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 to soil functions, we also show that network complexity and potentially competitive interactions are key drivers of soil multifunctionality. Global climate warming during the past few decades has resulted in the intensification of the hydrological cycle [1, 2], thereby leading to shifts in precipitation patterns at the global as well as regional scales [3, 4]. Such altered precipitation regimes are expected to have dramatic ecological consequences, including changes in microbial assembly and variations in biodiversity and ecosystem functions, especially in water-constrained areas, such as arid and semiarid ecosystems [5-7]. Accounting for more than 40% of the earth's terrestrial surface [8], dryland ecosystems are highly sensitive to altered precipitation regimes due to their persistent low precipitation inputs [6, 9]. A substantial body of literature has documented how dryland ecosystems respond to increased and decreased precipitation; however, most studies have largely focused on the response of above-ground plant communities [10-14]. Plant communities generally respond to altered precipitation in terms of productivity [15], phenology [16], community composition [17], and resource use [18], although the response is context-dependent. By comparison, our understanding of the implications of altered precipitation regimes regarding below-ground microbial communities in drylands and related soil functional dynamics is limited, even though this situation has improved in recent years [18-22]. A recent observational survey over large spatial scales underscored the importance of investigating the microbial responses to climate change in dryland ecosystems as they show that soil microbial diversity is a better predictor of the ecosystem function in relatively more arid regions compared to plant diversity [23]. In addition, the predicted rapid changes in precipitation regimes may pose a challenge for organisms (e.g., by inducing water or oxygen deficiency) [11, 24], while altering the biological interactions [25, 26]. Such effects may be more pronounced in already vulnerable, water-limited dryland ecosystems, where the stability of biological relationships depends greatly on water availability [27-29]. In such cases, altered precipitation may reshape the microbial coexistence patterns, thereby causing a strong cascade effect on microbial-mediated soil functions [6, 26, 30, 31]. Thus, conducting field studies in arid ecosystems to explore the general responses of microbial community structures to altered precipitation and the underlying mechanisms associated with ecosystem function is imperative. Ecosystem function is inherently multifunctional, reflecting the ability of an ecosystem to deliver multiple functions or services simultaneously, such as water and fertilizer availability, elemental cycling, and organic matter decomposition [23, 31]. Numerous recent studies have shown that biodiversity—plant or microbial—supports ecosystem multifunctionality at the microcosm, regional, and global scales [23, 32-34]. For instance, a global survey reported a significant positive correlation between soil microbial diversity and ecosystem multifunctionality in drylands [35, 36]. However, there is a paucity of information on the effects of climate change (e.g., altered precipitation) on ecosystem multifunctionality and its relationship with microbial biodiversity. More importantly, the soil microbiome is highly structured [37, 38]; complex interactions, such as competition or symbiosis, are formed by either one or multiple groups of microbes via exchanges of materials, energy, and information [39]. Such microbial interactions can act as a type of selection force to deterministically govern the community assembly and thus regulate microbial community structure [40-42]. Microbial co-occurrence networks can mechanistically unravel such complex ecological relationships and offer insights regarding the community structure and stability [39, 43, 44]. Recent awareness of this has led to a surge of studies exploring variations in the properties of microbe–microbe association networks under different habitats or stresses [28, 34, 38, 45, 46]. Although co-occurrence network analysis may not always indicate true interactions [47, 48], it can help to understand microbiome complexity and its responses to climate change [49-51]. For example, Wang et al. [26] investigated three habitats spanning 3700 km in northern China and showed that higher precipitation increased microbial network complexity. A recent study conducted at a global change experimental facility in Germany also showed that future climate conditions, including altered precipitation, increase the bacterial–fungal network complexity [52]. Importantly, the microbial ecological network complexity has been demonstrated to be an important driver of soil multifunctionality, and may even determine the direction and strength of diversity–function relationships [46, 53]. However, little is known regarding how the microbial network complexity responds to altered precipitation and thus participates in regulating ecosystem multifunctionality. This may weaken our capacity to predict ecosystem function variations under various future climate change scenarios, especially in dryland ecosystems that are extremely sensitive to altered precipitation regimes, such as the loess hilly region of China. Here, we conducted a manipulation experiment to simulate in situ precipitation changes (ambient conditions and ±50% precipitation treatments) in an abandoned grassland in the loess hilly region of China. Based on historical and predicted annual precipitation in the Loess Plateau region, we used ±50% precipitation treatment to simulate future precipitation patterns (Supporting Information: Figure S1). We examined soil bacterial and fungal community structures using high-throughput sequencing of 16S ribosomal RNA and internal transcribed spacer genes, respectively. We also obtained a data set of 17 ecosystem functions mediated by soil microbes, including nutrient provisioning, microbial growth efficiency, labile organic matter (LOM) decomposition, and recalcitrant organic matter (ROM) decomposition. We quantified the community assembly, constructed microbial co-occurrence networks, and evaluated the network complexity and microbial interactions. We aim to answer the two following questions: (i) how altered precipitation patterns affect soil ecosystem multifunctionality; and (ii) how microbial diversity, assembly processes, and microbial network properties respond to altered precipitation and participate in regulating soil multifunctionality. Decreased precipitation significantly suppressed the nutrient provisioning, microbial growth efficiency, LOM decomposition, and thus, the averaging multifunctionality by 42.6% (p < 0.001; Figure 1A); however, it significantly increased the ROM decomposition function by 38.9% (p < 0.001; Figure 1A). In contrast, increased precipitation significantly increased the microbial growth efficiency by 35.2%, but had no significant effect on the multifunctionality and other soil function groups. Analysis of C fractions further showed that decreased precipitation significantly increased the proportion of recalcitrant C structure (R-CS) and hydrochloric acid-resistant carbon (HCl-ROC), whereas labile C structure (L-CS) and permanganate oxidizable carbon (POXC) exhibited contrasting trends (Supporting Information: Figure S3). Additionally, considering the consistency of the trends among the three multifunctional indices, the averaging multifunctional index was thus used to characterize soil multifunctionality in the subsequent analysis (Supporting Information: Figure S4). Decreased precipitation resulted in markedly decreased soil bacterial and fungal richness by 7.9% and 20.9%, respectively (p < 0.01; Figure 1B). In contrast, increased precipitation resulted in significantly increased fungal richness by 5.3%, whereas bacterial richness remained stable. Spearman's correlation analysis indicated that the bacterial and fungal richness was significantly positively correlated with soil multifunctionality and each single function (Supporting Information: Figures S5A and S6A). Consistent results were also obtained when applying phylogenetic diversity or other multifunctionality proxies (Supporting Information: Figures S5 and S6A). Interestingly, our results showed that soil biodiversity indices that considered both bacteria and fungi together were generally more predictive of soil multifunctionality than those considering only one of the two (steeper slope; Supporting Information: Figures S5B and S6B). Random forest (RF) analysis also showed that soil biodiversity remained a significant and important predictor of ecosystem multifunctionality, even after accounting for multiple soil physicochemical properties (Figure 1C and Supporting Information: Figure S7). Null-model analyses revealed that decreased precipitation significantly reduced the stochasticity of bacterial assembly but increased that of fungi (p < 0.01; Supporting Information: Figure S8A). Furthermore, the fungal migration (m-value in the neutral community model) increased along the precipitation gradient, whereas the bacterial migration exhibited no significant changes (Supporting Information: Figure S8C). Decreased precipitation resulted in significantly decreased bacterial and fungal habitat niche breadths (p < 0.01; Supporting Information: Figure S8B). Conversely, increased precipitation had no significant effect on community assembly and habitat niche breadth (Supporting Information: Figure S8). Subsequently, we constructed a metacommunity cross-kingdom co-occurrence network of all samples and extracted subnetworks. We identified four dominant ecological clusters, including >80% of soil phylotypes that strongly co-occurred within the network (Figure 2A). We found a positive correlation between the richness of soil phylotypes within two of these ecological clusters (clusters 1 and 2) and multifunctionality, whereas cluster 3 exhibited a negative correlation with multifunctionality and a positive correlation with ROM decomposition (Figure 2B and Supporting Information: Figure S9). Further analysis at the phylum level revealed that Actinobacteria, Proteobacteria, and Ascomycota were the main representative taxa in the first two clusters, whereas Acidobacteria were predominant in cluster 3 (Supporting Information: Figure S10). The topological features of the subnetwork under decreased precipitation differed significantly from those found under the control and increased precipitation conditions (Supporting Information: Figure S11). Specifically, with decreased precipitation, the number of nodes and edges, average degree, clustering coefficient, graph density, and betweenness centrality, reflecting the complexity of the network, decreased significantly (Figure 2B); however, the average path length, denoting network sparsity, increased significantly (p < 0.01; Figure 2B). Spearman's correlation analysis showed that all the topological parameters representing network complexity were positively correlated with soil multifunctionality, whereas the average path length was negatively correlated (Figure 2C). In addition, B–F links significantly decreased in response to decreased precipitation and were significantly positively correlated with multifunctionality, while the opposite was true for Neg in B–F links (Figure 2B,C). Moreover, we also constructed bacterial, fungal, and cross-kingdom networks under different treatments to estimate the network stability in response to altered precipitation. All detected networks were scale-free and modular (Supporting Information: Table S1 and Appendix 3 of Supporting Information: Material S1). Decreased precipitation resulted in significantly decreased robustness of the bacterial, fungal, and cross-kingdom networks while increasing their vulnerability (Supporting Information: Figures S12–14). Kolmogorov–Smirnov (K–S) tests showed that the node-level features were significantly different between the decreased precipitation treatment and CK or increased precipitation treatment (p < 0.01; Supporting Information: Table S2). RF analyses indicated that soil biodiversity, network complexity, and Neg in B–F links could affect the soil function (Figure 3A). Soil moisture and SOC were identified as important abiotic factors affecting soil multifunctionality. Partial correlation analysis (Figure 3B) revealed a significant and robust effect of network complexity on soil multifunctionality. After controlling for network complexity, the correlation coefficients between the other categories and soil multifunctionality decreased by 36.28%, 45.38%, 66.27%, 47.36%, and 72.42%, respectively. In contrast, the correlation coefficients between network complexity and soil multifunctionality were almost unaffected after controlling for other properties. Moreover, the soil physical properties and Neg in B–F links were found to be the two key predictors next to network complexity. Piecewise structural equation modeling (SEM) analysis (Figure 3C and Supporting Information: Figure S15) further demonstrated that network complexity and Neg-Int directly positively and negatively regulated the soil multifunctionality, respectively. In contrast, soil biodiversity and soil properties had indirect effects through their associations with network complexity and Neg in B–F links. The response of soil multifunctionality to altered precipitation and its potential mechanism is a poorly explored ecological area [33]. Here, we explored the effects of ±50% precipitation variation with respect to the ambient conditions on soil multifunctionality in a semiarid grassland. The results indicate that decreased rather than increased precipitation had a significant impact on soil multifunctionality and biodiversity. Importantly, our study highlights that while taxonomic diversity is an important feature driving soil multifunctionality, it does so because diversity leads to greater microbiome complexity [34]. Our study provides insights into the mechanisms underlying the effects of global precipitation changes on soil multifunctionality. Bacteria and fungi are predominant decomposers in soil, whose distinct growth habits are likely to respond differently to climate changes [28, 54]. Here, bacterial and fungal richness exhibited asymmetric responses to altered precipitation. The adverse effects of decreased precipitation on microbial richness might be explained by drought as a powerful environmental filter [55, 56]. For example, drought deterministically drives the microbial community assembly and reduces its diversity by reducing the water and carbon availability [57, 58]. However, this interpretation applies only to bacteria, since the decrease in fungal richness was accompanied by an increase in stochasticity. This finding, albeit somewhat surprising, is consistent with the idea that the stochastic assembly, driven by dispersal limitation, characterizes the soil fungal communities across Scotland as well as in the forests and grasslands of southern and northern China [59-61]. The fungal migration rate varied with precipitation, in stark contrast to the invariant bacterial migration rate (Supporting Information: Figure S8C). Compared to bacteria, the dispersal of fungal spores is typically restricted to shorter distances; hence, this dispersal limitation largely shapes the fungal community structure [59, 62]. This may also partly explain the positive response of fungal richness to increased precipitation, since the fungal dispersal limitation may be weakened by increased moisture [63]. An alternative, but not mutually exclusive, explanation for the increased richness in fungal communities following increased precipitation is that fungi typically predominate over bacteria due to their special physiological features (e.g., the hyphal networks that can facilitate water transport and resource acquisition) [64]. In addition, both the bacterial and fungal habitat niche breadths decreased significantly with decreasing precipitation. Generally, a microbial group with a wider niche breadth is less affected by environmental filtration and is expected to be more metabolically flexible at the community level [63, 65]. Thus, our results suggest that decreased precipitation may also increase the vulnerability of microorganisms to environmental disturbances and the risk of diversity loss by inhibiting their The function relationships have been in recent years in both and below-ground ecosystems Here, we that greater soil biodiversity greater in soil functions in nutrient availability, microbial growth efficiency, and LOM decomposition. and have also demonstrated that diversity considering more improved ecosystem 34, 36]. This is related to the between bacteria and fungi with physiological leading to However, the increased ROM decomposition less co-occurred with decreased diversity under precipitation may be by the of microbial due to LOM This is by the analysis of organic carbon fractions and functional groups (Supporting Information: Figure S3). In our soil multifunctionality and biodiversity by considering fungi and bacteria were relatively to increased precipitation. climate features may help explain this For example, the semiarid climate may have relatively low richness through spatial capacity In this a precipitation increase may not in a significant positive effect on the because the number of has water positive have and over it is expected that the soil via increasing may have a more positive effect on the ecosystem biodiversity and function our most are in to the of microbial network properties in soil function under altered precipitation. we found a significant positive correlation between richness and multifunctionality for clusters by Actinobacteria, Proteobacteria, and while the opposite was true for clusters by taxa different thereby directly affecting different of the ecosystem function For example, Actinobacteria, Proteobacteria, and Ascomycota are more competitive for the LOM under conditions to In contrast, Acidobacteria in resource to especially for the of complex organic matter in (e.g., more show that network complexity is a and more robust affecting soil multifunctionality compared with biodiversity. This with that more complex microbial networks more to multifunctionality as ecological are not by the of these are the of by a of interactions among taxa (e.g., higher resource use efficiency and by more microbial to the network complexity, the of multifunctionality through interactions has been interactions have been reported to directly determine the strength and direction of the diversity–function may to the of bacterial communities in the of diversity cases, with the increased negative associations in response to decreased precipitation in this regarding the stability of ecosystem function under climate interactions under may to an of the effect of biodiversity on the ecosystem future studies interactions into ecosystem multifunctionality The interpretation of our results is by the following (i) Soil organisms are at (e.g., with higher Although our that including information from multiple organisms may further the of ecosystem studies have reported negative between the ecosystem function with other groups to be further explored in (ii) of an co-occurrence networks may results However, of interactions is Thus, the to network analysis to interactions In network analysis is considered a for between such as and competition Our results reveal the response and key factors of soil multifunctionality nutrient provisioning, microbial growth efficiency, and organic matter decomposition under altered precipitation in a semiarid grassland. We show that decreased precipitation exhibited a significant negative impact on microbial biodiversity, network complexity, and soil multifunctionality, while increased precipitation had a positive we found the important of microbial network properties in regulating soil multifunctionality. these results indicate that more microbial communities with networks may help to the adverse effects of climate change on ecosystem Our experimental was at the of the Loess region of the China. The is by and semiarid with an average annual and of and respectively. The average annual potential is The soil is as The region has and soil the the has a of of the important is to into abandoned grassland for The experimental of this study was on a grassland ecosystem that had been abandoned for years since Plant communities are by and The field experiment was on a largely in to the effects of climate change on soil The manipulation experiment used a in precipitation level annual precipitation) was the with warming warming and as the In the had three experimental each including were between was 3 3 in and from other in each within one a control climate change was and the were to one of the precipitation treatments and was further into one warming and one For this only the precipitation ambient precipitation and precipitation treatments were that each treatment had as biological For the multiple with an of were on a over each The of the soil surface and the by the was using from one be to the for treatment within after the of each Thus, each an of precipitation. In soil at a of were soil were from each and to a single After and were the soil samples were through a and into three was for physicochemical was at for the microbial and and the was at for Soil soil organic carbon and were to We also used permanganate oxidizable carbon (POXC) and hydrochloric acid-resistant carbon to characterize the labile and recalcitrant fractions of respectively analysis in the was used to characterize the labile and recalcitrant fractions of soil C were as in Supporting Information: Material S1 functions associated with various of soil ecosystem services were and into four including nutrient organic and and soil and microbial growth efficiency and carbon use efficiency, and LOM decomposition related to and and and ROM decomposition and the effects of in between functions, all functional were on a from to a multifunctionality index for each we each single function as well as the averaging multifunctionality index was obtained by averaging the of all ecosystem Additionally, to that each group of ecosystem functions to the multifunctionality, we also the multifunctionality index by four ecosystem function groups. we the multifunctionality index to the different of multifunctionality on the multifunctionality can be found in Supporting Information: Material S1 A of and sequencing is in Supporting Information: Material S1 the soil microbial diversity, the taxonomic were to a number of (Supporting Information: Figure S2). We used is a and used as an of soil microbial diversity We the bacterial and fungal richness on a from to 1 and the to the diversity of both groups of In addition, phylogenetic diversity was also evaluated to that the diversity selection in We quantified the effect of stochastic microbial communities using the stochasticity niche breadth was using the Moreover, the neutral community was used to further the potential importance of stochastic in the community assembly The evaluated the of a loss of an in a community by through dispersal from the metacommunity by the and neutral community were using the and respectively. estimate the network complexity of each a metacommunity co-occurrence network of the two microbial groups for all samples was on correlation using an on potential associations of from affecting data was correlation in of the samples were for We extracted the of the samples and their topological including the number of nodes and edges, average degree, clustering coefficient, average path length, graph density, and betweenness using the We analysis on the topological features to an index that the network complexity. that the average path length, denoting the network sparsity, was as the of the the of the We also extracted the proportion of negative links as well as the of links and negative links between bacteria and fungi links and Neg in B–F Moreover, we identified network clusters and the richness of soil microorganisms within each cluster across all the in networks across different precipitation regimes, we constructed ecological networks for each while the and robustness of these networks (Supporting Information: Appendix tests were to the in with different networks using the All networks were constructed using the Analysis and using the on the and of co-occurrence networks can be found in Supporting Information: Material S1 all analyses were conducted using the Spearman's correlation analysis and were used to the relationships between biodiversity and network properties with soil and were RF analysis was to determine the drivers of multiple function using the and the increase in was used to estimate the importance of the Partial correlation analysis was used to determine and how the effect of a on soil multifunctionality on other with a greater correlation between the and controlling correlation a effect of the Partial correlation analysis was conducted using the and Piecewise was further to the and indirect relationships among the soil properties and soil biodiversity, network complexity, and average multifunctionality using the The of was evaluated using information and C and the and the and Wang the and to the and and the We and for their help with This study was by the of China and the and of and the for of and in China of The no of The data that the of this study are from the The bacterial and fungal during this study are from the for under the and Supporting Information: and may be found in the or The is not for the or of information by the than be to the for the