IntroductionHuman pressures are numerous, highly diverse in nature1, and influence soil ecosystems at a global scale. One of the most important effects of global change (GC) are shifts in soil microbial populations, central to soil functioning. Several experiments have revealed the response of soil biota to alternative GC factors like warming2, drought3 or microplastics4, among others5. These microbial disturbances impact important soil functions, and monitoring them remains relevant for understanding anthropogenic impacts on soil ecosystems.However, most studies only include a limited number of GC factors, even though many may act concurrently in natural conditions. In order to address this gap, Rillig et al.(2019)6 designed a multifactor experiment including 10 GC factors of diverse nature1: warming (physical factor), drought, nitrogen deposition, increased salinity and heavy metal (inorganic chemical factors), microplastics (particle contamination) and antibiotics, fungicides, herbicides and insecticides (organic chemical toxicants). After applying them individually and in an increasing number of simultaneous combinations (up to 10) to soil grassland samples, results showed that multiple concurrent GC factors triggered directional shifts in soil properties. For instance, individual GC factors barely affected water drop penetration time, but the application of multiple concurrent factors caused a significant increase, more pronounced as the number of applied factors increased. These results highlighted the importance of studying not only the effect of individual GC factors, but also the combined effect of many.Increasing GC factors also triggered directional changes on soil fungal populations, but whether other microorganisms follow the same patterns remains unknown. This includes prokaryotes, which usually show different dynamics than fungi7,8 and are central to soil functioning. For instance, they are the only fixers of molecular nitrogen, a limiting soil nutrient, mediate phosphorus mobilization, critical for plant growth, decompose plant derived organic matter, and contribute to soil structure through the formation of aggregates9,10,11. Additionally, the response of viruses, understudied players of soil functioning with a key role in regulating microbial host dynamics and soil carbon pools12, to multiple GC factors has also not been studied.In order to understand whether prokaryotes and viruses show different responses to multiple GC factors than to individual GC treatments, we leverage 70 samples from the multi-factor experiment by Rillig et al (2019)6, including i) 10 controls, ii) 50 single GC factor samples (5 samples treated with each individual factor), and iii) 10 samples treated with random combinations of 8 concurrent GC factors (see Supplementary Fig. 1 for the factor composition of the samples), and analyse them following a comprehensive metagenomic exploration (Fig. 1). The 8-factor treatment de-emphasizes (through the random draws) the composition of factors, represents the changing multifactorial conditions observed in nature13, and yielded unexpected patterns in the original study by Rillig et al. (2019), making it appropriate for studying the effect of multiple concurrent GC factors.Fig. 1: Graphical summary of the experimental set-up.We sequenced the metagenomes of 10 control samples (treated with 0 factors, green), 50 samples treated with 10 individual GC factors (5 samples per factor, orange) and 10 samples treated with random combinations of 8 GC factors (purple), and sequenced their metagenomes for characterizing their prokaryotic and viral microbiome.Full size imageResultsDifferent prokaryotic composition under alternative GC scenariosSoils are the most biodiverse habitat on the planet14, and contain an immense number of uncultivated microbial species not present in reference databases15. These unknown species can be uncovered by constructing Metagenome-Assembled Genomes (MAGs) de novo, a method that is revealing a great degree of unknown biodiversity16. Hence, after sequencing the metagenomes of the 70 soil samples from the experiment by Rillig et al. (2019), trimming the reads and assembling them into contigs, we aimed at binning them into prokaryotic MAGs.After comparing the performance of different prokaryotic binning strategies (Supplementary Table 1), we restricted our analysis to the genomic bins computed with the multi-sample binning strategy by SemiBin217, which provided 653 medium quality (completeness ≥ 50%, contamination