Prolonged grazing reduces the diversity and weakens virus-host links of metagenome-assembled viral community in acidic karst soil Authors: Ran Xue, Zhou Li, Wangfei Qin, Yiling Wang, Kankan Zhao, Lin Liu, Yanfu Bai Published: 2025 ABSTRACT AbstractBackgroundSoil viruses are fundamental yet often overlooked components of terrestrial ecosystems, where they profoundly influence microbial diversity, community assembly, and biogeochemical cycling through predator–prey dynamics and the carriage of auxiliary metabolic genes (AMGs). However, the impact of human-driven land use change on viral communities and their functional potential remains poorly understood.ResultsIn this study, we analyzed viral diversity, virus-host linkage, and functional profiles of AMGs across different land management regimes in acidic karst soil after reclamation. The results showed that both bacterial and viral communities significantly differ across the four land management regimes, which are unused land, grazing pastureland, abandoned pastureland, and fertilized grazing pastureland. Grazing led to a lower viral diversity, a decrease in lysogenic viral abundance, and fewer virus-host linkages. The diversity and function of AMGs were also significantly affected by land management regimes, with grazing leading to a reduction in both AMGs diversity and relative abundance. Furthermore, both the bacterial community and soil physiochemical properties were essential factors that shaped viral diversity and lifestyle.ConclusionsOverall, our findings indicate that inappropriate land use (long-term grazing) disrupts the virus-host balance, thereby altering the functional profiles of the soil viral community.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12915-025-02489-0.Keywords: Grazing, Karst, Virus-host linkage, Viral lifestyle Background Viruses are highly abundant and diverse, playing pivotal roles in shaping microbial structure, genetic exchange, and nutrient cycling [20, 39]. While most previous studies focused on viral diversity and composition in marine or intestinal ecosystems, the study of viral ecology in soil ecosystems has been limited [5, 8, 12, 34]. Understanding the dynamics of soil viruses and their interactions with hosts under environmental stress is essential for gaining insights into biogeochemical cycling and biochemical processes within soil ecosystems [11]. As research into soil viral ecology has accelerated, distinct macroecological patterns have started to emerge. Notably, recent studies have highlighted the biogeographic characteristics of soil viruses across ecosystems and continents [11, 31]. It is supposed that viral diversity in soil is much higher than in other ecosystems due to greater heterogeneity and host diversity [20]. Although soil viral diversity varies significantly across different samples and biomes, large-scale studies have shown that common macroecological patterns also apply to viral communities. Significant distance-decay relationships have been observed in viral communities across agricultural lands, forests, grasslands, and wetlands within continents [31]. While host diversity and structure are the primary factors shaping viral communities, environmental stress gradients such as pH and moisture are also strongly correlated with virus community at global scales [24, 56]. These findings provide broad insights into the turnover of virus community in response to spatial and environmental heterogeneity. However, fewer studies have explored the impact of human disturbance on viral communities at local scales. Soil viruses play crucial roles in soil biogeochemical cycling by releasing bacterial cellular materials through host lysis [55] or by altering host metabolic pathways via auxiliary metabolic genes (AMGs) [19, 25]. It has been proven that viruses contribute significantly to carbon and nitrogen losses through their impact on hosts in both alpine soil and sediment [1, 4, 18]. Virus-carried AMGs are also linked to nutrient transformation, the degradation of organic pollutants, and microplastic breakdown [50, 58]. Environmental stresses are key factors that affect both viral lifestyle and AMG diversity. Environmental stress disrupts the connection between viruses and their hosts and leads soil viruses toward the lysogenic lifestyle [16]. AMGs are also supposed to play more important roles in nutrient transformation, heavy metal resistance, organic pollutant degradation, and antibiotic resistance under stress [15, 16, 46, 47]. While these researches primarily aimed to illustrate the ecological significance of soil viruses or compare viral communities across different ecosystems, the impact of land use change on the viral functional dynamics of soil viral communities remains an area that requires further investigation. One-third of the world’s land use has been altered by humans over the past six decades [53]. Land use change is a key driver of environmental transformation, influencing various ecological components. Extensive research has elucidated that land use changes significantly impact bacterial structure and function by altering the heterogeneity of soil factors such as pH, moisture, and nutrient availability [24, 33, 42, 49]. Since viruses rely on hosts for replication, viral community diversity and structure are likely to be changed across different land management regimes. Moreover, changes in soil nutrient conditions and physical properties can alter viral lifestyles (lytic or lysogenic), which in turn affect host community turnover, metabolic potential, and ecosystem functions [18]. Therefore, understanding how soil virus communities respond to land use change is essential for effective ecosystem management and sustainable land use practices. Therefore, understanding how soil virus communities and their functional gene repertoire respond to land use change is essential for effective ecosystem management. However, an integrated view of how land use alters viral community structure, viral-host interactions, and associated functional potentials (AMGs) remains elusive. Herein, we collected soil samples from four soil use patterns (unused land, grazing pastureland, abandoned pastureland, and fertilized grazing pastureland) in a karst region. Using metagenomic sequencing, we aimed to investigate the changes in soil viral diversity, composition, and viral-host link across these land use types. Additionally, we explored potential linkages between shifts in soil viral community dynamics and environmental parameters influenced by land use change. We hypothesize that (1) prolonged grazing reduces viral diversity and virus-host linkage and (2) abandoned pastureland increases the proportion of lysogenic virus and AMG diversity. These hypotheses are based on the results of our previous study that soil nutrient content was higher in grazing treatments and lower in abandoned pastureland [37]. Results Bacterial diversity and structure across the four land management regimes We analyzed the relative abundance of different bacterial taxa across the four treatments to assess the impact of soil management on bacterial community. In total, we identified 10,936 bacterial taxa at species level, with 9085 shared across all four treatments (Additional file 1: Fig. S1). Pseudomonadota and Actinomycetota dominated in all treatments, accounting for over 85% of all total bacterial abundance. Pseudomonadota had a higher relative abundance in GPL and APL, while Actinomycetota was more abundant in UL and FGPL. When comparing bacterial alpha diversity, APL had lower richness and higher Shannon index compared to UL and FGPL (Fig. 1a, b). In contrast, FGPL had higher richness and lower Shannon index than GPL and APL. Bacterial community structures were clearly differentiated among the different treatments, forming distinct clusters in two-dimensional ordinations (Fig. 1c, Adonis, p < 0.001). A total of 1042 taxa were selected as indicator taxa for the four treatments (LDA score > 2.0). Generally, Pseudomonadota was more abundant in UL; Acidobacteriota and Actinomycetota were more abundant in APL; Alphaproteobacteriota was more abundant in FGPL; and Rhodanobacter was more abundant in GPL (Fig. 1d). Bacterial diversity and composition across the four land management regimes (UL, unused land; GPL, grazing pastureland; APL, abandoned grazing pastureland; and FGPL, fertilized grazing pastureland). a Bacterial community richness of the four land use treatments. Different letters indicate significant differences (one-way ANOVA, LSD test, p < 0.05). b Shannon index of bacterial community in the four and use patterns. Different letters indicate significant differences (one-way ANOVA, LSD test, p < 0.05). c Principal coordinate analysis of the bacterial community composition based on Bray Curtis dissimilarity (Adonis, p < 0.001). d LEfSe analysis of bacterial abundance across the four land management regimes Viral diversity and structure across the four land management regimes A total of 60,370 contigs (500 bp) were identified as viral contigs, which were filtered (> 5 kb) and clustered into 2831 vOTUs. Generally, CheckV quality improved with increasing sequence length (Additional file 1: Fig. S2a). More than 82% (2331) of the vOTUs were unique to a single treatment, with only 37 vOTUs shared across all four treatments (Additional file 1: Fig. S2b). A large portion (2463) of the vOTUs were unassigned, accounting for more than 85% of the total viral abundance (Fig. 2a). The majority of the assigned vOTUs were classified as Caudoviricetes (332), which made up more than 10% of the total viral abundance across all treatments (Fig. 2b). In terms of viral alpha diversity, APL had significantly higher richness and Shannon index compared to FGPL (Fig. 2c, d). Similar to bacterial communities, viral communities also exhibited distinct structural differences across the four treatments (Fig. 2e, Adonis, p < 0.001). Viral diversity and composition across the four land management regimes (UL, unused land; GPL, grazing pastureland; APL, abandoned grazing pastureland; and FGPL, fertilized grazing pastureland). a Classification of the total viral population at the family level. b Relative abundance of viral population in the four land management regimes at the family level. c Viral community richness in the four and use patterns. Different letters indicate significant differences (one-way ANOVA, LSD test, p < 0.05). d Shannon index of the viral community in the four and use patterns. Asterisks indicate significant differences (Dunn’s test, p < 0.05). e Principal coordinate analysis of viral community composition based on Bray Curtis dissimilarity (Adonis, p < 0.001) Response of virus-host links, viral lifestyle and AMGs to the four land management regimes In total, 145 vOTUs and 330 hosts were identified at the genus level across all the treatments, with 131, 90, 241, and 39 links identified in UL, GPL, APL, and FGPL, respectively (Fig. 3a, b). The identified viral hosts belonged to 24 bacterial phyla, most of which were assigned to Acidobacteriota and Pseudomonadota at the phylum level. The resulting virus-host bipartite network showed 12 virus hubs (3, 3, 4, and 2 in UL, GPL, APL, and FGPL, respectively) and 4 host hubs (1, 1, and 2 in UL, GPL, and APL, respectively) that linked to > 10 hosts or viruses (Additional file 1: Fig. S3). Koribacteraceae was the most linked host at the family level, susceptible to infection by as many as 92 viruses. Additionally, an unassigned vOTU (GPL_k141_1125439) was identified as a generalist, capable of infecting more than 20 bacterial species. Significant differences in the relative abundance of putative lysogenic phages were observed among the treatments, with UL and APL showing significantly higher relative abundances of lysogenic vOTUs compared to GPL and FGPL (Fig. 3c). Virus-host interactions across the four land management regimes (UL, unused land; GPL, grazing pastureland; APL, abandoned grazing pastureland; and FGPL, fertilized grazing pastureland). a Predicted phage-host associations. Left: viral contigs matching hosts at the family level. Right: the distribution of hosts at the phylum level. b Number of virus-host links in the four land management regimes. c Relative abundance of lysogentic viruses in the four land management regimes. Different letters indicate significant differences (one-way ANOVA, LSD test, p < 0.05) To assess the potential contribution of viruses to microbial community functioning, we compared the diversity and functional annotation of AMGs across the four treatments. A total of 254 putative AMGs were annotated, of which 49 were annotated and assigned to 24 KOs spanning 8 metabolism pathways. The highest number of AMGs was identified in APL, while the fewest were identified in FGPL (Fig. 4a, b). Most of the AMGs were unique to each treatment, with only 5 AMGs being ubiquitous across all four treatments (Fig. 4c). The vOTUs primarily encoded AMGs related to amino acid metabolism, followed by carbohydrate metabolism and energy metabolism. When comparing the relative abundance of each metabolism pathway, we found that carbohydrate metabolism pathway was more abundant in UL and FGPL, while the lipid metabolism pathway was more abundant in APL and GPL (Fig. 4d). Functional annotation of AMGs across the four land management regimes (UL, unused land; GPL, grazing pastureland; APL, abandoned grazing pastureland; and FGPL, fertilized grazing pastureland). a The number of AMGs in the four land management regimes. b Relative abundance of AMGs in the four land management regimes. Asterisks indicate significant differences (Dunn’s test, p < 0.05). c Venn diagram showing the overlap of annotated AMGs in the four land management regimes. d Heatmap showing the relative abundance of functional genes annotated by KEGG in the four land management regimes Factors associated with bacterial and viral communities across the four land management regimes We first correlated bacterial and viral diversity to check the relationships between bacterial and viral turnover. The results showed that virus richness decreased as bacterial richness increased (Pearson, p < 0.01, Fig. 5a). Furthermore, the Mantel test results revealed a significant correlation between the dissimilarities of bacterial and viral communities (Fig. 5b). When evaluating the effect of environmental variables on viral communities and viral lifestyle, we found that viral diversity was positively correlated with BD, WHC, AP, K, AK, TC, SIC, and SOC, while negatively correlated with N and NO3−. The abundance of lysogenic viruses showed strongest positive correlations with soil nutrient factors, including NO3−, AP, K, TC, and SOC (Fig. 5c). Biotic and abiotic factors associated with viral divesity and lifestyle. a Correlation between bacterial and viral richness. b Correlation between bacterial and viral community dissimilarity. c Spearman correlation for environmental factors; Mantel test results for the relationship between environmental factors with viral divesity and lifestyle Discussion Most previous studies have shown that land use change significantly affects both bacterial and viral diversity due to the differences in organic matter supplied by plants and alterations in soil physiochemical properties (Muñoz‐Arenas et al., 2020). In contrast, some studies have revealed that land use change has little effect on bacterial diversity. For example, the conversion of shrubland to grassland did not affect bacterial diversity when the dominant grass species were similar [14]. Our findings revealed significant differences in both bacterial and viral structure across different land management regimes. For bacterial communities, APL exhibited a higher Shannon index but lower richness. This discrepancy suggests that the bacterial community in APL was dominated by a few species, with the remaining species present at relatively low but evenly distributed abundances [51]. For the viral community, we found that only a few vOTUs were ubiquitous across all four treatments, indicating a strong effect of land use changes. The majority of vOTUs remained unclassified in all treatments. This lack of taxonomic annotation and the high variation in viral composition aligned with the findings of most other soil virus studies, which suggest that such variability can be attributed to the highly localized adaptation, rapid coevolution of viruses with their hosts, and high spatial heterogeneity in different soil management systems [26, 31]. As obligate intracellular parasites, viral community composition is strongly influenced by their hosts [40, 41]. On the other hand, viruses also affect the taxonomic and functional composition of bacterial communities through predation (cell lysis) and coevolutionary interactions such as horizontal gene transfer [6, 23]. Virus-host links have been found closely related to environmental stresses, with soil pH, moisture, and nutrient conditions being the major determinants [9, 42, 48]. Although most of the karst area is known for its alkaline soil, our sampling site had a pH of less than 5.5, likely due to the dissolution of calcium carbonate caused by persistent subtropical monsoon rainfall [37]. Therefore, it is not surprising that Acidobacteriota emerged as the primary host, in contrast to Pseudomonadota when compared with other studies [26, 58]. Further, we found that viruses were associated with a higher number of bacterial hosts in APL, which we speculate may be due to the higher viral diversity in this treatment. When examining the viral life style in the four treatments, we found a lower proportion of lytic viruses in GPL and FGPL. Soil particles form a vast network of microhabitats and pores of varying sizes, which act as a barrier that isolates microbial cells into small and fragmented populations. This intact soil (typically with high bulk density) reduces the encounter rate between viruses and their hosts. Under these conditions, the lysogenic strategy offers a survival advantage for viruses. In contrast, anthropogenic activities such as grazing and fertilization not only increase soil nutrient availability but also disrupt soil aggregates, mixing viruses and hosts into a more homogeneous environment [10]. These conditions favor rapid viral replication, promoting a shift toward the lytic strategy. Thus, the higher relative abundance of lysogenic viruses in APL and UL could also be a result of the fewer anthropogenic activities in these treatments. It is believed that AMGs can influence host metabolism and ecosystem function, thereby enhancing the fitness of both hosts and viruses [30, 59]. Functional annotations of AMGs in our study showed that the viral community is equipped with genes from a number of cellular metabolic pathways. Viral-encoded AMGs are crucial for compensating host metabolism under environmental stress such as nutrient deficiency and heavy metal pollution [17, 18]. In our study, the viral community in APL harbored more AMGs, indicating poor nutrient conditions in abandoned pasture soil. This is further supported by our previous study, which found that both soil N and organic C contents were lowest in APL compared to other treatments [37]. The AMGs encoded for xenobiotic biodegradation and metabolism pathways were exclusively found in FGPL; this could be because higher nutrient content may promote the growth and reproduction of certain microorganisms capable of degradation. Moreover, long-term fertilization may also drive adaptive evolution in soil microbes, developing more efficient metabolic pathways for xenobiotics [27]. Viruses have important ecological impacts on soil function through regulating host populations [52]. In this study, we found that viruses were more sensitive to land use change than bacteria, as the dissimilarity in viral communities was higher than that in bacterial communities. This could be due to the higher diversity and spatial heterogeneity of viruses compared to bacteria in soil. In fact, because vOTUs recovered from total metagenomes are more likely to miss the rare virosphere and underestimate most of the viral diversity, viral turnover among the four treatments may still be underestimated in our study [43]. Our analysis of the correlation between bacterial and viral diversity revealed that viral richness was negatively correlated with bacterial richness. These results support the Kill-the-Winner (KtW) theory, which suggests that the lysis of dominant hosts by viruses creates more ecological niches for the rare biosphere, thereby increasing host diversity [54]. Conversely, the dissimilarity of the bacterial community was positively correlated with the dissimilarity of the viral community, indicating that similar bacterial communities tend to harbor similar viral communities. Recent studies have shown that changes in environmental factors, such as shifts in moisture, temperature, and soil nutrient condition, influence viral diversity in soil [7, 9, 57]. Soil physical properties can directly affect viral communities through altering the viral lifestyles or indirectly through influencing host communities [48]. Viruses tend to be more lysogenic in dry soils because of the isolation of host-virus interactions and reduced host activity under water stress [57]. In our study, both viral diversity and lifestyle were correlated with soil water holding capacity. Although viral diversity was also correlated with most nutrient indices, there was no evidence suggesting that soil nutrient conditions can directly affect viral communities. Therefore, we attribute these results to the indirect effects of host community dynamics and soil physiochemical properties, both of which are strongly influenced by land management regimes. While total metagenomes assembled, the viral population is likely to underestimate viral taxonomic diversity, host range, and the rare virosphere, future studies using viral-metagenome (viromes) could provide deeper insight into the interactions and co-evolution of viruses and their host as enrichment steps specifically target viral particles, increasing the proportion of viral DNA/RNA in the sample. Conclusions Our findings provide an initial view of viral-host relationships and the dynamics of soil viral diversity under different land management regimes. Grazing decreased viral diversity, virus-host linkages, and the proportion of lysogenic virus. Abandoned pastureland increased viral diversity, virus-host linkages, and AMG’s abundance. In addition to the host community, the dynamics of soil properties, including bulk density, moisture, and nutrient condition associated with land management regimes, were key factors determining viral diversity. These results are critical for understanding the dynamics of soil viruses in complex ecosystems. The insights from this study may offer valuable implications for maintaining soil health, conserving biodiversity, and supporting ecosystem functioning amidst rapid land use changes. Methods Study area and experimental design The experimental area was located at Sequoia River Ranch in Dafang County, Bijie City, Guizhou Province, China (27°23′N, 105°52′E; altitude 1760 m). This area features a humid subtropical monsoon climate and is representative of the Guizhou Karst region. During the experimental period, the average annual temperature was 12 ℃, and the average annual precipitation was 1064 mm. The dominant soil type is Ali-Perudic Argosols (APA) according to the International Society of Soil Sciences standards. The native vegetation mainly consists of shrubs (Salix wallichiana, Rosa omeiensis, Rubus fockeanus), and warm-season grasses (Macrothelypteris oligophlebia, Dipsacus asperoides, Arthraxon hispidus). The experimental area, covering a total area of 80 ha, was established in late 2012 following prescribed burning and rotary mowing. A rotational pasture system was then implemented using a mix of red clover (Trifolium pratense), white clover (Trifolium repens), duck grass (Dactylis glomerata), and perennial ryegrass (Lolium perenne). A compound fertilizer (N: P2O5: K2O, 15:15:15) was applied at a rate of 150 kg/ha per year. In September 2015, four land management regimes were established: (i) unused land (UL) with no burning and rotary mowing anthropogenic intervention throughout the study period; (ii) Grazing Pastureland (GPL) with a rotational grazing system without external fertilizer inputs from 2012 to 2021; (iii) Abandoned Pastureland (APL) with no grazing activities from 2015 to 2021; (iv) Fertilized Grazing Pastureland (FGPL), a rotational grazing system with annual fertilizer inputs of N: P2O5: K2O (15:15:15) at 300 kg/ha, split evenly in March and September from 2012 to 2021. The study followed a randomized complete block design with repeated measures. Each block included four treatments (unused land, grazing pastureland, abandoned pastureland, and fertilized grazing pastureland), resulting in a total of 12 sampling plots (4 treatments with 3 replicates, about 7 ha per plot). Analysis of soil chemical and physical properties In August 2021, five topsoil samples (0–15 cm) were collected from each plot. All samples were collected within a single day to ensure consistent seasonal and weather conditions. Samples were taken at least 5 m away from the fence, with a minimum interval of 10 m between sampling points. After removing roots and stones, the soil samples were combined and stored at −80 °C until further processing. A portion of the soil sample was sieved through a 2 mm mesh and air-dried for subsequent analysis of physical and chemical properties. The physicochemical properties measured included soil pH, bulk density (BD), water holding capacity (WHC), inorganic nitrogen (NO3−, NH4+), total carbon (TC), soil inorganic/organic carbon (SIC/SOC), total potassium (K), available potassium (AK), total phosphorus (P), and available phosphorus (AP) as described in our previous study Qin et al. [37]. Metagenomic sequencing Soil microbial DNA was extracted using the FastDNA SPIN kit (MP Biomedicals, Solon, OH, USA) following the manufacturer’s protocol. The extracted DNA samples were then subjected to quality control using agarose gel electrophoresis and a Qubit®2.0 Flurometer. Paired-end sequencing was performed with a sequencing depth of 20 G using the Illumina NovaSeq PE150 platform (Illumina, San Diego, CA, USA). Raw reads were quality controlled using Trimmomatic (v.2.39) to eliminate adaptors and primers, as well as to remove short (< 50 bp) and low-quality reads (Phred score < 20). The filtered reads from each treatment were assembled into contigs using Megahit (version 1.1.2) through a co-assembly strategy with a minimum contig length of 500 bp (k-step = 10; k-min = 27). On average, 3,540,133 contigs were yielded for each treatment (pooled data of 3 replicates). Bacterial taxonomic assignment and virus identification Bacterial classification was performed using Kraken2, and the relative abundance of detected taxa in each sample was estimated using Bracken [29]. Phage contigs (dsDNA and ssDNA phage) in the assembled contigs were identified using DeepVirFinder and Virsorter2 with default parameters [13, 38]. The quality of the results was assessed with CheckV, applying a minimal score cutoff of 0.5. The remaining viral contigs were filtered (> 5 kb) and clustered into vOTUs based on 95% average nucleotide identity and ≥ 80% coverage of the shortest sequence using CD-HIT [32]. The longest sequence in each cluster was selected as the representative sequence of the cluster. Viral taxonomy annotations were assigned using vConTACT (v 3.0) by applying a majority-rules approach as previously described [3]. A viral population was assigned to a viral family if > 50% of its proteins matched to the family in the Viral RefSeq Virus database, using a BLASTp bitscore ≥ 50. To estimate viral relative abundance, paired-end reads from each sample were mapped to the viral contigs using BWA with default parameters. Subsequently, CoverM (https://github.com/wwood/CoverM) was employed to eliminate reads that aligned for less than 90% of their length and with < 95% nucleic identity. The filtered bam files were then processed using SAMtools to determine the coverage of positions by reads. Virus lifestyle and virus-host linkage analysis Putative lysogenic vOTUs were identified by the presence of integrases or by being identified as prophage through Checkv and VIBRANT using default parameters [22]. Three methods were used to analyze putative virus-host linkages: (1) Host and virus tRNA genes were predicted using tRNAscan-SE v1.3.1 [28]. tRNAs matching promiscuous tRNAs from the Earth virome dataset were removed from the dataset [36]. General and bacterial/archeal models were used to explore tRNA genes (-G/-A/-B), using BLAST (100% coverage and 100% sequence identity) after deleting self-hits and duplicates; 2) Host CRISPR spacers were sorted using MinCED v0.4.2 (-minNR 2) [2] from host genomes. BLASTn was used to determine alignment between viral genomes and CRISPR spacers (100% nucleotide identity, mismatch ≤ 1 and E-value ≤ 10−5; 3 Integrated prophages were identified by comparing vOTUs against host sequences using BLASTn. Aligned regions needed to be at least 2500 bp with ≥ 90% identity. A bipartite network was constructed on high and perfect score results from the three methods using Cytoscape (v.3.8.0, with edges presenting virus–host linkages. Virus-encoded auxiliary metabolic genes (AMGs were annotated using DRAM-v (contig length > 1 kb based on the virus cluster vOTU sequence [45]. In this study, AMG abundance was calculated as the sum of the abundance of AMG- derived viral contigs as suggested by Jian et al. [21]. Data analysis All statistical analyses were conducted using R (v4.0.2) (R Core Team, 2020). Bacterial and viral diversity were assessed with Richness and Shannon index in vegan [35]. The assumptions of normality and homogeneity of variance were assessed using the Shapiro–Wilk normality test and Bartlett test prior to the main analysis. The data met the assumptions of normality and homogeneity of variance (bacterial Richness and Shannon index, viral Richness and percentage of lysogenic virus) were compared with one-way ANOVA. The data failed to meet the assumptions of normality and homogeneity of variance (viral Shannon index and relative abundance of AMGs) were compared with Dunn’s test. Bray–Curtis dissimilarity matrices were used for principal coordinate analysis (PCoA) and Adonis tests (999 permutations) were performed to compare among-group differences of the four treatments. Linear discriminant analysis effect size (LEfSe) was used with the Kruskal–Wallis rank sum test to identify taxa with higher relative abundance in specific treatments [44]. We defined “indicator taxa” as those OTUs with linear discriminant analysis (LDA) scores higher than 2 in each treatment. The correlations between viral diversity/relative abundance of lysogenic virus with environmental parameters were analyzed using ggcor R package v0.9.8.1 (https://github.com/houyunHuang/ggcor). Supplementary Information Additional file 1. Figures S1-S3.FigS2- Viral community information. a Distribution of viral sequences quality and length. b Venn diagram showing the overlap of viral richness in the four land use patterns (UL: Unused land, GPL: Grazing pastureland, APL: Abandoned grazing pastureland, and FGPL: Fertilized grazing pastureland)FigS2- Viral community information. a Distribution of viral sequences quality and length. b Venn diagram showing the overlap of viral richness in the four land use patterns (UL: Unused land, GPL: Grazing pastureland, APL: Abandoned grazing pastureland, and FGPL: Fertilized grazing pastureland)FigS2- Viral community information. a Distribution of viral sequences quality and length. b Venn diagram showing the overlap of viral richness in the four land use patterns (UL: Unused land, GPL: Grazing pastureland, APL: Abandoned grazing pastureland, and FGPL: Fertilized grazing pastureland). Acknowledgements We would like to acknowledge Shengjiang He (Farm Manager, Oil fir River Ranch) for providing support for establishing and maintaining the research sites. We are grateful to Lili Niu, Jie Pan, Qishun Mo, Guang Xu for their contributions to the field data collection. Authors’ contributions RX: Conceptualization, Formal analysis, Visualization, Writing-original draft. ZL: Conceptualization, Investigation, Writing-review & editing; WFQ: Investigation, Writing-review & editing. YLW: Methodology, Formal analysis, Writing-review & editing. KKZ: Methodology, Formal analysis, Writing-review & editing. LL: Writing-review & editing. YFB: Writing-review & editing. All authors read and approved the final manuscript. Funding This research was financed by the National Natural Science Foundation of China (41907025, 32160337), Natural Science Foundation of Sichuan Province (2025ZNSFSC0202). Data availability Raw sequence reads are deposited in the SRA (BioProject: PRJNA1185441). Declarations Ethics approval and consent to participate This article does not contain any studies with human participants or animals performed by any of the authors. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Additional file 1. Figures S1-S3.FigS2- Viral community information. a Distribution of viral sequences quality and length. b Venn diagram showing the overlap of viral richness in the four land use patterns (UL: Unused land, GPL: Grazing pastureland, APL: Abandoned grazing pastureland, and FGPL: Fertilized grazing pastureland)FigS2- Viral community information. a Distribution of viral sequences quality and length. b Venn diagram showing the overlap of viral richness in the four land use patterns (UL: Unused land, GPL: Grazing pastureland, APL: Abandoned grazing pastureland, and FGPL: Fertilized grazing pastureland)FigS2- Viral community information. a Distribution of viral sequences quality and length. b Venn diagram showing the overlap of viral richness in the four land use patterns (UL: Unused land, GPL: Grazing pastureland, APL: Abandoned grazing pastureland, and FGPL: Fertilized grazing pastureland). Data Availability Statement Raw sequence reads are deposited in the SRA (BioProject: PRJNA1185441).