Moisture modulates soil reservoirs of active DNA and RNA viruses

A diverse and active DNA virosphere

We first leveraged two existing metagenomes that were constructed from the Konza native prairie soil^14,15 to screen for viral sequences at the site. Each of the metagenomes was obtained from a composite of all the replicate soils collected at ambient field moisture conditions. One of the metagenomes was de novo assembled from deep sequence data (1.1 Tb)¹⁴ and the second was a hybrid assembly of short and long reads (267.0 Gb)¹⁶. The combination of the two metagenomes was used to maximize the coverage of viral sequences from the Konza prairie site. To balance between the detection limits of the viral detection tools and the wide range of viral genome size, the viral contigs > 2.5 kb in length were combined with those obtained from screening of the two largest public viral databases (i.e., IMG/VR¹⁷ and NCBI Virus¹⁶) to further increase the coverage of DNA viral sequences. We acknowledge that the length cutoff of 2.5 kb would preclude detection of some ssDNA viruses with small segmented genome sizes (e.g., Nanoviridae¹⁸). As a result, a DNA viral database for the site was curated that included 726,108 de-replicated viral contigs. The DNA viral database then served as a scaffold for mapping of metatranscriptome and metaproteome datasets to determine the activities of soil DNA viruses and their responses to differences in soil moisture. This approach was also recently applied to detect the transcriptional activity of marine prokaryotic and eukaryotic viruses^19,20,21,22 and giant viruses in soil⁵.

The metatranscriptome reads from both wet and dry treatments were mapped to a total of 416 unique DNA viral contigs using stringent criteria (% sequence identity > 95% and % sequence coverage > 80%). The 416 DNA viral contigs with an average sequence length of 19 kb were highly diverse and grouped into 139 clusters, with 111 of the clusters being singletons (Supplementary Data 1).

We aimed to assign putative host taxa to the viral clusters by combining several approaches: CRISPR spacer matching, and screening for host and viral sequence similarities to respective databases (details in ‘Methods’). As a result, we assigned putative viral host taxa to 160 out of the 416 transcribed DNA viral contigs. Some of these were assigned to more than one host (Supplementary Data 1), resulting in a total of 181 virus–host pairings (Fig. 1a). Of these, 79 host–virus pairs were detected only in the dry soil treatment, 51 were only in the wet soil treatment, and an additional 51 were found in both dry and wet treatments (Fig. 1a). Consistent with previous reports⁴, the majority of the transcribed DNA viral contigs were annotated as bacteriophage sequences. Different sets of transcribed DNA viral contigs were unique to wet or dry soils and assigned to specific hosts at the phylum level, whereas others were shared (Fig. 1a). However, the dominant soil taxa, i.e., Proteobacteria and Actinobacteria that were previously identified by 16S rRNA gene sequencing in this soil environment, were predicted as hosts under both wet and dry conditions (Supplementary Fig. 1a). Eukaryotic DNA viruses, such as Bracovirus and Ichnovirus belonging to a family of insect viruses within the Polydnaviridae family, were also transcribed in the soils (Fig. 1a and Supplementary Data 1). Most of these insect viruses were only detected in dry soil conditions. These differences in virus–host pairings suggest that some of the respective hosts were impacted differently by the dry and wet incubation conditions.

There were 21 DNA viral contigs that were assigned to hosts across multiple bacterial phyla suggesting the presence of viral generalists^1,23 (Supplementary Data 1). We recognize that host assignment based on CRISPR spacer matching, however, is limited to detection of recent or historical virus–host interactions that were captured at the time of sampling²⁴. As bioinformatics assignment of virus–host linkages only suggests possible pairings based on sequence features, there are also chances of introducing false positives. However, we applied the most stringent criteria possible to provide confident host assignments.

Increased activity of a subset of DNA viruses in wet soil

Soil moisture has a strong influence on the community structures of transcribed DNA viruses. The majority of the transcriptionally active DNA viral contigs were unique to wet or dry conditions, with only 111 viral contigs (~ 26.7%) detected in both wet and dry soils, suggesting that the different soil moisture conditions may shape the activity of the DNA viral community differently (Fig. 1b). Interestingly, although a significantly higher number of transcribed DNA viral contigs were detected in dry soils (Fig. 1b, c), the levels of transcriptional activity were significantly higher (based on the normalized abundance of RNA reads that mapped to the viral contigs) for DNA viruses in wet soils irrespective of sampling site location (Fig. 1d). DNA viral contigs with mapped transcripts could represent either prophages that are passively replicated along with their host genomes, or (lytic) viruses that are actively regulating early/middle/late expression of viral gene clusters²⁵. In soil, a lysogenic lifestyle is considered to be an adaptive strategy for viruses to cope with long periods of low host activity^26,27. Therefore, the 1.5-fold increase in the number of transcribed DNA viral contigs representing transcriptionally active DNA viruses, but with lower levels of overall transcription, in dry soil suggests that the increase was due to a higher prevalence of lysogeny in dry conditions. This hypothesis is strengthened by our finding of a 20-fold increase in transcripts for lysogenic markers (i.e., integrase and excisionase) in one of our replicates (A-2) in dry compared to wet conditions (Supplementary Data 2). High number of lysogenic phages were also previously reported in dry Antarctic soils using a cultivation-independent induction assay²⁸. By contrast, under wet soil conditions we found a 2-fold increase in transcription of fewer viral contigs representing a subset of DNA viruses, suggesting that those viruses were more transcriptionally active in response to higher soil moisture. In addition, there was a higher correlation between prokaryotic abundances, as estimated by 16S rRNA gene sequencing, with DNA viral transcript counts in wet soils (R² = 0.593, Supplementary Fig. 1d) in comparison to dry soils (R² = 0.069, Supplementary Fig. 1d), supporting this hypothesis.

We then identified which soil DNA viruses were most transcriptionally active and how they responded to the differences in soil moisture. As the majority of the transcribed DNA viral contigs (97%) were environmental viruses with unclassified taxonomy assignment, we were not able to calculate the taxonomic abundance of each and instead compared the differential abundances of the transcribed viral contigs. There were four DNA viral contigs with significantly different levels of transcription under wet and dry conditions (VC_1, VC_19, VC_282, VC_412; Fig. 2a). Contigs VC_1 and VC_19 correspond to unclassified viral contigs deposited in IMG/VR (identifiers of ‘REF:2547132004_2547132004’ and ‘3300010038_Ga0126315_10000854’) that were previously detected in metagenomes from the Rifle site²⁹ and from serpentine soil in the UC McLaughlin Reserve³⁰, respectively. Contigs VC_282 and VC_412 were extracted from our Kansas metagenomes. Contigs VC_1 and VC_19 had significantly higher levels of transcriptional activity in wet soils compared to dry soils (p < 0.01, Fig. 2b), whereas VC_412 had significantly more transcripts in dry soils (p < 0.01, Fig. 2b). However, VC_282 was detected with higher transcript levels only in the dry soil replicates from site C. Five specific regions of contig VC_1 had the highest transcriptional activity in both wet and dry soil conditions, with lengths of 546, 338, 175, 420, and 663 bp, respectively (Fig. 2c). The finding that the same specific regions had differential transcript mapping frequencies is suggestive of active/lytic viruses with highly regulated transcription of early and/or late genes^25,31, in comparison to lysogenic viruses that are passively transcribed along with their host genomes. VC_1 was originally detected from the Rifle site, an aquifer environment. Therefore, this virus–host pair may be better adapted to wet conditions, reflecting our finding of higher transcription levels for VC-1 in wet soil (p < 0.05, Fig. 2b). A similar increase in activity of DNA viruses together with a bloom of their respective hosts has also previously been observed following laboratory wetting of soil biocrusts³², suggesting that this may be a common phenomenon to soil wetting.

It is interesting to note that the quality-filtered transcripts were mapped to both protein coding and noncoding regions of the DNA viral contigs. Noncoding RNAs with a phage origin have previously been reported to regulate viral replication at the early stage of infection and to maintain a lysogenic state by silencing the expression of late structural genes^33,34. We observed a trend towards a higher percentage of viral noncoding RNAs in drier soils from site A, along with higher transcriptional levels of lysogenic markers (i.e., integrase and excisionase) in these samples (Fig. 3a and Supplementary Data 2). These findings suggest higher levels of lysogenic phages in these samples, but this hypothesis needs further experimental validation. Interestingly, recent studies identified prophage-encoded noncoding RNAs that can also contribute to the virulence of bacterial hosts³⁵ and protective functions such as superinfection-immunity³⁶. Future studies are therefore needed to better characterize the functions of viral noncoding RNAs.

After removal of noncoding RNAs, the remaining transcripts were mapped to 314 viral genes. Only 149 of those transcribed viral genes were annotated to functional gene categories (Supplementary Data 3), reflecting that a large proportion of viral genes remain uncharacterized. The annotated viral genes with transcripts detected are shown in Fig. 3b with the majority having low transcript read depths (< 20 per sample). A range of viral structural genes (e.g., phage tail, head, capsid), genes encoding DNA/RNA polymerase, and genes related to DNA recombination or re-arrangement (resolvase, Rhs element vgr protein) were transcribed, indicative of active viruses in the soil incubations. After verifying the gene positions on the respective viral contigs, one auxiliary metabolic gene (AMG), acyl-CoA dehydrogenase, was also found to be transcribed (Fig. 3b and Supplementary Data 3). This gene encodes an enzyme that is involved in initial steps of fatty acid metabolism. Similarly, virus-encoded AMGs involved in the fatty acid oxidation were previously reported in viromes from the Pacific Ocean³⁷ and have been reported as a conserved strategy for a wide range of viruses in ocean³⁸. Our findings suggest that viruses may similarly be involved in host metabolism in soil.

In addition to transcripts, we detected several viral peptides in our soil metaproteome data that were indicative of viruses with a lytic lifestyle. To our knowledge this is the first report of using an untargeted, metaproteomic approach to detect viral proteins in soil, although it has previously been used with success for aquatic samples³⁹. For highly confident assignment of viral peptides with low abundances in the metaproteome, quality searching criteria (i.e., 5% false discovery rate (FDR), < 5 ppm mass error) were applied with manual inspection (Supplementary Data 4 and 6). Confident peptides for phage tail proteins, a virus-encoded protease, an RNA polymerase, and numerous chaperonins are captured in Fig. 3b. Inferring viral activity from metaproteomic data is supported by the fact that substantial components of viral structures (e.g., capsids and tails) are comprised of proteins³⁹. Due to the low abundances of viral proteins detected, we refrained from statistical comparisons of the impact of soil wetting and drying on the proteome data.

Interestingly, we detected a collection of chaperonin-like genes (K01802, K03554, K04043, and K04077) that were expressed by soil viruses at both the transcript and peptide level. Examples of confidently identified viral chaperonin peptide sequences (PSM < 1.19E-13, PPM < 2.83) are shown in Fig. 3d. These soil viral chaperonins shared similar sequence regions with bacterial, eukaryotic, and marine viral GroELs (Group I chaperonin), and also contained conserved features that clearly distinguished them as a novel group of soil viral chaperonins, as shown in the alignments (Supplementary Data 6). One example of a conserved region is displayed as an outer circle of the GroEL phylogenetic tree (Fig. 3c). Chaperonins have previously been expressed from a phage⁴⁰ and a plant virus (Closteroviruses)⁴¹. A recent study also reported a high prevalence of viral chaperonins including ones related to bacterial (GroELs) and archaeal (thermosome) chaperonins in an aquatic system⁴². Our phylogenetic analysis of bacterial, viral and eukaryotic GroEL genes (Fig. 3c) suggests that although chaperonins from the marine ecosystem are more phylogenetically similar to bacterial chaperonins, the majority of the soil viral chaperonins that we detected were most likely derived from eukaryotes, with some having bacterial origins. Such an apparent evolutionary separation of viral chaperonins in marine and soil ecosystems warrants further study. Because of the recognized functional importance of viral chaperonins in viral assembly and a lysogenic-lytic lifestyle transition⁴², the detection of soil viral GroEL at all levels, genomic, transcriptional, and translational, in our study suggests that the respective DNA viruses were actively infecting their hosts.

A diverse soil RNA virosphere

Our second aim was to examine the soil metatranscriptome data for RNA viruses. To date there have been few reports of RNA viruses in soil. Recently, RNA viruses in a marine study were found to be more abundant than DNA viruses implying important ecological functions in marine ecosystems⁷. The current knowledge of RNA viruses in soil is fragmentary, mainly focusing on culturable viruses and crop pathogens^43,44. Recently, Starr et al. (2019) reported detection of RNA viruses from metatranscriptomes of a California annual grassland soil³. Here we reassembled RNA viruses from the quality-filtered metatranscriptomic reads and uncovered a diverse RNA viral community in the Kansas native prairie soil.

The taxonomies of the identified soil RNA viral contigs are shown in rooted phylogenetic trees based on the alignments of the marker gene, RNA-dependent RNA polymerase (RdRP) for double-stranded RNA viruses, negative single-stranded RNA viruses, and positive single-stranded RNA viruses⁴⁵ (Fig. 4). Most transcripts from double-stranded RNA viruses were mapped to viral contigs annotated as Reoviridae, which are known to infect a wide variety of eukaryotic hosts⁴⁶ (Fig. 4a). One Reoviridae member that is closely related to a Bluetongue virus (MH559812, max. E-value of 1.86E-09, min. % identity of 88%) had the highest cumulative genomic coverage (Fig. 4a and Supplementary Data 7). This could be explained as either due to a bloom of the respective host during incubation, or due to their highly segmented genomes, i.e., ~10 dsRNA segments per genome, making them more resilient to sample processing steps⁴⁷. This is in contrast to Starr et al. (2019) who did not detect Reoviridae in the California grassland study³. The negative single-stranded RNA viruses had highest representation of Nairoviridae, Peribunyaviridae, and Hantaviridae across both wet and dry soils and Paramyxovirade in dry soil only (Fig. 4b). The positive single-stranded RNA viruses had highest representation of Secoviridae, Bromoviridae, Closteroviridae, Ifaviridae, Piconaviridae, Hypoviridae, and Leviviridae (Fig. 4c). The Picornavirales-like RNA viruses detected in Kansas soil were not previously found in California soil². More study is therefore needed to further extend our understanding of soil RNA viruses and their potential ecological functions.

More than half of the identified RNA viral sequences (124 out of 209) were assigned to Leviviridae, a family of negative-sense single-stranded RNA bacteriophages and the most diverse family found in Kansas soils (Fig. 4). The Leviviridae-like RNA viral sequences were clustered into two deeply branched major clades (Fig. 4d, e), supporting the recent hypothesis that the current Leviviridae family is evolving into two distinct lineages^45,48. Similar to our study, Starr et al. (2019) also found Leviviridae phages were the most diverse in California grassland soil. Leviviridae are only known to infect Proteobacteria⁴⁹ (Virus–Host DB, https://www.genome.jp/virushostdb/, accessed on 11 August 2020), one of the most abundant and diverse bacteria phyla in the Kansas soil (898 out of the total 4419 OTUs at 97% 16S rRNA gene sequence identity, Supplementary Fig. 1a). Similar observations of environmental viruses assigned to the most diverse and abundant taxa (e.g., Proteobacteria) have also been reported previously^1,2,4.

RNA viruses are responsive to soil moisture

Soil wetting and drying treatments impacted the RNA viral communities. Soil moisture treatments shaped the RNA viral communities from each of the Kansas soil locations (A and C) differently (Fig. 5a). The RNA viral communities generally grouped together according to sampling locations, except for one outlier from site A (Fig. 5a). The total RNA viral abundances were strongly correlated with the abundance of active eukaryotic species (based on 18S rRNA gene transcript counts from the metatranscriptomes, R² = 0.829 in Supplementary Fig. 1e). The correlation was higher in soils under wet conditions (R² = 0.856 in Supplementary Fig. 1f) compared to dry conditions (R² = 0.811 in Supplementary Fig. 1f). Among the detected RNA viruses, there were two viral families that responded significantly to the experimental soil wetting and drying. The estimated abundances of Leviviridae were higher in wet soils (p < 0.01, Fig. 5b). By contrast, a family of eukaryotic viruses (Paramyxoviridae) was more abundant in dry soils (p < 0.01, Fig. 5b). Due to the predominantly lytic lifestyle ascribed to Leviviridae³, higher abundances in wet soil compared to dry soil may reflect a higher degree of host lysis in wet soils. Proteobacteria, the host of Leviviridae⁵⁰, were also more abundant in wet soil (Supplementary Fig. 1a). Sampling across multiple time-points is needed to specifically investigate the continuous population dynamics of the virus–host communities, to better resolve interactions and transcriptional regulation at the community scale, and to elucidate infection dynamics.

Source: Ecology - nature.com