – Environmental Microbiology 75: 7537 7541. diterpene compounds #25 and #28 in S. altissima leaves. Schneider CA, Rasband WS, Eliceiri KW. 2012. NIH image to ImageJ: 25 years of image analysis. Nature Methods 9: 671–675. Stockinger H, Kruger€ M, Schußler€ A. 2010. DNA barcoding of arbuscular Methods S1 Characterisation of bacterial and fungal communi- mycorrhizal fungi. New Phytologist 187: 164. ties via amplicon sequencing of the 16S rRNA and ITS gene Uesugi A, Kessler A. 2013. Herbivore exclusion drives the evolution of plant regions. competitiveness via increased allelopathy. New Phytologist 198: 916–924. Uesugi A, Kessler A. 2016. Herbivore release drives parallel patterns of Methods S2 Volatile metabolite profiles of plants in the microevolutionary divergence in invasive plant phenotypes. Journal of Ecology 104: 876–886. biome transplant experiment. Uesugi A, Poelman EH, Kessler A. 2013. A test of genotypic variation in specificity of herbivore-induced responses in Solidago altissima L. (Asteraceae). Table S1 Leaf secondary metabolites examined via HPLC. Oecologia 173: 1387–1396. United States Department of Agriculture. 2014. Kellogg soil survey laboratory Table S2 ANOVA table for effects of successional age and season methods manual. Soil survey investigations report no. 42, v.5.0. R. Burt and Soil Survey Staff (ed.) IUSA Department of Agriculture, Natural Resources on Trirhabda sp. abundance on S. altissima plants in the field. Conservation Service. [WWW document] URL https://www.nrcs.usda.gov/ Internet/FSE_DOCUMENTS/stelprdb1253871.pdf. Table S3 ANOVA table for effects of successional age and season Vance ED, Brookes PC, Jenkinson DS. 1987. An extraction method for on herbivore damage on S. altissima plants in the field. measuring soil microbial biomass C. Soil Biology and Biochemistry 19: 703–707. Vessey JK. 2003. Plant growth promoting rhizobacteria as biofertilizers. Plant and Soil 255: 571–586. Table S4 Odds ratios for the likelihood of Trirhabda sp. con- Wagner MR, Lundberg DS, Coleman-Derr D, Tringe SG, Dangl JL, Mitchell- suming S. altissima plants from different stages of succession in a Olds T. 2014. Natural soil microbes alter flowering phenology and the feeding choice test. intensity of selection on flowering time in a wild Arabidopsis relative. Ecology – Letters 17: 717 726. Table S5 Confidence intervals (95%) of odds ratios for the likeli- Wagner MR, Lundberg DS, del Rio TG, Tringe SG, Dangl JL, Mitchell-Olds T. 2016. Host genotype and age shape the leaf and root microbiomes of a wild hood of Trirhabda sp. consuming S. altissima plants from differ- perennial plant. Nature Communications 7:1–15. ent stages of succession in a feeding choice test. Wright JP, Fridley JD. 2010. Biogeographic synthesis of secondary succession rates in eastern North America. Journal of Biogeography 37: 1584–1596. Table S6 Pairwise contrasts between Trirhabda sp. beetle larvae Yamada Y, Kuzuyama T, Komatsu M, Shin-ya K, Omura S, Cane DE. 2015. performance of on Solidago altissima plants from communities in Terpene synthases are widely distributed in bacteria. Proceedings of the National Academy of Sciences, USA 112: 857–862. different stages. Zhu F, Heinen R, van der Sluijs M, Raaijmakers C, Biere A, Bezemer TM. 2018. Species-specific plant–soil feedbacks alter herbivore-induced gene expression and Table S7 ANOVA table for effects of successional age and season defense chemistry in Plantago lanceolata. Oecologia 188:801–811. on S. altissima plant heights in the field.

Supporting Information Table S8 ANOVA table for effects of successional age and season on the total number of leaves of S. altissima in the ﬁeld. Additional Supporting Information may be found online in the Supporting Information section at the end of the article. Table S9 Soil properties over succession.

Fig. S1 Comparison of sizes of S. altissima plants over succession Table S10 ANOVA table for effects of successional on soil prop- in the ﬁeld. erties.

Fig. S2 Compositional shifts in S. altissima rhizosphere microbial Table S11 ANOVA table for effects of successional age and sea- communities over succession in the ﬁeld (normalised via propor- son and the relative abundance of basidiomycetes in S. altissima tions). rhizospheres in the ﬁeld.

Fig. S3 Compositional shifts in S. altissima rhizosphere microbial Table S12 PERMANOVA table for effects of successional age communities over succession in the field (normalised via cumula- and field season on S. altissima rhizosphere bacterial composition tive sums scaling). in the field (normalised via proportions).

Fig. S4 Microbial communities of successional inoculants and Table S13 PERMANOVA table for effects of successional age resulting S. altissima rhizospheres in the microbiome transplant and ﬁeld season on S. altissima rhizosphere fungal composition in experiment. the ﬁeld (normalised via proportions).

Fig. S5 Phylum-level composition of successional inoculants and Table S14 PERMANOVA table for effects of successional age resulting S. altissima rhizospheres in the microbiome transplant and ﬁeld season on S. altissima rhizosphere bacterial composition experiment. in the ﬁeld (normalised via cumulative sums scaling).

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com New 14 Research Phytologist

Table S15 PERMANOVA table for effects of successional age Table 21 Confidence intervals (95%) of odds ratios for the likeli- and field season on S. altissima rhizosphere fungal composition in hood of Trirhabda virgata consuming S. altissima plants inocu- the field (normalised via cumulative sums scaling). lated with microbiomes from different stages of succession in a feeding choice test. Table S16 PERMANOVA table for effects of successional inoculants and plant genotype on S. altissima rhizosphere bacterial Table S22 PERMANOVA table for effects of successional inocu- composition in the microbiome transplant experiment. lants and plant genotype on the overall leaf secondary metabolites of S. altissima plants. Table S17 PERMANOVA table for effects of successional inoculants and plant genotype on S. altissima rhizosphere fungal com- Table S23 Results of ANOVAs assessing the effects of succes- position in the microbiome transplant experiment. sional inoculants and plant genotype on the concentrations of individual secondary metabolites in S. altissima leaves. Table S18 ANOVA table for effects of successional inoculants and plant genotype on early season height-based growth rates of Table S24 PERMANOVA table for effects of successional inocu- S. altissima plants in the microbiome transplant experiment. lant and plant genotype on the overall VOC secondary metabolites of S. altissima plants. Table S19 ANOVA table for effects of successional inoculants and plant genotype on early season leaf number-based growth rates of S. altissima plants in the microbiome transplant experi- Please note: Wiley Blackwell are not responsible for the content ment. or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be Table S20 Odds ratios for the likelihood of Trirhabda virgata directed to the New Phytologist Central Office. consuming S. altissima plants inoculated with microbiomes from different stages of succession in a feeding choice test.

New Phytologist is an electronic (online-only) journal owned by the New Phytologist Trust, a not-for-proﬁt organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews and Tansley insights.

Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication ‘as ready’ via Early View – our average time to decision is <26 days. There are no page or colour charges and a PDF version will be provided for each article.

The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for table of contents email alerts.

If you have any questions, do get in touch with Central Office (np-centraloffi[email protected]) or, if it is more convenient, our USA Office (np-usaoffi[email protected])

For submission instructions, subscription and all the latest information visit www.newphytologist.com

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust