Soil microplastic pollution can have negative effects on organisms, including plants, but the underlying mechanisms are not fully understood. We tested whether structural or chemical properties of a microplastic cause its effects on plant above- and belowground growth and whether these effects can be influenced by earthworms. We conducted a factorial experiment in a greenhouse with seven common Central European grassland species. Microplastic granules of the synthetic rubber ethylene propylene diene monomer (EPDM), a frequently used infill material of artificial turfs, and cork granules with a comparable size and shape to the EPDM granules were used to test for structural effects of granules in general. To test for chemical effects, EPDM-infused fertilizer was used, which should have contained any leached water-soluble chemical components of EPDM. Two Lumbricus terrestris individuals were added to half of the pots, to test whether these earthworms modify effects of EPDM on plant growth. As response parameters aboveground, belowground, and total biomass, as well as several root traits were measured, and further parameters derived from the measured values. EPDM granules had a clear negative effect on plant growth, but since cork granules had a negative effect of similar magnitude, with an average decrease in biomass of 37 % in presence of granules, this is likely due to the structural properties of granules (i.e., size and shape). For some belowground plant traits, EPDM had a stronger effect than cork, which shows that there must be other factors playing into the effects of EPDM on plant growth. The EPDM-infused fertilizer did not have any significant effect on plant growth by itself, but it had in interaction with other treatments. Earthworms had an overall positive effect on plant growth and mitigated most of the negative effects of EPDM. These results show that EPDM microplastic can have negative effects on plant growth, and that these might be more related to its structural than to its chemical properties.

1. Material and methods
2.1  Study species and pre-cultivation

The study was conducted on seven grassland species (Alopecurus pratensis L., Daucus carota L., Festuca guestfalica Boenn. Ex Rchb., Galium album Mill., Leucanthemum ircutianum Turcz. ex DC., Lolium perenne L., Prunella vulgaris L.) from five families (Table S1). These seven species were chosen because they are very common in central European grasslands, represent different guilds (grasses and forbs), and because a previous study had shown that their growth was reduced by the presence of EPDM microplastics in the soil (van Kleunen et al. 2020). Using multiple species from different guilds increases the generality of the results (van Kleunen et al. 2014), and therefore the results should be more representative for Central European grassland communities in general. Seeds were sown in cultivation pots (Kulturschalen TK1214; Pöppelmann GmbH & Co. KG TEKU ®, Lohne, Germany) filled with peat soil (Einheitserde ® CL P; Einheitserdewerke Patzer Gebr. Patzer GmbH & Co. KG, Sinntal-Altengronau, Germany). As the species were known to differ in the time required for germination, they were sown on different dates (Table S1). The germination took place in a greenhouse with a temperature of 18°C during the night and 20°C (±2°C) during the day. Additional lighting was used to extend the daily light period to 14 hours.

2.2  Experimental set-up

We used three main treatments according to a full factorial design: granules treatment (none, cork, ethylene propylene diene monomer [EPDM]), EPDM-infusion (yes, no) and earthworm presence (yes, no). We used four replicates of every treatment combination for each of the seven species, resulting in a total of 336 pots. The pots were randomly allocated to positions on three benches in the greenhouse.

For the experiment, we used 3-L pots (SMV pots; Soparco, Chaingy, France). In each pot, we put a circular piece of chiffon fabric (Chiffon Stoff weiß; Modern Textiles GmbH, Berlin, Germany) at the bottom to avoid the substrate leaking out from the bottom of the pots. We filled all pots with 1.2 L of a 1:1 mix of sand (Quarzsand; Emil Steidle GmbH & Co.KG, Sigmaringen, Germany) and vermiculite (Isola Vermiculite GmbH, Sprockhövel, Germany). The sand-vermiculite mixture had a grain size of 0.3 to 0.8 mm. As substrate for the earthworms, we added a layer of 600 ml of organic substrate to each pot on top of the sand-vermiculite mix. The organic substrate was a 1:1 mix of soil (Oberbodengemisch; Corthum Nordschwarzwald GmbH, Marxzell, Germany) and special earthworm cultivation soil on basis of peat-mix, containing food, proteins, vitamins and minerals for the earthworms (Regenwurmzuchterde; Superwurm e.K., Düren, Germany). For the treatments with cork and EPDM, this organic top layer consisted of 30 ml of granules mixed into 570 ml of the organic substrate. The black EPDM granules (Gummi Appel GmbH + Co.KG, Kahl am Main, Germany; Figure S1), which are sold for use as infill material for artificial turfs (e.g., sports fields) had a size of 1.0 mm to 2.5 mm (approximate size distribution according to the producer: 1% 1.0 – 1.2 mm; 3% 1.2 – 1.4 mm; 15% 1.4 – 1.6 mm; 33% 1.6 – 1.8 mm; 48% 1.8 – 2.5 mm). The 5 % volume concentration of EPDM in the organic substrate was chosen, because it was shown to have negative effects on plant growth without resulting in high plant mortality (van Kleunen et al. 2020). Since this concentration is equal to a concentration of ca. 53.4 g microplastic per kg substrate, it is comparable to the 67.5 g microplastics per kg found in soils of an industrial area (Fuller & Gautam 2016). The cork granules were used to determine whether the effects of EPDM on plant growth result from the physical presence of granules, irrespective of the material. The cork granules (cork-shop.com, Friedbert Bleile Handel und Dienstleistungen, Umkrich, Germany; Figure S1) consisted of natural cork without any additives and ranged from 0.5 mm to 2.0 mm in size and are also sold for use as infill material for artificial turfs.   

On the 6th of May 2019, we transplanted the seedlings into the pots. Within the first week, 14 seedlings that had died were replaced. The temperature in the greenhouse was set to 16°C/18°C (night/day) with ventilation starting at temperatures above 20°C/22°C (night/day), and no artificial lighting was used. The plants were watered regularly until saturation of the substrate with approximately equal amounts of tap water using a hose throughout the whole experiment to keep the soil permanently moist and avoid drought stress.

To test the effect of the presence of earthworms, we added two Lumbricus terrestris individuals to half of the pots (Wurmwelten.de, Stadtoldendorf, Germany). This number is representative of the number of earthworms typically found in Central European grasslands (Jänsch et al. 2013). As the first batch of earthworms we had ordered did not include enough individuals to have two worms per pot, we first added only one worm per pot four days after transplanting the seedlings. Three days later, after we had received a second batch of earthworms, we added the second worm to each pot. Before adding the earthworms to the pots, we washed and weighed each one of them separately. The total earthworm mass per earthworm-treatment pot was between 6.13 g and 9.77 g, with a mean ± SE of 7.96 ± 0.05 g. To reduce earthworm escape from pots, we put a sleeve of white fleece (Thermovlies M85; Hermann Meyer KG, Rellingen, Germany) with an approximate height of 6 to 9 cm around each pot (i.e., also around the pots without earthworms, Figure S2).

To test whether effects of EPDM on plant growth could result from  water-soluble chemical components of the EPDM granules (including any additives added during production or residual reactants), we added EPDM-infused fertilizer to half of the pots. We soaked 5 % volume concentration of EPDM in 1 ‰ Universol ® Blue fertilizer solution (ICL Deutschland Vertriebs GmbH, Nordhorn, Germany; main macronutrients contained in the fertilizer: 18 % nitrogen, 11 % phosphate, 18 % potassium oxide) for one week. Afterwards, it was stored at 15°C in a dark room. As control, we used 1 ‰ Universol ® Blue fertilizer solution, stored in the same room. Before applying 50 ml of EPDM-infused fertilizer to each plant in the infusion treatment, the EPDM granules were removed from the fertilizer solution with a 200 μm sieve. We applied the infusion treatment once a week, starting from the second week, resulting in five applications in total. No significant difference in pH was observed between the infused fertilizer and control fertilizer, as measured before each application with a pH-meter.

2.3  Measurements
2.3.1 Initial measurements

Two and three days after transplanting, we took initial size measurements of the plants. For this, we measured on each plant the length and width of the biggest cotyledon, the length and width of the biggest true leaf (if present) and counted the numbers of cotyledons and true leaves. These measurements were used to calculate a proxy of total initial leaf area of each plant as number of cotyledons × length of cotyledon × width of cotyledon + number of true leaves × length of true leaf × width of true leaf.

2.3.2 Harvest

The plants were harvested from the 17th to the 21st of June 2019, after a growth period of at least 6 weeks. As two plants had died, the total number of harvested plants was 334. The aboveground and belowground parts were harvested separately. The aboveground parts were cut off, dried in a drying oven at 70°C for at least three days and weighed. The belowground parts were washed free from substrate and stored at 6°C in 60 ml plastic cups filled with demineralized water. Since the top third of the substrate was organic soil, it was not possible to get the roots completely clean. To measure root morphological traits, we scanned a subsample of distal fine roots, which are most likely the absorptive roots, from the lower 1/3 of the root system of each plant, using an EPSON ® Scanner (Expression 12000 XL & 11000 XL). This was done within four days after harvesting the roots. We analysed the scanned root subsamples with WinRHIZOTM (Pro Version 2017a; Regent Instruments Canada Inc.), which calculated the root length, root volume and average diameter for each root subsample. After scanning, the subsample and remaining root system were dried separately in a drying oven at 70°C for at least two days and weighed. From the WinRHIZOTM data and the root biomass, we calculated the total root length, specific root length (SRL; root length/root dry weight) and root tissue density (RTD: root dry weight/root volume).  Seventy-four of the initial 336 earthworms were missing (9 pots with no earthworms, 56 pots with one earthworm instead of two), indicating that they had died or escaped. At the time of harvest the total earthworm mass per earthworm-treatment pot ranged from 0 g, when both earthworms had died or escaped, to 10.66 g, with a mean ± SE of 5.84 ± 0.18 g (7.36 ± 0.11 g when considering only the pots that still had two earthworms).

2.4 Statistical analysis

All analyses were conducted with R 4.2.2 (R Core Team 2022). To test the effects of the different treatments and their interactions on aboveground biomass, belowground biomass, total biomass, root weight ratio (RWR), total root length, SRL, RTD and average root diameter, we fitted a linear mixed effects model for each of these response variables with the lme function of the R package nlme (Pinheiro et al. 2018). As fixed factors, we used the three treatments and all possible two- and three-way interactions. To account for differences in initial size and in the number of days until harvest, we added total initial leaf area and growth duration (i.e., number of days between initial size measurements and harvest) as scaled and centred covariates. As we did not have specific expectations for differences among the seven study species, and we considered them merely as representatives of common grassland species in Central Europe, we included species identity as well as its interactions with the three treatments as random terms. From the analyses, we excluded nine pots in which both earthworms had disappeared, two pots in which the plants had died, and one pot of the non-earthworm treatment in which we found an earthworm. As preliminary analyses did not reveal any difference between pots with one or two surviving earthworms, we did not include the number of alive earthworms as covariate.

Since none of the models met the assumptions of normality, the response variables were transformed (square-root transformation for biomass parameters, RWR, total root length and SRL; log-transformation for RTD and average root diameter). Since most of the models with did not meet the assumption of homoscedasticity due to differences in variance among the species, we modelled different variances per species with the varIdent function of the R package nlme (Pinheiro et al. 2018). The latter, however, was not required for SRL. To estimate the standard deviations of the random terms, the restricted maximum likelihood method (REML) of the linear mixed effects model was used. As the granules treatment had three levels, we used a priori orthogonal contrasts to compare the presence of granules (EPDM + cork) with the absence of granules (C_Present), and the presence of EPDM granules with the presence of cork granules (C_EPDM). To test for significance of each fixed term, the maximum likelihood method (ML) of the linear mixed effects model was used. We then used the anova function to compare the log-likelihood of models including and excluding the factor of interest. The test statistic used to compare the models was the log-likelihood ratio, which is approximately chi-squared distributed.

In addition to the orthogonal contrasts among the granule treatments, we also added to the figures letters to indicate the significant differences for all pairwise combinations of treatments based on the Sidak multiple comparison method. It should be noted, however, that these pairwise multiple comparisons have a lower statistical power than the planned contrasts in the mixed effect models. Therefore, the result and discussion sections focus on the outcome of the latter approach.  

In addition to the R packages that were already mentioned, we used RVAideMemoire (Maxime Hervé 2019) for visual inspection of models’ assumptions, plyr (Wickham 2011) and dpylr (Wickham et al. 2019) for data wrangling, and plotrix (Lemon 2006), ggplot2 (Wickham 2016) and ggpattern (FC et al. 2022) for data visualization, and emmeans (Lenth 2021) and multcompView (Graves et al. 2019) for pairwise comparisons.

The data file is a txt file which can be opened in Microsoft Excel, R and most other statistical softwares.