The effects of drought and nutrient addition on soil organisms vary across taxonomic groups, but are constant across seasons

Anthropogenic global change alters the activity and functional composition of soil communities that are responsible for crucial ecosystem functions and services. Two of the most pervasive global change drivers are drought and nutrient enrichment. However, the responses of soil organisms to interacting global change drivers remain widely unknown. We tested the interactive effects of extreme drought and fertilization on soil biota ranging from microbes to invertebrates across seasons. We expected drought to reduce the activity of soil organisms and fertilization to induce positive bottom-up effects via increased plant productivity. Furthermore, we hypothesized fertilization to reinforce drought effects through enhanced plant growth, resulting in even drier soil conditions. Our results revealed that drought had detrimental effects on soil invertebrate feeding activity and simplified nematode community structure, whereas soil microbial activity and biomass were unaffected. Microbial biomass increased in response to fertilization, whereas invertebrate feeding activity substantially declined. Notably, these effects were consistent across seasons. The dissimilar responses suggest that soil biota differ vastly in their vulnerability to global change drivers. Thus, important ecosystem processes like decomposition and nutrient cycling, which are driven by the interdependent activity of soil microorganisms and invertebrates, may be disrupted under future conditions.

reduction of the diversity of microbial communities 15 . Furthermore, drought was shown to cause a decline in soil microarthropod abundances 16 . In contrast, drought seems to have only marginal effects on nematode community composition 17 . Yet, a reduction of soil moisture content can induce community shifts via lower trophic levels, often favouring fungal-feeding nematodes over bacteria-feeders, as fungi perform relatively better under dry conditions 17,18 .
Nutrient enrichment is another key factor that affects the soil community by altering the physical and chemical properties of the soil, e.g., by influencing pH, soil porosity, and organic fractions [19][20][21] . Nitrogen addition has been identified to decrease soil microbial respiration and biomass, often leading to shifts in the soil microbial community composition under the use of mineral fertilizer (NPK) [22][23][24] . On the other hand, fertilization treatments were shown to increase soil microbial catabolic and functional diversity 25,26 . Furthermore, nitrogen addition alters the nematode community structure towards bacterivores, thus promoting the bacterial-dominated decomposition pathway 27 , and was shown to simplify communities 17 . Similarly, nitrogen was reported to create unfavourable conditions for soil microarthropods, leading to declines in abundance as well as diversity 28,29 . At the same time, nitrogen enrichment is one of the major drivers determining aboveground primary production 30 . Nitrogen and phosphorous addition are known to increase total aboveground biomass and consequently the quantity and quality of plant litter input to the soil 26,31 . In addition, nitrogen also affects plant rhizodeposition by altering the amount and quality of substrates released from roots 32,33 . Both mechanisms can potentially enhance resource availability via bottom-up effects and can therefore increase soil microarthropod abundances 34 . Concurrently, the fertilization-induced increase in aboveground biomass may cause higher transpiration rates, which are likely to reinforce drought effects on soil organisms 35 .
To investigate the interactive effects of extreme drought events and fertilization (NPK), we established a field experiment at the UFZ Experimental Research Station (Bad Lauchstädt, Germany), which combines the treatments of two globally distributed networks -the Drought-Network and the Nutrient Network 36 . Here, we tested the responses of soil microorganisms, nematodes, and overall soil invertebrate activity to the interactive effects of extreme drought and fertilization (NPK) across all seasons. Based on prior research, we hypothesized that (1) drought will reduce the activity of soil organisms, whereas (2) fertilization will increase their activity, owing to enhanced plant litter input that subsequently increases resource availability for soil organisms. Furthermore, we predicted that (3) the interactive effects of drought and fertilization will result in detrimental conditions for soil organisms as the negative effects of drought were expected to be further enhanced by increased plant growth under fertilization, resulting in reduced soil water availability for soil organisms.

Research site. The study site is located at the Experimental Research Station of the Helmholtz Centre for
Environmental Research (UFZ), which is situated in Bad Lauchstädt, Germany. The field site is located in the central German dry area with a mean annual precipitation of 487 mm and an average annual daily temperature of 8.9 °C (Meteorological data of Bad Lauchstädt, Helmholtz Centre for Environmental Research GmbH -UFZ, Department of Soil System Science, 1896-2017). The area represents an anthropogenic grassland, which is maintained by moderate mowing (twice a year since 2012). It is a successional plant community dominated by Vulpia myuros (L.) C. C. Gmel., Picris hieracioides (L.) and Taraxacum officinale (F. H. Wigg.) with Apera spica-venti (L.) P. Beauv. and Cirsium arvense (L.) Scop. being very common. The soil is classified as a haplic chernozem, developed upon carbonatic loess substrates, distinguished by a composition of 70% silt and 20% clay 37 . Within the upper 30 cm, the soil contains 0.18% total nitrogen, 1 g kg −1 total phosphorus, and 20 g kg −1 total potassium. For more details see Altermann et al. 37  Experimental design and treatments. The experimental site was established in March 2015. The experimental design consists of five blocks with five plots each. The plots have a size of 2 × 2 m and are arranged at a distance of 3 m from each other (Fig. S1). The experiment includes two treatments with two levels each (first applied in March 2016): drought (control/drought) and fertilization (no NPK/NPK addition), as well as their interaction (drought × fertilization). Notably, this experiment crosses treatments of two globally distributed experimental networks: the full NPK fertilization treatment of the Nutrient Network 36 and the drought treatment of the Drought-Network (http://www.drought-net.colostate.edu/) [38][39][40] .
In order to simulate drought, a rainfall manipulation system was established 39 using corrugated acrylic strips. The roofs have a size of 3 × 3 m and reduce precipitation by 55% throughout the year, simulating a severe long-term reduction in precipitation. Roofs were built with a slope of 20° to ensure water runoff and account for the expected snow load in the region. Exclusion of potential artefacts was realized by equal roof constructions using inverted acrylic strips intended to let rainfall pass 41 (Fig. S2). To control for possible infrastructure effects of the roof constructions itself, a fifth plot was added to each block without any roof construction (ambient plots), thus receiving ambient precipitation (not crossed with the fertilization treatment and thus only used to assess if the roof construction itself affected soil water content, see Fig. S1). To validate the drought treatment, soil water content was quantified on all plots in every sampling campaign. All three precipitation levels differed significantly in their soil water content (Tukey's HSD test, p < 0.05): as intended, the lowest soil water content was found for the drought treatment (−19.4% compared to the ambient plots). Also the infrastructure control plots (with SCIENTIFIC RepoRts | (2019) 9:639 | DOI:10.1038/s41598-018-36777-3 concave roof constructions) differed significantly from the ambient plots (without roof construction), indicating that there were effects of the roof construction itself (−13.4%). Furthermore, soil water content varied significantly between seasons (Table S1; Fig. S3). The fertilization treatment was realized by annual addition of a mixture of separate fertilizers for nitrogen (N), phosphorus (P) and potassium (K) (i.e. NPK fertilization; applied at 10 g m −2 y −1 by elemental mass for each of the three main fertilizer elements) before each growing season. Following the protocol of the Nutrient Network 36 , we used coated urea (CO(NH 2 ) 2 ) as nitrogen fertilizer (Multicote, Haifa -Pioneering the Future Haifa, Matam-Haifa, Israel), P 2 O 5 as phosphorus fertilizer (Triple Super Phosphate, Delitzscher Landhandels-und Dienste GmbH, Delitzsch, Germany), and K 2 SO 4 as potassium fertilizer (KaliSOP gran. max. 1.0% Cl, K + S KALI GmbH, Kassel, Germany). In addition, the micronutrient mix "Micromax Premium" (Everris, Geldermalsen, The Netherlands) was applied in the first treatment year 36  Soil sampling. The first soil sampling took place in March 2016. Sampling campaigns were repeated every three months to cover every season (spring, summer, fall, winter) from March 2016 to December 2017 (i.e., eight samplings across two years). Samples were taken on all plots with roof construction (drought and control) with a steel core sampler (1 cm in diameter; 15 cm deep). Seven subsamples per plot were homogenized, sieved at 2 mm, and stored at 4 °C. Soil samples were used to determine soil water content and microbial respiration. In addition, nematodes were extracted from the soil samples in spring and summer of 2017, and pH was measured for all four seasons in 2017.
The Bait Lamina Test. Feeding activity of soil invertebrates was surveyed using the bait lamina test (Terra Protecta GmbH, Berlin, Germany), which presents a commonly used rapid ecosystem function assessment method 42 . The test uses rigid PVC sticks (1 mm × 6 mm × 120 mm) with 16 holes of 1.5 mm diameter in 5 mm distance. Original sticks were filled with a bait substrate consisting of 70% cellulose powder, 27% wheat bran, and 3% activated carbon, which was prepared according to the recommendations of Terra Protecta. The bait substrate is primarily consumed by mites, collembolans, nematodes, enchytraeids, millipedes, and earthworms, whereas microbial activity plays a minor role in bait loss [43][44][45][46] . The bait strips were inserted vertically into the soil with the topmost hole just below the ground surface. To avoid damaging the strips, a steel knife was used to prepare the ground prior to insertion. Five strips were used per plot to account for spatial heterogeneity 47 . For each sampling campaign, the bait lamina strips were removed from the soil after three weeks of exposure and evaluated directly in the field. Bait consumption was recorded for each of the 16 holes per strip as empty (1), partly empty (0.5), or filled (0). Thus, soil invertebrate feeding activity per bait strip could range from 0 to 16 (maximum feeding activity). Mean bait consumption per plot (averaged among the five strips) was calculated prior to statistical analyses.
Microbial biomass and activity. An O 2 -microcompensation system was used to measure the respiratory response of soil microorganisms in two separated steps using approximately 6.5 g of fresh soil 48 . First, basal respiration was determined as a measure of soil microbial activity (µl O 2 h −1 g −1 soil dry weight) without the addition of any substrate. Second, the maximal respiratory response to a single addition of glucose (4 mg g −1 dry weight soil, solved in 1.5 ml distilled water) allowed us to determine soil microbial biomass (μg Cmic g −1 soil dry weight) 49 . For an overview of the experimental setup see Fig. S4. Nematode analysis. Nematode extraction was conducted with a modified Baermann method 50 .
Approximately 25 g of soil per plot were transferred to plastic vessels with a milk filter and a fine gauze (200 µm) at the bottom and placed in water-filled funnels. More water was added to saturate the soil samples and to ensure a connected water column throughout the sample and the funnel. Hence, nematodes migrated from the soil through the milk filter and the gauze into the water column and gravitationally-settled at the bottom of a closed tube connected to the funnel. After 72 h at 20 °C, the nematodes were transferred to a 4% formaldehyde solution. Nematodes were counted at 100x magnification using a Leica DMI 4000B light microscope. Identification was conducted at 400x magnification. For identification, sediment material from the bottom of each sample vial was extracted with a 2 ml plastic pipette and examined in temporary mounted microscope slides. At least 100 well-preserved specimens (if available in the sample) were randomly selected and identified to genus (adults and most of the juveniles) or family level (juveniles), following Bongers (1988) 51 . Nematode taxa were then arranged into trophic groups (bacteria-, fungal-and plant-feeders, omnivores and predators) 52,53 . Due to low densities, omnivorous and predatory nematodes were grouped into a combined feeding type for most analyses, which was based on the fact that both groups often share similar traits, like being carnivores (i.e. representing higher trophic levels) and being persisters rather than colonizers on the c-p scale (see explanation below), thus responding similarly to environmental disturbances 54 . Nematodes were ordered according to the colonization-persistence gradient (c-p values) 55,56 , which classifies nematode taxa based on their life history strategy (i.e. r or K strategists). Cp-1 taxa are distinguished by their short generation cycles and high fecundity. They mainly feed on bacteria. Cp-2 taxa have longer generation times, lower fecundity and consist of bacterivores and fungivores 57 . Both are categorized as r-strategists. Cp-3 to cp-5 are classified as K-strategist nematodes with longer generation times, higher trophic feeding levels and increasing sensitivity against disturbances 57 . The c-p-values can be used to calculate the Maturity Index (MI) as weighted means of nematode families assigned to c-p-values. It is used to describe soil health and as an indicator of overall food web complexity 55,56 .
with v(i) being the c-p-value of a taxon i and f(i) being the frequency of that taxon in a sample. Furthermore, nematode taxa were assigned to functional guilds according to Ferris et al. 57 , which then served as a basis to calculate additional indices. Functional guilds refer to the following trophic groups: bacterial feeders (Ba X ), fungal feeders (Fu X ), omnivores (Om X ), and carnivores (Ca X ). Associated numbers (i.e., the x of the respective trophic group) are again referring to the c-p values described above. The Enrichment Index (EI) indicates the responsiveness of the opportunistic bacterial (Ba 1 and Ba 2 ) and fungal feeders (Fu 2 ) to food web enrichment 57 and is calculated as follows: with e as weighted frequencies of Ba 1 and Fu 2 and b as weighted frequencies of Ba 2 and Fu 2 nematodes 57 . The Channel Index (CI) reflects the nature of decomposition channels through the soil food web. High values indicate a predominant decomposition pathway of organic matter dominated by fungal-feeding nematodes, whereas low values refer to bacterial-dominated decomposition pathways 57 .

Statistical analyses. Soil microorganisms and invertebrates.
Linear mixed-effects models (LMM) were used to analyse the effects of drought, NPK fertilization, season, and their interactions on invertebrate feeding activity, microbial activity, and microbial biomass using the R-package "nlme" 58 . The random intercept of the model was structured with plots nested within blocks, nested within year (year as a categorical factor). To account for repeated measurements within plots, we compared first-order autoregressive and compound symmetry covariance structures based on the Akaike information criterion (AIC). As differences between AIC values were lower than 2, the simplest covariance structure (i.e. compound symmetry) was used. Based on the importance of soil water content for microbial activity and biomass 59 , soil water content was added as an additional explanatory variable to the LMMs (Tables S3 and S4, Figs S5 and S6). As we were expecting a strong relation between aboveground plant biomass and microbial biomass 60 , additional LMMs were used to test the influence of plant biomass on microbial biomass (Table S5, Fig. S7). To evaluate model variation explained by fixed and random effects, marginal and conditional R 2 were calculated using the "MuMIn" package 61 ; marginal R 2 represents model variation explained by fixed effects in the final model and conditional R 2 represents model variation explained by both random and fixed effects.
Nematode community indices. LMMs were also used to assess the effects of drought, NPK fertilization, season (spring and summer 2017), and their interactions on nematode indices, i.e. Enrichment Index, Structure Index, Channel Index, and Maturity Index. A random intercept with plots nested within block was included in the models. We accounted for repeated measurements within plots by using a compound symmetry covariance structure, which fitted the data better than a first-order autoregressive covariance structure based on the AIC. To evaluate model variation explained by fixed and random effects, marginal and conditional R 2 were calculated using the "MuMIn" package 61 .
Soil pH. We also used LMMs to assess the effects of drought, NPK fertilization, and their interaction on soil pH for all seasons in 2017. Additionally, we tested the influence of soil pH on invertebrate feeding activity. A random intercept with blocks nested within season was included in the models. We accounted for repeated measurements within plots by using a compound symmetry covariance structure, which fitted the data better than a first-order autoregressive covariance structure based on the AIC. To evaluate model variation explained by fixed and random effects, marginal and conditional R 2 were calculated using the "MuMIn" package 61 .
Nematode density, richness, and trophic groups. Generalized mixed-effects models (GLMM) were used to assess the effects of drought, NPK fertilization, season (spring and summer 2017), and their interactions on nematode richness, total density (i.e. total number of individuals in the nematodes community) and the abundance of each trophic group (i.e. percentage of individuals in each trophic group). Nematode richness and total density of nematodes were modelled with Poisson distribution (link = "log"), while the trophic groups were modelled with Binomial distribution (link = "logit"). The random intercept of the model was structured with plots nested within blocks. To account for over-dispersion, an observation-level random effect was used in the model with omnivorous and predatory nematodes as a response variable.
Nematode functional guilds and cp groups. GLMMs were also used to assess the effects of drought, NPK fertilization, and their interactions on nematode functional guilds and cp-groups (Table S6)  (link = "logit"). The random intercept of the model was structured with plots nested within blocks, nested within sampling (sampling as a categorical factor). GLMMs were performed using the "lme4" package 62 .
Model assumptions of the LMMs and the GLMMs were checked by visually inspecting residual plots for homogeneity and quantile-quantile plots for normality (i.e. no correlation between the residuals and the fitted parameters of the model). To meet the assumptions of the model, invertebrate feeding activity and microbial activity were log-transformed (log(x + 1)). Figures are based on mixed-effects model fits extracted using the package "ggeffects" 63 . All statistical analyses were conducted using R version 3.4.2 64 .

Results
Soil microbial responses. Soil microbial activity ranged from 0.7 to 5.1 µl O 2 h −1 g −1 dry weight soil with an average of 1.7 µl O 2 h −1 per g dry weight soil across all measurements. We could not detect a significant effect of the drought or the fertilization treatment on soil microbial activity (Fig. 1a). However, microbial respiration was significantly affected by season, with lowest activity in summer and highest activity in winter (Table 1; Fig. 1b). In addition, we found a positive relation between microbial activity and soil water content (F 1, 111 = 170.83, p < 0.001; Table S3) that was independent of the drought and fertilization treatment (Fig. S5).
Soil microbial biomass ranged from 168.0 to 979.8 µg Cmic g −1 dry weight soil with an average of 530.6 µg Cmic g −1 dry weight soil across all measurements. Overall, soil microbial biomass increased significantly with NPK fertilization (Fig. 1c). Microbial biomass was also significantly affected by season with lowest biomass in summer and highest biomass in fall (Fig. 1d, Table 1). Furthermore, fertilization and soil water content interactively affected microbial biomass (F 1, 111 = 10.60, p = 0.002; Table S4); while microbial biomass increased with higher soil water content under ambient conditions, it slightly decreased with higher soil water content on plots with NPK fertilization (Fig. S6). In addition, soil microbial biomass increased significantly with aboveground plant biomass (F 1, 59 = 8.81, p = 0.004; Table S5; Fig. S7).

Nematode responses.
Neither total nematode density nor richness were significantly affected by any of the experimental treatments (Fig. 2a,b; Table 2); we could only detect significant differences between spring and summer (Fig. S7a,b). Among the nematode trophic groups, only bacteria-feeders were significantly increased by the fertilization treatment (Fig. 2c), which was mainly due to a significant increase of the r-strategic Ba 1 -nematodes (χ 2 = 4.57, p = 0.032; Fig. S8a). In addition, bacteria-feeders were highly abundant in summer (Fig. S9c). Plant-feeders were affected by the interactive effects of fertilization and season: while fertilization favoured plant-feeding nematodes in spring, it had a negative effect in summer (Figs 2d and S9d). The combined group of omnivorous and predatory nematodes marginally significantly decreased under drought and NPK fertilization (Fig. 2e), with a stronger negative effect of fertilization in spring (Fig. S9e). Fungal-feeders were not significantly affected by any of the treatments (Fig. 2f). A closer look at changes in the nematode community composition revealed that cp2 plant-feeding nematodes increased with drought (χ 2 = 6.65, p = 0.0099; Fig. S10c;   Table 1. Results of linear mixed-effects models for the effects of drought, NPK fertilization, season, and their interactions on soil invertebrate feeding activity, soil microbial activity, and soil microbial biomass. A random intercept with plots nested within blocks, which were nested within year was added to the model. A compound symmetry covariance structure was used to account for repeated measurements within plots. Marginal R 2 : model variation explained by fixed effects; conditional R 2 : model variation explained by both fixed and random effects. Logarithmic transformations were used for soil invertebrate feeding activity and soil microbial activity. NPK = NPK fertilization. *p < 0.05; **p < 0.01; ***p < 0.001. While the increase in the Enrichment Index was marginally significant under drought conditions (Table 3; Fig. 3a), drought led to a marginally significant decrease of the Structure Index and the Maturity Index (Fig. 3b,c). Fertilization increased the importance of the bacterial decomposition channel as indicated by a marginally significant decrease of the Channel Index (Fig. 3d). In addition, fertilization decreased nematode complexity as indicated by a significant decrease of the Structure Index and Maturity Index. (Fig. 3b,c).
Soil invertebrate feeding responses. Mean soil invertebrate feeding activity per plot ranged from 0 to 60% of consumed bait substrate with an average of 11% bait consumption. Feeding activity was significantly affected by an interactive effect of drought x fertilization; overall, fertilization decreased invertebrate feeding activity at ambient precipitation, but had no significant effect under drought conditions (Fig. 4a). These treatment effects were consistent across seasons (no treatment x season interaction), however, the level of soil invertebrate feeding activity varied strongly between seasons and tended to be highest in summer and strongly decreased in winter (Table 1; Fig. 4b).
Soil pH. Soil pH was significantly reduced by NPK fertilization across all seasons in 2017 (F 1,56 = 14.11; p < 0.001; Table S7; Fig. S11a). In addition, we could show that an decrease in soil pH led to a marginally significant decrease in invertebrate feeding activity across all seasons in 2017 (F 1,58 = ; p = 0.07; Table S8; Fig. S11b).

Discussion
We studied the impacts of two major global change drivers, namely drought and fertilization, and their interactive effects on soil communities and functions. By investigating the responses of a wide range of soil organisms across all seasons within a two-year timeframe, we gained a comprehensive picture of how soil ecosystem functions may be altered in a changing world. Intriguingly, we saw vastly different responses among trophic levels that were constant across seasons.
Our first hypothesis, predicting detrimental effects of drought on soil organisms, was confirmed to some extent: drought reduced soil invertebrate feeding activity and led to a more disturbed soil nematode community structure, whereas soil microbial activity and biomass were not significantly affected. The dependency of soil invertebrates on soil moisture content is well documented 65 as is the fact that soil microfauna is more prone to water stress than bacteria and fungi 66 . The detrimental drought effects on faunal activity are thus in line with previous studies, which claim that abiotic conditions shape the performance of the soil faunal community 67 . Microarthropods (mites and collembolans), enchytraeids, and earthworms are some of the most relevant groups  Table 2. Chi-squared values (χ 2 ) of the generalized mixed-effects models for the effects of drought, fertilization, season (spring and summer 2017), and their interaction on soil nematode density and richness using Poisson distribution and percentage of each nematode trophic group using Binomial distribution. Plots nested within blocks served as a random intercept in the model. NPK = annual NPK fertilization. (*) p < 0.1; *p < 0.05; ***p < 0.001.  Table 3. Results of linear mixed effects models for the effects of drought, fertilization, season (spring and summer 2017), and their interaction on soil nematode indices. Plots nested within blocks served as a random intercept in the model. A compound symmetry covariance structure was used to account for repeated measurements within plots. Marginal R 2 : model variation explained by fixed effects; conditional R 2 : model variation explained by both fixed and random effects. NPK = annual NPK fertilization. (*) p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001. found in the upper soil layer in temperate regions 43 and are likely to account for most of the feeding activity leading to bait perforation 68,69 . Extreme drought not only forces them to migrate to deeper soil layers, but also interferes with their reproduction and development success, which is possibly the reason why they are highly susceptible to drought [70][71][72][73] . Furthermore, drought conditions entail drier food sources for detritivores, which are more difficult to digest 47 .

Nematode response variable
In line with our expectations, drought was also responsible for (moderate) shifts in nematode indices. The environmental conditions were changed towards an enriched, more disturbed, less structured system, with a higher proportion of opportunistic nematodes. Already at ambient precipitation levels, the guilds of the opportunistic Fu 2 -and Ba 2 -nematodes accounted for the highest shares at our experimental site, indicating a basal food web that is capable to cover a wide ecological amplitude and is already adapted to some environmental stress 57 . The positive response of the Enrichment Index to drought suggests mortality at higher trophic levels, which subsequently promoted nutrient enrichment and gave further rise to opportunistic nematodes 57 . Thus, overall, drought led to simplified trophic structures of the nematode community.
In contrast to our initial hypothesis, soil microbial activity and biomass were not affected by the drought treatment. This was unexpected given the intensity of the drought treatment applied in the experiment (precipitation was reduced by 55% during the entire study period) and the fact that most soil microbes are strongly dependent on soil moisture 59,74 . However, this finding is in line with Pailler et al. 75 , who also found little soil microbial functional responses to drought. In addition, we observed distinct variation of microbial activity and biomass across seasons and could reveal that soil water content explained a significant proportion of the variability in microbial  activity, yet irrespective of the treatments. Notably, soil water content itself was strongly varying among seasons (more than in response to the drought treatment; Tables S9 and S10). This suggests that microorganisms residing in the top soil are in general highly dependent on soil moisture levels, which are fluctuating across the year. Therefore, they must be able to sustain drought periods, for instance, through physiological modifications. Such adaptations may comprise an adjustment of internal water potential, sporulation, or production of exopolysaccharides that provide protection against exsiccation 76,77 . This provides evidence that temporal environmental variability is a strong predictor of species activity and emphasizes the dependency of soil organisms on seasonal patterns [78][79][80] . We therefore infer that seasonal fluctuations in natural precipitation at our site (Table S9) may have led to acclimatization of the microbial communities to natural drought periods as they are part of their climatic history 81,82 . We speculate that this may explain the negligible responses to the experimentally induced drought, which may still be within the adaptive range of the community (Table S10). The responses of soil biota were again ambiguous with regard to our second hypothesis, in which we expected fertilization to promote the activity of soil biota. Consistent with our hypothesis, soil microbial biomass increased under fertilization. Soil invertebrate feeding activity, however, substantially declined under elevated nutrient supply, questioning the universal validity of our initial hypothesis. Notably, in a recent meta-analysis, Yian et al. 83 reported negative effects of nitrogen fertilization on soil microbial biomass, which may be attributable to changes in the community composition, but this requires further investigation, because microbe-mediated activity markers were, on the other hand, increased. The pronounced responsiveness of soil microbial biomass to NPK fertilization that we found in our study is in line with similar studies reporting positive effects of nitrogen fertilization on microbial biomass and changes in microbial community structure and function 84,85 . Fertilization certainly enhanced nutrient availability, resulting in higher yields of aboveground plant biomass (Berger et al., in prep.), which is often accompanied by an expanded root system 30 . Subsequently, this increases the release of organic compounds into the soil 86 , providing substrates that support the growth of microbial communities towards higher population densities 60 .
Also nematode indices were highly responsive to NPK fertilization, resulting in a less structured and more disturbed system with an enhanced bacteria-driven decomposition pathway. The latter was additionally supported by a strong increase in bacteria-feeding nematodes (especially Ba 1 ) and an equally strong decrease in fungal-feeding nematodes (Fu 3 ) under fertilization, indicating that fertilization had detrimental effects on soil fungi. This is in line with the general notion that fertilization favours bacteria-dominated decomposition 18,27,57,87 . The simplification of the nutrient-enriched soil food web is also reflected by declines of the omnivorous nematodes (Om 4 ) and the cp3 and cp4 nematodes, which consist of long-living, pollutant sensitive, rather immobile organisms that are prone to environmental stress 56 . When linking these changes reported for the nematode indices to the responses of the microbial community, our results suggest that fertilization promoted the growth of bacteria, which then accounted for a strong increase in microbial biomass. As a result, this restructuring of the microbial community may thus have provided the basis for the observed increase in bacteria-feeding nematodes and some of the shifts in nematode indices.
Current evidence for soil invertebrate responses to fertilization is equivocal. Several studies reported an enhancing effect of fertilization on soil invertebrate activity and diversity 88,89 , whereas others revealed no such effects 16,90 , or recorded declines in soil fauna abundances and diversity after fertilizer application 43 . Most likely, the diminished feeding activity of soil invertebrates can be explained by alterations of soil physicochemical properties. In our study, we found that fertilization decreased soil pH, which can serve as one potential explanation of the reported decline of invertebrate feeding activity. This is in line with previous studies showing that reduced soil pH as a result of mineral fertilization was negatively correlated with the abundance of most soil invertebrates 91,92 . Especially earthworms, which represent a significant part of soil invertebrate biomass 93 , are known to decline noticeably in numbers under reduced soil pH 94 . Overall, environmental constraints, such as altered energy channels, simplified food web structures, and reduced soil pH outweighed potential positive bottom-up effects through enhanced plant growth, which we expected to find.
Prior studies emphasized the importance of interactive effects of global change drivers as they will profoundly alter ecosystem functions and services 7,95 . Accordingly, our third hypothesis predicted fertilization to reinforce drought effects, resulting in reduced soil water content and thus aggravated living conditions for soil organisms. Indeed, drought and fertilization interacted significantly in restraining soil invertebrate feeding activity, partially supporting our hypothesis. However, as discussed above, fertilization obviously created an unfavourable environment for most soil invertebrates at our site. Although both global change drivers individually decreased soil invertebrate feeding activity, the interaction, however, did not result in further declines. We therefore conclude that the combined effects may have led to a distinct restructuring of the soil faunal community by promoting species better adapted to adverse conditions, and thus revealing no measurable change in net effects. Since the combined global change drivers are likely to modify aboveground plant communities [96][97][98] , alterations in the quality of aboveground litter inputs can be expected. Leaf litter quality, in turn, affects soil fauna and might therefore be responsible for reshaping the faunal community 99,100 . With the methods applied in our study, however, we can only speculate about potential changes in the soil faunal community composition. This highlights the need for future research to detect which specific groups are responsible for bait perforation. This could be done, for instance, by exposing bait lamina strips with a labelled substrate under controlled laboratory conditions 101,102 . Building on that, the abundances of the most important groups of soil organisms could be monitored in the field, while being exposed to different global change drivers.
In contrast to the invertebrate feeding activity, microbial activity was not significantly affected by the interaction of the two global change drivers. Moreover, we could not detect any interactive effects on nematode indices or nematode groups. This illustrates the robustness of a large portion of the soil community to interactive global change effects, which might therefore be able to buffer prospective global change effects to a certain extent. In conclusion, the main groups of soil organisms investigated in the present study responded differently to the individual and interacting effects of global change drivers. Soil invertebrate activity was strongly impaired by both global change drivers and their interaction, while microbial biomass benefited from enhanced nutrient availability, and microbial activity was surprisingly unaffected by all treatments. Despite the strong seasonal dynamics of temperate regions, these treatment effects remained constant across all seasons within two years. Notably, nematode indices pointed to changes in the state of the ecosystem, shifting towards simplified and more disturbed systems under drought and especially under fertilization that mostly facilitated opportunistic species. We could show that soil biota differ considerably in their sensitivity to global change drivers and in their seasonal dynamics -also highlighting the importance of integrating seasonal effects into experimental frameworks. This may lead to far-reaching alterations of crucial ecosystem processes, since decomposition and nutrient cycling are driven by the interdependent concurrence of soil microbial and faunal activities 46 . By covering a range of different taxonomic and trophic levels of soil organisms, we could therefore show that single as well interacting global change drivers induce complex changes in soil food webs and functions.

Data Availability
The datasets analysed during the current study are available from the corresponding author on reasonable request.