Mesophilic and thermophilic viruses are associated with nutrient cycling during hyperthermophilic composting

The study investigates how mesophilic and thermophilic bacteria and their viruses jointly drive nutrient cycling during hyperthermophilic composting (HTC). Although bacterial and fungal roles in organic matter decomposition are established, viral contributions in terrestrial systems remain unclear compared to oceans. Viruses can influence nutrient cycling via top-down control through host lysis and by encoding auxiliary metabolic genes (AMGs) that augment host metabolism, potentially affecting carbon and nitrogen turnover and microbial succession. HTC provides a model with distinct temperature-driven phases (hyperthermophilic, thermophilic, maturation) and dynamic microbial communities, allowing assessment of virus-bacteria coupling and functional impacts on nutrient cycling. The research aims to quantify viral and bacterial diversity, abundance, activity, virus-host linkages, and viral AMGs over time, and to relate these to carbon and nitrogen cycling during industrial-scale HTC.

Prior work shows viruses are abundant and regulate microbial mortality, composition, and biogeochemical cycles, with well-established roles in oceans and emerging evidence in soils. Viral AMGs, including glycoside hydrolases and genes affecting methane metabolism, indicate potential participation in complex carbon degradation in soils. Viruses can also enhance host survival under stress via AMGs. Ecological models such as Kill-the-Winner explain viral top-down control of dominant bacteria. Transcriptomics has revealed active lytic and lysogenic cycles and AMG expression in marine and subsurface systems, and giant viruses active in coastal systems. However, the contributions of viruses to nutrient cycling and organic matter mineralization in terrestrial systems, particularly across temperature-driven successions, remain underexplored.

Design: An industrial-scale HTC plant in Beijing, China processed sewage sludge and rice husk over 45 days encompassing initial (D0), hyperthermophilic (>90 °C; D2–D9), thermophilic (>55 °C; D10–D26), and maturation (<45 °C; D27–D45) phases. Eight time points were sampled from five piles (D0, D4, D7, D9, D15, D21, D27, D33, D45). Composite samples were collected at 40–50 cm depth and split for biological (liquid N2) and physicochemical (4 °C) analyses; temperature was continuously monitored. Physicochemical analyses: Carbon cycling metrics (TC, TOC, IC, WSC) and nitrogen cycling metrics (TN, WSN, NH4+, NO3−) were quantified (TOC analyzer; elemental analyzer for TC/TN; OM by loss-on-ignition at 550 °C; pH, water content, EC). C/N ratio and WSC/WSN derived from component measurements. Amplicon sequencing: 16S rRNA V4–V5 amplicons (515F/907R) were sequenced (Illumina NovaSeq PE250). QIIME2 with DADA2 generated ASVs; taxonomy assigned via SILVA v138. Diversity (alpha: Shannon, observed OTUs; beta: weighted UniFrac) and null model analyses (βNTI) assessed community assembly. Shotgun metagenomics and metatranscriptomics: Three replicates at D0, D4, D15, D27 were selected for DNA and RNA sequencing (Illumina NovaSeq PE150). DNA and total RNA were extracted (Qiagen kits); RNA treated with RiboMinus and prepared with TruSeq stranded mRNA kit. Libraries (~300 bp) constructed (Illumina DNA Library Prep Kit V2; predominantly dsDNA capture). Reads were quality filtered (Trimmomatic). Assembly and MAGs: Co-assembly and phase-specific assemblies with SPAdes --meta. Contigs >2 kb binned using MetaBAT2, MaxBin2, Concoct via metaWRAP, curated (Bin_refinement). MAG quality assessed by CheckM; dereplicated by dRep; taxonomy via GTDB-Tk. Genes predicted by Prodigal; annotated against KEGG, Pfam; CAZymes via dbCAN (hmmscan); proteases via MEROPS. rRNAs via RNAmmer. Optimal growth temperature (OGT) predicted using Tome; mesophiles OGT <50 °C, thermophiles ≥50 °C. Phylogenetic trees via GTDB marker genes and FastTree. Viral identification and annotation: Viral contigs >5 kb identified using DeepVirFinder (score ≥0.7, p<0.05) feeding VirSorter2 (score ≥0.95), with CheckV quality assessment; manual curation trimmed host regions. Criteria: hallmark genes or VirSorter2 ≥0.95 or ≥80% unknown genes (eggNOG). VIBRANT virome mode cross-checked. Non-redundant clustering by CD-HIT at 95% ANI and ≥85% coverage yielded 1297 vOTUs. Taxonomy via PhaGCN2 and CAT (LCA) using RefSeq v216; functions annotated with VIBRANT (Pfam, dbCAN, KEGG, eggNOG). Lifestyle predicted via VIBRANT, PhaTYP (BERT), and manual BLAST of lysogeny genes; prophage/vConTACT2 clustering also informed temperate assignment. Abundance profiling: CoverM mapped quality-filtered reads to vOTUs (contig mode) and MAGs (genome mode) with identity ≥95% and aligned percent ≥75%. Abundances as RPKM for metagenomes; matrices used for diversity and Mantel tests. Virus–host linkages: Three in silico methods: (1) shared genomic content (bitscore ≥50, e-value <1e-3, identity ≥70%, length ≥2500 bp); (2) CRISPR spacer matches (CRT; BLASTn with 100% identity, ≤1 mismatch); (3) tRNA matches (tRNAscan; BLAST with ≥95% coverage, ≥90% identity). Viruses linked to hosts with OGT ≥50 °C were considered thermophilic. Viral AMGs: DRAM-v (scores 1–3; flags -M, -F) identified candidate AMGs; manual curation removed illegitimate categories (DNA/RNA replication/repair, viral invasion, modification of viral components). Retained AMGs were functionally annotated (eggNOG, CAZy via dbCAN HMMdb 10.0, KEGG). Selected GHs structurally modeled with Phyre2; only models with 100% confidence retained. Metatranscriptomics mapping: Quality filtering (Trimmomatic) and rRNA removal (SortMeRNA). mRNA mapped to MAGs/vOTUs with minimap2 via CoverM; activity as TPM. Entities deemed active if TPM>0 in at least two of three replicates. Gene-level expression quantified with HISAT2 and featureCounts; TPM computed. RNA viruses and transcript-derived dsDNA phages: RdRp-based discovery from metatranscriptome contigs (>1 kb) via DIAMOND BLASTX against curated RdRp databases (coverage ≥70%, E≤1e-10, score ≥70); clustered by CD-HIT (95% ANI, 85% coverage). Also assembled dsDNA phage contigs from transcriptomes (>5 kb) and clustered to compare with metagenomic vOTUs. Statistics: R v3.6.1; alpha/beta diversity (vegan), PERMANOVA (adonis, 999 permutations), ANOVA/Tukey HSD, Wilcoxon as needed. DESeq2 for differential expression (FDR p=0.05). Partial Mantel tests between MAG/vOTU matrices and nutrient turnover (Bray-Curtis; 999 permutations; BH correction). Random Forest regression (rfPermute) to assess importance of bacterial and viral community composition/activity for nutrient cycling (MSE% importance; 1000 permutations). PLS-PM (plspm) modeled direct/indirect effects of mesophilic/thermophilic bacterial/viral communities and activities on carbon and nitrogen cycling; model selection by GoF.

Process and physicochemistry: Temperature rose to ~90 °C by day 2 (hyperthermophilic), then declined through thermophilic and maturation phases. Total carbon and nitrogen decreased by 32% and 28%, respectively, by end of HTC; OM declined from 51.3% to 38.7% (all p<0.0001). WSC and WSN peaked in hyperthermophilic phase; OM degradation correlated positively with temperature, WSC, and WSN. Bacterial community: 17 phyla represented in MAGs (227 dereplicated; 180 mesophilic, 47 thermophilic). Composition and richness tracked HTC phases (PERMANOVA p<0.001). Thermophilic genera (Thermus, Planifilum) increased from 5.3% at D0 to 91.4% by D15 (p<0.001); maturation enriched Actinobacteriota. Viral community: 1297 vOTUs (≈97% dsDNA), predominantly lytic (66.2%); only 7.7% matched known RefSeq taxa and 2.6% IMG/VR, indicating novelty. 78.6% of vOTUs detected in non-thermophilic phases; 21.3% in thermophilic phases. Family-level shifts included Matshushitaviridae (Thermus phages) rising from 1.4% at D0 to 66.3% at D15 (p=0.0245). Viral richness and composition tracked HTC phases (richness p=0.036; PERMANOVA R2=0.78, p<0.001). Virus–host coupling: 21.3% of vOTUs linked to 228 MAGs via genome similarity, CRISPR spacers, and tRNA; hosts spanned major phyla. Viral and bacterial abundances were strongly correlated (R2=0.74, p<0.001); beta-diversity matrices correlated (Mantel r=0.71, p<0.001). Number of active MAGs and active vOTUs correlated negatively (R2=0.31, p=0.035), consistent with top-down control and succession. Activity patterns: 98.5% of vOTUs were transcriptionally active and followed composting temperature. Mesophilic viruses were more active early; thermophilic viruses peaked during hyperthermophilic and maturation phases. Bacterial activity of mesophiles correlated with carbon cycling and OM degradation (Mantel r=0.25–0.80), while thermophiles correlated with temperature and nitrogen cycling (Mantel r=0.25–0.60). Bacterial functional genes: All MAGs encoded CAZymes (avg: mesophiles 117, thermophiles 89 per MAG) and proteases (avg: mesophiles 52.1, thermophiles 46.3). 70.6% of MAGs carried nitrogen metabolism genes (denitrification 53.7%, dissimilatory nitrate reduction 44.4%, nitrogen fixation 3.5%). Thermophiles expressed CAZymes during high-temperature phases (p<0.0001), while mesophiles expressed nitrogen metabolism early (p=0.00017). Viral AMGs: Identified 194 high-confidence AMGs overall; 90 AMGs on 75 mesophilic phages, including 34 AMGs from 10 CAZy families (e.g., hemicellulose/chitin degradation), 14 peptidase/amino acid metabolism genes, and 2 phosphorus metabolism genes (phoH, phoD). No inorganic nitrogen AMGs detected. 99.5% of AMGs were expressed. Viral CAZyme activity decreased in thermophilic phase (p=0.0053) and positively correlated with carbon cycling (R2=0.55, p<0.0001), but not nitrogen cycling. Top-down control and VHR: Virus-host abundance ratios (VHRs) exceeded 1 and increased in thermophilic phases (mean 73.6–190.1; p=0.006). Highest VHRs for Deinococcota at D15 and Firmicutes at D27. For 12 dominant MAGs, viral and host abundances were tightly coupled (R2=0.79–0.87, p<0.001), and activities were positively correlated (R2=0.93–0.96, p<0.001). VHRs correlated positively with temperature (R2=0.63, p=0.0012), WSC (R2=0.31, p=0.034), WSN (R2=0.41, p=0.015), and OM degradation rate (R2=0.56, p=0.0056). Predicting nutrient cycling: Changes in bacterial and viral community activity explained 45.3% of variance in nutrient turnover; viral activity contributed more strongly overall and especially to carbon cycling, while bacterial activity associated more with nitrogen cycling. Community activity vs nutrient turnover showed significant positive relationships (MAGs R2=0.21, p<0.001; vOTUs R2=0.52, p<0.001). PLS-PM indicated positive links between bacterial community composition and both bacterial and viral activity; mesophiles and thermophiles contributed to carbon and nitrogen cycling, with mesophiles stronger for nitrogen and thermophiles stronger for carbon. Mesophilic viruses associated with carbon cycling; thermophilic viruses affected cycling indirectly via host regulation. RNA viruses: Only ~0.23% of metatranscriptomic reads mapped to RNA viral contigs. Most RNA viruses were eukaryote-associated (Virgaviridae dominant); few bacterial RNA phages (Fiersviridae). RNA viral abundance did not correlate with composting properties (Mantel r=0.0173, p=0.35), indicating minimal role in nutrient cycling during thermophilic phases.

The findings demonstrate that viruses are integral to nutrient cycling during HTC, tightly coupled to bacterial dynamics and activity. Mesophilic viruses contributed directly to carbon turnover by encoding and expressing CAZyme AMGs, while thermophilic viruses primarily influenced nutrient cycling through top-down control of thermophilic bacterial densities, consistent with Kill-the-Winner dynamics that facilitate microbial succession across composting phases. Virus–host ratios positively tracked temperature and indicators of carbon and nitrogen turnover, suggesting that relative viral abundance and activity can serve as indicators for ecosystem functioning and compost process efficiency. The dominance of DNA viruses, their high activity even under hyperthermophilic conditions, and the novelty of detected vOTUs underscore a substantial, previously uncharted viral contribution to terrestrial biogeochemical cycles. The differential associations—viral activity and AMGs with carbon cycling versus bacterial activity with nitrogen cycling—provide mechanistic insight into how viruses and bacteria partition functional roles across temperature regimes in HTC.

Viruses, particularly DNA phages, play pivotal roles in HTC by (i) exerting top-down control that drives microbial succession between mesophilic and thermophilic communities, and (ii) directly augmenting carbon turnover via expression of viral CAZyme AMGs associated with mesophilic hosts. Virus–host ratios correlate with efficient organic matter degradation and nutrient turnover, indicating potential utility of viral metrics as process indicators to optimize biotechnological and agricultural composting systems. The study expands the known viral diversity in compost ecosystems and clarifies how viral activity aligns with carbon versus nitrogen cycling. Future research should enrich for phages to better capture viral diversity, experimentally validate the functions of discovered viral AMGs, quantify the roles of ssDNA and RNA viruses under varying thermal and stress conditions, and test how manipulating viral communities or VHRs could enhance composting performance and nutrient recovery.

Metagenomes were generated from unfiltered DNA, likely including prophages and lysogenic signatures that can confound interpretation of lytic interactions and may underestimate free phage diversity; phage enrichment was not performed. Library preparation favored dsDNA and likely underrepresented ssDNA viruses. Only a small fraction of vOTUs were high quality, potentially underestimating viral functional diversity and AMG content. Thermophilic viral AMGs linked to metabolism were sparse, and AMG functions were not experimentally validated. RNA viruses were assessed only via RdRp-based metatranscriptomics and appeared largely eukaryote-associated, but their full contribution, especially under different conditions, may remain unresolved.