Viromes outperform total metagenomes in revealing the spatiotemporal patterns of agricultural soil viral communities

Viruses are ubiquitous and abundant in soils and can shape microbial community assembly, dynamics, and function through infection, lysis, auxiliary metabolic genes, and horizontal gene transfer. Soils may harbor up to 10^10 viruses per gram and play roles in terrestrial carbon and nutrient cycling. Yet, soil viral communities and the factors shaping them remain poorly understood. Because viral replication depends on infecting suitable hosts, viral community structure is linked to host community composition, which in agricultural soils is influenced by rhizosphere recruitment and soil amendments such as biochar and nitrogen fertilization. Environmental factors (temperature, pH, nutrients, moisture), decay, and adsorption to soil particles may also directly shape viral communities and their spatial distributions by limiting virion movement. This study asks whether viromes outperform total metagenomes for recovering soil viral diversity and whether soil viral communities exhibit spatiotemporal patterns similar to or distinct from coexisting microbial communities across a tomato growing season with biochar and nitrogen treatments.

Given the absence of a universal viral marker gene, metagenomics is required to survey viral diversity. Many soil studies have mined viral sequences from total shotgun metagenomes, which provides convenience and concurrent microbial profiling but suffers from low viral signal amidst abundant cellular DNA and high sequence complexity that hampers assembly. Viromics—physical separation of virions by filtration prior to DNA extraction—enriches viral sequences and has been successful in aquatic systems, but soil applications have been challenging due to low DNA yields and biases introduced by amplification methods like MDA. Recent advances (low-input library kits, improved elution buffers disrupting virion–soil particle interactions, and optimized extraction workflows) now enable preparation of soil viromes from manageable soil masses. These developments motivate direct comparisons of viromes versus total metagenomes for resolving soil viral diversity and ecological patterns.

Study site and design: An agricultural tomato field in Davis, CA (38°32′08″N, 121°46′22″W) was established with three blocks and forty 6.1 m × 4.6 m plots per block. Eight plots in the western block were sampled, crossing four biochar treatments (650 °C coconut shell, 650 °C pine, 800 °C almond shell, and no biochar) and two nitrogen fertilization regimes (150 or 225 lbs N/acre). Biochar (4.2 kg per plot) was subsurface banded on November 8, 2017. Tomato seedlings (cultivar H-8504) were transplanted May 2, 2018; nitrogen fertilizer was applied via drip on five dates from May 31 to July 24, 2018. Sampling: Composite soil samples (0–30 cm depth) were collected from the same eight plots at two time points: April 23, 2018 (preplanting, prior to N additions) and August 28, 2018 (ripening; rhizosphere-influenced). Each composite comprised eight 2.5-cm cores along transects adjacent to drip tape. Samples were sieved (8 mm) and partitioned for chemistry, moisture, viromics, and total metagenomics. Soil chemistry: Gravimetric moisture; total C and N (combustion); extractable NH4+ and NO3− (2 M KCl, FIA); pH (1:1 soil:water); organic matter (loss on ignition); P (weak Bray and bicarbonate-P); extractable cations (K, Mg, Ca, Na). Virome preparation: From 50 g soil, virus-like particles were resuspended in amended potassium citrate prime (AKC′) buffer, filtered through 0.22 µm to remove cells, concentrated by ultracentrifugation, treated with DNase to remove free DNA, then virions lysed and DNA extracted with the PowerSoil kit. Total metagenome DNA: Extracted from 0.5 g soil using the PowerSoil kit. Library prep and sequencing: April libraries used KAPA Hyper Prep; August libraries used Nextera DNA Flex. Illumina HiSeq 4000, 150 bp paired-end. April: viromes and total metagenomes pooled equimolarly; August: pooled at 1:2 (virome:total) to increase total metagenome throughput. Bioinformatics: Reads quality-filtered with Trimmomatic and BBDUK. Assemblies with MEGAHIT; clustering with PSI-CD-HIT (0.95 global identity). Viral detection with VirSorter and DeepVirFinder; taxonomic classification with vConTACT2. Read mapping with BBMap; coverage and abundance tables with BamM. vOTU definition: contigs ≥10 Kbp, clustered at ≥95% global identity; detection thresholds: ≥75% contig covered ≥1× by reads at ≥90% ANI. 16S rRNA fragments detected with SortMeRNA and classified with RDP. K-mer profiling via sourmash (k=31). Statistics in R using vegan and DESeq2. Data deposition: SRA BioProject PRJNA646773. One August virome (CS-H) with very low throughput and vOTU recovery was excluded from downstream analyses.

• Sequencing throughput: Total metagenomes averaged 8,741,015 paired reads (April) and 14,551,631 (August); viromes averaged 9,519,518 (April) and 5,770,419 (August).
• Cellular DNA depletion: Only 0.006% of virome reads versus 0.042% of total metagenome reads were classified as 16S rRNA fragments, indicating substantial depletion of bacterial/archaeal DNA in viromes.
• Microbial taxa differences in retained 16S reads: Total metagenomes enriched for Acidobacteria, Actinobacteria, Firmicutes, Thaumarchaeota; viromes enriched for Armatimonadetes, Saccharibacteria, Parcubacteria (small-celled CPR taxa more likely to pass 0.22 µm).
• Lower sequence complexity in viromes: K-mer spectra showed fewer singletons and higher k-mer occurrences in viromes, facilitating assembly.
• Assembly outcomes: Viromes yielded ~800 Mbp across 169,421 contigs (250 Mbp in ≥10 Kbp contigs), versus total metagenomes ~65 Mbp across 22,951 contigs (1.5 Mbp in ≥10 Kbp contigs).
• Viral enrichment of contigs: 52.4% of virome contigs vs 2.2% of total metagenome contigs identified as viral by VirSorter/DeepVirFinder.
• vOTU catalog: 4,065 vOTUs (median length 17,870 bp; max 259,025 bp; median 27 ORFs). Mapping recovered 2,961 vOTUs in at least one sample.
• Mapping rates: On average, 23.4% of virome reads vs 0.04% of total metagenome reads mapped to vOTUs.
• Recovery by approach: Of 2,961 detected vOTUs, 2,864 were exclusive to viromes, 94 were shared, and only 3 were exclusive to total metagenomes (these three occurred in an April metagenome lacking a successful paired virome).
• Diversity capture: Virome-based accumulation curves approached richness asymptotes; total metagenomes did not, indicating more complete viral diversity recovery by viromes.
• Abundance–occupancy: In viromes, abundant vOTUs had high occupancy; rare vOTUs were found in few samples. Over 30% of vOTUs were found in all plots, suggesting a core field virosphere. In total metagenomes, >80% of vOTUs were singletons, indicating sparse viral recovery.
• Ecological patterns: Both viral and microbial communities differed significantly over time (preplanting vs ripening), suggesting coupled responses to rhizosphere recruitment and/or nitrogen amendments. Viral communities also exhibited spatial structuring along an ~18 m gradient across the field.
• Total metagenomes predominantly captured the most abundant and ubiquitous vOTUs, missing much of the rare virosphere.

The study demonstrates that size-fractionated viromes substantially enrich viral sequences and reduce cellular DNA and sequence complexity, yielding superior assemblies and enabling comprehensive profiling of soil viral diversity compared to total metagenomes. The markedly higher proportion of viral contigs, greater mapping rates to vOTUs, and the recovery of nearly all detected viral populations from viromes show that total metagenomes, at typical sequencing depths, largely capture only the most persistent and abundant viruses and miss the rare virosphere. Ecologically, both viral and microbial communities shift over the growing season, indicating that viral dynamics likely respond to host community changes driven by rhizosphere processes and nitrogen fertilization. However, viral communities also display clear spatial structure along an 18 m gradient, whereas microbial communities may not mirror these spatial patterns to the same extent, implying a partial decoupling of spatial drivers between viruses and hosts. This decoupling could arise from differences in dispersal constraints (e.g., adsorption to soil particles), decay rates, or sensitivities to microenvironmental heterogeneity. Overall, the findings address the central questions by validating viromics as the preferred method for soil viral community profiling and by revealing distinct temporal and spatial patterns in agricultural soil viromes.

Viromics outperforms total metagenomics for characterizing dsDNA viral communities in agricultural soils, enabling recovery of far greater viral richness, broader taxonomic and host-range diversity, and access to the rare virosphere. Across a tomato growing season, viral and microbial communities exhibit significant temporal shifts, while viral communities also show pronounced spatial structuring at the field scale, suggesting both shared and distinct ecological drivers relative to microbial hosts. These results underscore the utility of soil viromics for uncovering viral ecology and biogeography. Future work should disentangle mechanisms underlying spatial decoupling of viruses and hosts, including quantifying viral dispersal and decay in heterogeneous soils, refining virome extraction to minimize biases, and integrating host–virus linkages to connect viral dynamics to ecosystem functions.

• Library preparation differed between time points (April: KAPA Hyper Prep; August: Nextera DNA Flex), which could introduce methodological batch effects in sequence composition and assembly independent of biological change.
• One August virome sample had low sequencing throughput and was excluded, reducing replication.
• Size-fractionation may allow very small cells (e.g., CPR taxa) to pass 0.22 µm filters, contributing residual cellular DNA to viromes.
• Potential biases in virome preparation, such as differential adsorption of viruses to soil particles and variable recovery during resuspension, could affect relative abundance estimates.
• Total metagenomes had different pooling ratios (August 1:2 virome:total), and lower virome throughput in August may affect comparative sensitivity, although viromes still outperformed.
• 16S rRNA gene copy-number variation could influence comparisons of cellular contamination levels between approaches.