Decreased soil multifunctionality is associated with altered microbial network properties under precipitation reduction in a semiarid grassland

Our results reveal different responses of soil multifunctionality to increased and decreased precipitation. By linking microbial network properties functions, we also show that complexity potentially competitive interactions are key drivers multifunctionality. Global climate warming during the past few decades has resulted in intensification hydrological cycle [1, 2], thereby leading shifts precipitation patterns at global as well regional scales [3, 4]. Such altered regimes expected have dramatic ecological consequences, including changes assembly variations biodiversity ecosystem especially water-constrained areas, such arid semiarid ecosystems [5-7]. Accounting for more than 40% earth's terrestrial surface [8], dryland highly sensitive due their persistent low inputs [6, 9]. A substantial body literature documented how respond precipitation; however, most studies largely focused on response above-ground plant communities [10-14]. Plant generally terms productivity [15], phenology [16], community composition [17], resource use [18], although is context-dependent. comparison, our understanding implications regarding below-ground drylands related functional dynamics limited, even though this situation improved recent years [18-22]. observational survey over large spatial underscored importance investigating change they diversity a better predictor function relatively regions compared [23]. In addition, predicted rapid may pose challenge organisms (e.g., by inducing water or oxygen deficiency) [11, 24], while altering biological [25, 26]. effects be pronounced already vulnerable, water-limited ecosystems, where stability relationships depends greatly availability [27-29]. cases, reshape coexistence patterns, causing strong cascade effect microbial-mediated functions 26, 30, 31]. Thus, conducting field explore general structures underlying mechanisms associated with imperative. Ecosystem inherently multifunctional, reflecting ability an deliver multiple services simultaneously, fertilizer availability, elemental cycling, organic matter decomposition [23, Numerous shown biodiversity—plant microbial—supports microcosm, regional, 32-34]. For instance, reported significant positive correlation between [35, 36]. However, there paucity information precipitation) its relationship biodiversity. More importantly, microbiome structured [37, 38]; complex interactions, competition symbiosis, formed either one groups microbes via exchanges materials, energy, [39]. can act type selection force deterministically govern thus regulate structure [40-42]. Microbial co-occurrence networks mechanistically unravel offer insights [39, 43, 44]. Recent awareness led surge exploring microbe–microbe association under habitats stresses [28, 34, 38, 45, 46]. Although analysis not always indicate true [47, 48], it help understand [49-51]. example, Wang et al. [26] investigated three spanning 3700 km northern China showed higher complexity. study conducted experimental facility Germany future conditions, precipitation, increase bacterial–fungal [52]. Importantly, been demonstrated important driver multifunctionality, determine direction strength diversity–function [46, 53]. little known responds participates regulating This weaken capacity predict various scenarios, extremely regimes, loess hilly region China. Here, manipulation experiment simulate situ (ambient conditions ±50% treatments) abandoned grassland Based historical annual Loess Plateau region, used treatment (Supporting Information: Figure S1). We examined bacterial fungal using high-throughput sequencing 16S ribosomal RNA internal transcribed spacer genes, respectively. obtained data set 17 mediated microbes, nutrient provisioning, growth efficiency, labile (LOM) decomposition, recalcitrant (ROM) decomposition. quantified assembly, constructed networks, evaluated interactions. aim answer two following questions: (i) affect multifunctionality; (ii) diversity, processes, participate Decreased significantly suppressed LOM thus, averaging 42.6% (p < 0.001; 1A); ROM 38.9% 1A). contrast, efficiency 35.2%, but had no other groups. Analysis C fractions further proportion (R-CS) hydrochloric acid-resistant carbon (HCl-ROC), whereas (L-CS) permanganate oxidizable (POXC) exhibited contrasting trends S3). Additionally, considering consistency among multifunctional indices, index was characterize subsequent S4). markedly richness 7.9% 20.9%, respectively 0.01; 1B). 5.3%, remained stable. Spearman's indicated positively correlated each single Figures S5A S6A). Consistent were when applying phylogenetic proxies S5 Interestingly, indices considered both bacteria fungi together predictive those only (steeper slope; Supporting S5B S6B). Random forest (RF) after accounting physicochemical (Figure 1C S7). Null-model analyses revealed reduced stochasticity S8A). Furthermore, migration (m-value neutral model) along gradient, S8C). habitat niche breadths S8B). Conversely, breadth S8). Subsequently, metacommunity cross-kingdom all samples extracted subnetworks. identified four dominant clusters, >80% phylotypes strongly co-occurred within 2A). found these clusters (clusters 1 2) cluster 3 negative 2B S9). Further phylum level Actinobacteria, Proteobacteria, Ascomycota main representative taxa first Acidobacteria predominant S10). The topological features subnetwork differed from control S11). Specifically, number nodes edges, average degree, clustering coefficient, graph density, betweenness centrality, network, 2B); path length, denoting sparsity, 2B). parameters representing length negatively 2C). B–F links opposite Neg 2B,C). Moreover, bacterial, fungal, treatments estimate All detected scale-free modular Table S1 Appendix Material robustness increasing vulnerability S12–14). Kolmogorov–Smirnov (K–S) tests node-level CK S2). RF biodiversity, complexity, could 3A). Soil moisture SOC abiotic factors affecting Partial 3B) robust After controlling coefficients categories 36.28%, 45.38%, 66.27%, 47.36%, 72.42%, almost unaffected properties. physical predictors next Piecewise structural equation modeling (SEM) 3C S15) Neg-Int directly regulated indirect through associations links. potential mechanism poorly explored area [33]. variation respect ambient grassland. rather impact highlights taxonomic feature driving does so because leads greater [34]. provides into Bacteria decomposers soil, whose distinct habits likely differently 54]. asymmetric adverse might explained drought powerful environmental filter [55, 56]. drives reduces reducing [57, 58]. interpretation applies bacteria, since decrease accompanied stochasticity. finding, albeit somewhat surprising, consistent idea stochastic driven dispersal limitation, characterizes across Scotland forests grasslands southern [59-61]. rate varied stark contrast invariant Compared spores typically restricted shorter distances; hence, limitation shapes [59, 62]. partly explain weakened [63]. An alternative, mutually exclusive, explanation predominate special physiological hyphal facilitate transport acquisition) [64]. decreasing Generally, group wider less affected filtration metabolically flexible [63, 65]. suggest microorganisms disturbances risk loss inhibiting metabolic capacity. biodiversity–ecosystem extensively studied above- [36, 66]. provide evidence ensured performance Previous large-scale observations microcosm experiments biomes facilitates [30, 31, mainly labor division unique characteristics complementarity [34, 67]. (a desirable function) reduction caused transfer substrates depletion [68, 69]. supported experiments, (here simultaneously) insensitive Historical phenomenon. prolonged shaped species limited carrying [70]. case, short-term result quick local community, requirements [7, 71]. Nevertheless, minor impacts emerged, time, favorable microenvironment, moisture, [20, 22]. Perhaps, intriguing findings relation role mediating First, dominated Ascomycota, Acidobacteria. Different employ life-history strategies, aspects [43, 72]. resource-rich (analogous Y strategist) [72, 73]. preferentially invests resources acquisition A-strategists), depolymerization nutrient-poor habitats. aromatic C) 74]. Second—but importantly—we stronger factor aligns previous work suggesting contribute 34], microbially derived processes necessarily captured sum coexisting individuals [75, 76]. Instead, integrated pathways carried out myriad regulation conferred tightly connected members) [50, 77]. Contrary prediction rarely reported. Competitive relationship, which lead collapse presence high [53, 78]. These coupled bacterial-fungal study, raise concerns change. overestimation function; therefore, should incorporate studies. hindered limitations: widely distributed multitrophic levels predators trophic [31, 36]). alludes improve predictions function, correlations some [31]. Therefore, unassessed need manipulative experiments. As simple representations intricate system, yield spurious [79]. direct inference difficult [80]. motivation infer remains strong. particular, valuable tool identifying symmetric species, mutualism involving impact. Crucially, Together, diverse complicated alleviate functioning. site located Wuliwan watershed Hilly (36°51′–36°52′ N, 109°19′–109°21′ E, 1061–1371 m elevation) Ansai Country, Shaanxi Province, characterized temperate climates, temperature rainfall 8.8°C 505 mm, evapotranspiration 962.3 mm. classified Calcaric Cambisols [81]. historically experienced severe vegetation destruction erosion. Since 1970s, Chinese government implemented series restoration measures. One approaches convert farmland natural [82]. built 13 (i.e., 2006). Astragalus melilotoides, Artemisia scoparia, Poa sphondylodes, Patrinia heterophylla. initiated flat hilltop July 2017 investigate processes. blocked split-plot design, (−50%, −25%, ambient, +25%, +50% mean primary nested (whole-year unwarming) secondary factor. short, blocks, six plots. There >5 buffer zones adjacent blocks. Each plot × m2 size 2 away plots block. Among block, without any selected, remaining five randomly assigned divided subplot unwarming subplot. −50% (DP), (Control), (IP) is, subsample blocks replicates. DP treatment, U-shaped transparent plexiglass sheets inclination approximately 10° placed metal hanger plot. covered 50% area, precipitate collected plastic containers. Water will subsequently manually added nearest designated IP 24 h end event. eventually received additional 2019, cores depth 0–10 cm sampled. Three (5 diameter) taken then thoroughly mixed form sample. visible roots rocks gently removed, fresh passed 2-mm mesh sieve separated parts. part air-dried analyses, another maintained 4°C measuring biomass enzyme activities, last stored −80°C DNA extraction. pH, (SOC), total nitrogen (TN), phosphorus (TP) measured according standard testing methods. (HCl-ROC) SOC, [83, 84]; diffuse reflectance infrared Fourier-transform spectroscopy midinfrared range structures. variables determined detailed (Appendix 1). Seventeen assessed distinguished groups, provisioning (soil dissolved C, available P, inorganic N [NH4+-N NO3−-N]), (microbial turnover rate), (extracellular activities sugar degradation [β−1,4-glucosidase β−1,4-d-cellobiohydrolases], chitin [β−1,4-N-acetylglucosaminidase leucine aminopeptidase], P mineralization [alkaline phosphatase], (lignin degradation) [peroxidases polyphenol oxidases]. To eliminate differences measurement scale standardized 0 1. produce quantitative sample, calculated [31], scores individual functions. ensure contributed equally alternative weighted Finally, principal coordinate identify dimensions [85]. Details calculations 2). description extraction MiSeq presented 3). Before calculating operational unit (OTU) tables rarefied minimum sequences per sample OTU richness, conservative metric, indicator [36]. values averaged represent taxa. metric did bias. shaping null-model-based normalized ratio (NST) [86]. Levins' [87] estimated R package “spaa” [88]. model assess determining [89]. probability random being compensated replacement estimating parameter m. NST packages “NST” “minpack.lm”, interkingdom consisting constructed, based Pearson's approach matrix theory [38]. avoid rare OTUs reliability, filtering performed before calculation. Only present 9 18 included subnetworks features, centrality “igraph.” obtain reflects Note inverse calculation index. (Neg) proportions interacted (B–F links, respectively). samples. evaluate separate relative modularity, vulnerability, 4.2). K–S compare node attributes “stats” [43]. Molecular Ecological Network Pipeline [41], visualized interactive Gephi platform (https://gephi.org). construction characterization 4). Unless otherwise stated, statistical software v.4.2.0 [90]. linear regressions slopes calculated. major “rfPermute” [91], percentage mean-squared error variables. whether particular variable depended [92], partial coefficient difference zero-order implying controlled. “ggm” “psych.” SEM (pH, temperature, TN, TP, OC fraction), “piecewiseSEM” [93]. goodness fit Akaike criterion corrected Fisher's statistics. Xing Wang, Zekun Zhong, Xinhui Han, Gaihe Yang conceived project. Wenjie Li, Zhong study. Qi Zhang, Zhenjiao Weichao Liu, Hanyu Leyin analyzed data. Jing Ma, Zhenxia Quanyong Chengjie Ren, Naijia Xiao ideas experiment. Han wrote revised manuscript. thank Xinyi Shaojun Wu, Miaoping Xu, Xuqiao Lu, Chenghong Li sampling. financially National Natural Science Foundation (Nos. 41877543 42207368), Agricultural Technology Innovation Program Province NYKJ-2022-YL(XN)16), Open Fund Key Lab. Land Degradation Restoration northwestern Ningxia University LDER2022Q01). authors declare conflict interest. support corresponding author upon reasonable request. generated Center Biotechnology Information's GenBank database project accession numbers SRP392706 (https://www.ncbi.nlm.nih.gov/sra/?term=SRP392706) SRP392707 (https://www.ncbi.nlm.nih.gov/sra/?term=SRP392707). Materials (figures, tables, scripts, graphical abstract, slides, videos, translated version updated materials) online DOI iMeta http://www.imeta.science/. Please note: publisher responsible content functionality supporting supplied authors. Any queries (other missing content) directed article.