We established 160 experimental ecosystems (mesocosms), manipulated interactions between plants, soil biota and invertebrate herbivores in a fully factorial design. Each mesocosm was grown in a 125 L pot (575 mm diameter, Fig. 1B), and comprised one of 20 unique, eight-species plant communities varying orthogonally in the proportion of exotic and woody shrub/tree species (0-100% and 0-63%, respectively). These plants were taken from a pool of 20 exotic and 19 native/endemic New Zealand plant species. Soil biota were manipulated using a modified plant-soil feedback approach, where each plant species was grown in monoculture in 10 L pots containing field-collected soil for 9-10 months, allowing the conditioning of typical associated soil biota for each of the plant species. We created ‘home’ soils by taking the conditioned soil from each of the eight representative species in a mesocosm and mixing it together to create a single inoculum. Each ‘home’ soil mixture was also used as an ‘away soil’ in a different mesocosm that did not contain any of the representative plants in that inoculum. These soils were intended to increase the relative biomass in inocula of specialized and preferred interaction partners of the resident (or non-resident) plant species. Invertebrate insect herbivore populations were added into half of the mesocosms with home soils and half with away soils. Thirteen invertebrate herbivore species introduced into the mesocosms successfully established, along with seven self-colonizing species, totaling 20 species in all. All mesocosms were sealed with mesh cages (15% shade factor) designed to retain added herbivores and exclude others from entering. In each mesocosm, we measured distinct ecosystem properties and processes that are relevant to carbon cycling dynamics: above and belowground biomass; total in situ soil respiration; basal respiration; microbial biomass (measured as substrate-induced respiration); decomposition (rate of standardized substrate mass loss); soil organic matter; nitrogen availability; herbivore biomass; arbuscular mycorrhizal and other fungal biomass (using neutral- and phospho-lipid fatty acid [NLFA and PLFA] biomarkers, respectively); and bacterial biomass (PLFA).

Additional details of the methods can be found in the supplementary materials of Waller et al. 2020: science.sciencemag.org/content/368/6494/967/suppl/DC1

Soil conditioning

The soils to be used in the larger mesocosm experiment were initially conditioned in the glasshouse by each of our focal species in order to create unique 1) ‘home’ soil inocula that would contain the suite of specialist and generalist soil biota cultured by the resident species growing in a given community and 2) ‘away’ soil inocula that would contain soil organisms not typically associated with the resident plants in a community. These soils could be practically interpreted as ‘previously-invaded’ vs. ‘uninvaded’. We grew 12-20 replicates of each individual plant species in 10 L pots containing live, field-collected soil mixed in equal parts with pasteurized field-collected soil and pasteurized sand. The soil was collected in late June and early July of 2016 from twelve subalpine, grass and shrub dominated sites located across western Canterbury and southern Marlborough (details in table S3). We chose sites where subsets of our focal plant species (table S2) were present. At each site, we located three sampling areas that were representative of the site, but differed slightly in species composition, so that our final soil sample included rhizosphere soil from as many of our focal species as possible. At each sampling area, we used shovels to excavate chunks of sod measuring approximately 1 x 0.5 x 0.2 m, retaining as much loose soil that had fallen off the sod as possible. Soil and sod were stored outdoors for two weeks in weatherproof bags where temperatures averaged 1/11^o C (low/high). Aboveground vegetation was removed from sod chunks and soil and roots were passed through (2.5 cm²) sieves in mid-July to remove rocks and other debris, but retain any roots. This sampling provided us with approximately 375 L of soil from each site, which was mixed together in equal parts for a total of approximately 3000 L of inoculum. We also collected 3000 L of soil from a field on the Lincoln University campus for pasteurization in mid-July of 2016. Sand was collected from the Rakaia River valley, washed and sieved. The soil collected from the Lincoln site was mixed with the sand in equal parts and pasteurized in 500 L batches on a modified trailer bed fitted with steam pipes below a metal sheet and covered with a large tarpaulin. Each batch was brought up to a temperature of at least 100^oC, held at temperature for 60 minutes and cooled for 24 hours, before being treated a second time. Live soil was mixed with pasteurized soil in a 1:2 ratio.

After seeds had germinated and seedlings had at least one true leaf, they were transplanted into their own 1 L pot containing the soil inoculum. Some seedlings were too small to transplant directly into the large pot at this stage, so they were potted into a 500 mL pot of live inoculum until they were strong enough to go into the larger pot.  Plants were added to pots beginning in July 2016 and grew for approximately 9-10 months.

Mesocosm experimental design

Our mesocosm experiment incorporated a fully factorial design, with 20 plant communities X soil manipulation (home/away) X herbivore manipulation (herbivores added/not added). To begin, we designed 20 unique plant communities (table S1), each containing one individual of each of eight species, varying orthogonally in their proportion of exotic species (0, 25, 50, 75, 100%) and their proportion of woody species (0, 25, 38, 63%). Although we initially included seven nitrogen (N)-fixing exotic species, two species (one woody and one herbaceous) had extremely low survival, so effectively there were only five exotic N-fixers overall. Home soils contained a mixture of conditioned soils from each of the eight species occurring in that community, whereas away soils contained a mixture of conditioned soils from eight species occurring in one of the other 19 communities, but where a focal species did not occur (table S4). Twenty herbivore species established across the +Herbivore mesocosms (n = 80) and mesh cages (see fig. 1A) were secured over each pot to ensure the insects remained in their pots. Seven of these herbivore species colonized from outside of cages (i.e. slugs and aphids) and we controlled these additions in -Herbivore but not +Herbivore mesocosms, where we allowed them to maintain populations. Thus, we consider our herbivore treatment to be ‘herbivores added’ vs. ‘herbivores reduced’.

Mesocosm establishment

The 160 mesocosms were established in a field on the campus of Lincoln University in Lincoln, New Zealand. We established the mesocosm communities in two phases. First, we germinated all of our plant species a second time, using the same method as described above. This time, however, after germination and two true leaves emerged, we planted each of our species into small pots containing their respective mesocosm soil inoculum before adding them to the mesocosms. This was done so the plants could be “hardened off” in treated soil before planting outdoors. To accomplish this, we harvested four pots of each plant species grown in the soil-conditioning phase, combining all soil and roots into a single bag for each individual species. We simultaneously pasteurized another batch of field-collected background soil from our field at Lincoln University to double this inoculum volume. Each seedling for the mesocosm phase was then allocated to a particular community (table S1) and treatment combination (table S4): plants that were to go into home soil communities were planted in individual pots in soil conditioned by themselves, and seedlings to grow in away communities were planted in soil from one randomly chosen species from their predetermined away mixture. These seedlings grew for approximately one month and were moved outside to a shade house during their last week in the small pots. While seedlings were hardening off, we harvested the rest of the plants from the conditioning phase as before and created home and away soil inoculum mixtures that would go into each mesocosm pot. Finally, to complete the planting, we constructed steel pots (575 mm diameter), filled each with a bottom layer of 22 L of crushed gravel, then 88 L of pasteurized soil:sand, and finally 12 L of either home or away soil inoculum, mixed uniformly across the top. We mixed pasteurized soil with sand to improve the drainage in the soil. Seedlings were planted in a ring, equally spaced around the center of the pot in March of 2017.

Mesocosm harvest and sampling

After one year of growth, we harvested all above and belowground plant material from each mesocosm community. Using spades, we carefully excavated each plant, disentangling roots of different species when necessary. Plants were bagged in the field and taken offsite to wash free of soil and other debris, then cut into root and shoot fractions. All shovels used for harvesting were scrubbed and rinsed in bleach for at least 10 minutes between mesocosms. Although the communities grew in pots for a year, there was no evidence that the plants were pot bound at the end of the experiment. The different plant species’ roots were easily separated from one another at harvest, and there was no root coiling around the pots whatsoever. After all plants were removed, soil was homogenized by turning over for several minutes with the shovel, after which time a soil sample was extracted and taken into the laboratory. There, approximately 50 g of soil was passed through a 4 mm sieve, and stored at 4^oC until processed (i.e. substrate induced respiration, basal respiration). Ten grams of soil was freeze dried for PLFA/NLFA analysis, 50 g dried and ground for nutrient analyses, 2.5 g added to 96 well plates from the Mo-Bio Power Soil Extraction Kit, and 10 g frozen at -80^oC for archiving.

In the laboratory, a small subsample of root was taken from each individual plant by taking approximately ten fine-root fragments from random places on the root ball, which were then bundled together, and a 1 cm cross section cut from the bundle using a sterile razor blade. This sample was placed into a 96 well plate from the Mo-Bio Power Soil Extraction Kit and frozen at -80^oC as soon as a plate was full. After this sampling, all plant material was dried at 65^oC and weighed.

Decomposition (percentLoss):

We measured the rate of a standard substrate lost over time using buried tea bags that were placed into each mesocosm at the initiation of the experiment and removed at harvest (approximately 12 months later). We used Harney & Sons Black English Breakfast tea bags. These bags were chosen because the tea bag material is plastic and did not degrade in the soil. Our goal was to use a standard substrate across all mesocosms that was equally novel to each mesocosm to avoid any home field advantage in decomposition. Camellia sinensis, being a member of the Ericales, was sufficiently distinct from any species in the mesocosms that the phylogenetic similarity would be low across all mesocosms. Tea bags were oven dried at 65^oC and weighed immediately before adding to the mesocosms. We added two tea bags to the center of each mesocosm, spaced 150 mm apart. At harvest, bags were shaken free of soil, oven dried again and weighed.

NLFAamf: AMF biomass as measured by 16.1ω5 NLFA biomarkers

Bacterial.biomass: Bacterial biomass as measured by PLFA biomarkers

X18.2w6c: Non-AM fungal biomass as measured by 18:2ω6 PLFA biomarkers

GramNegative: Biomass of gram-negative bacteria, grouped using iso and anteiso saturated branched PLFAs

GramPositive: Biomass of gram-negative bacteria, grouped using monounsaturated and cyclopropyl 17:0 and 19:0 PLFAs

tC: total carbon; AN_kg.ha: available nitrogen; OM: organic matter; CN: carbon:nitrogen 

Total carbon and nitrogen in dried soil material were analyzed using the Dumas type combustion process with an Elementar Vario-Max CN Elemental Analyzer. The sample was combusted at 900°C in an oxygen atmosphere. The combustion process converts any elemental carbon and nitrogen into CO₂, N₂ and NO_x. The NO_x species were subsequently reduced to N₂. These gases were then passed through a thermal conductivity cell to determine CO₂ and N₂ concentrations and the percentages of C and N calculated from the sample weights.

CWSLA and CWSRL: Community-weighted specific leaf area and specific root length

To calculate community-weighted SLA and SRL for each mesocosm, we multiplied each species’ mean trait value by its relative aboveground biomass in the mesocosm and then summed these values across all species in each mesocosm community. 

R10: total respiration:

Total soil respiration was measured in situ from each mesocosm pot using a PPSystems EGM-4 infrared gas analyzer (IRGA). In October 2017, a single PVC collar (50 mm deep, 103 mm diameter) was carefully inserted into the soil in the center of each pot so that 1-2 cm of the ring remained above the surface. Rings were left undisturbed throughout the course of the experiment. Surface CO₂ efflux (R10), soil temperature and moisture were measured during peak growth in December 2017 and again at harvest in April 2018. Pots were well-watered ahead of the measurements to ensure they were all at approximately similar moisture levels. Total soil respiration (R10) was calculated using the equation: R10 = Flux/EXP(308.56*((1/56.02)-((1/(Ts+273.15-227.13))))), where flux=the measured CO₂ efflux in µmol/m²/s and Ts is the soil temperature (^oC).

Basal respiration:  

Basal respiration was measured as the amount of CO₂ evolved from a 130 mL bottle containing 10 g dry weight equivalent of soil at 55% water holding capacity over a 3 h period, at 25°C. Carbon dioxide concentrations were measured by injecting 1 mL gas samples into a LiCor LI-7000 CO₂ analyser 1 and 3 h after capping.

Substrate-induced respiration:

Substrate-induced respiration (a measure of microbial biomass) was measured in the same way as basal respiration, except that 0.1 g of glucose was added to each bottle before capping

Key to headings in data spreadsheet:

• Mesocosm: Unique mesocosm number
• Soil.treatment: Home/Away/Sterile
• Invasion.treatment: Invaded with Cytisus scoparius or uninvaded
• Herbivore.treatment: Herbivores added (HERB) or not (NO_HERB)
• Plant.community: Each mesocosm was in one of 20 unique plant communities
• Soil.community: Denotes the 'away' soil community the mesocosm was grown in
• prop.exotic: Proportion of exotic plants (0, 25, 50, 75 or 100%) planted in the community
• prop.woody: Proportion of woody plants (0, 25, 38 or 63%) planted in the community
• Above, below, total biomass: Total shoot, root or root and shoot biomass produced in the community
• prop.Nfix: Proportion of N-fixing plant species planted in the community
• Herbivore.biomass: Total herbivore biomass (mg) produced in the mesocosm
• percentLoss: percent of standard substrate lost over the course of the experiment
• NLFAamf: AMF biomass as measured by 16.1ω5 NLFA biomarkers
• Bacterial.biomass: Bacterial biomass as measured by PLFA biomarkers
• X18.2w6c: Non-AM fungal biomass as measured by 18:2ω6 PLFA biomarkers
• GramNegative: Biomass of gram-negative bacteria, grouped using iso and anteiso saturated branched PLFAs
• GramPositive: Biomass of gram-negative bacteria, grouped using monounsaturated and cyclopropyl 17:0 and 19:0 PLFAs
• tC: total carbon; AN_kg.ha: available nitrogen; OM: organic matter; CN: carbon:nitrogen 
• CWSLA and CWSRL: Community-weighted specific leaf area and specific root length
• R10: total respiration
• Basal.Respiration: basal respiration
• SIR: Microbial biomass