Marine viruses disperse bidirectionally along the natural water cycle

The study investigates whether and how marine viruses residing at the air–sea interface (surface microlayer, SML, and sea foams) become aerosolized, disperse through the lower atmosphere, and return to Earth via wet deposition (rain/snow). Marine viruses are abundant and influential in ocean biogeochemistry and microbial diversity, yet their dynamics at the air–sea boundary, where neuston organisms modulate gas and organic matter exchange, are poorly understood. Prior work suggests viruses and bacteria can be selectively enriched in the SML and aerosolized, and microorganisms can serve as ice nucleating particles (INPs) that influence cloud processes. Precipitation is a major vector for microbial deposition, but viral transmission pathways from sea surface to atmosphere and back via rain remain speculative. The purpose is to quantify exchange of viral populations among sea surface ecosystems, boundary layer aerosols, and precipitation; identify signatures of aerosolization and adaptation (e.g., genomic G/C content); and link viruses to hosts across ecosystem boundaries.

Background work shows: (i) viruses are abundant in marine waters, shape microbial diversity and metabolism, and drive carbon cycling via the viral shunt; (ii) the SML (<1 mm) concentrates organisms (neuston) via bubble transport and particle attachment; (iii) sea foams can be widespread, concentrate viruses up to 300×, and may harbor pathogens, potentially facilitating dispersal and aerosolization; (iv) taxon-specific aerosolization of microbes has been observed in mesocosms, but field metagenomics for marine virioneuston is scarce; (v) microorganisms, potentially including viruses, can act as INPs affecting cloud formation and precipitation; (vi) atmospheric deposition delivers substantial microbial loads, with marine sources important; (vii) storms and rainfall can alter viral assemblages in marine and freshwater systems. These studies collectively motivate field-based, metagenomic assessment of viral exchange along the water cycle.

Study area and sampling: Coastal Skagerrak near Tjärnö Marine Laboratory, Sweden, February 2020. Eleven marine stations sampled for SML (glass plate method), sea foams (surface patches), and subsurface water (SSW, 1 m). Boundary layer aerosols were sampled land-based at ~2 m above ground using a constant-flow sampler onto 0.1 µm PTFE filters. Precipitation (rain, snow) was collected via funnels into glass bottles; rain was filtered on 0.2 µm membranes with viral fraction recovered by FeCl3 flocculation. Some rain samples across days were pooled to obtain sufficient DNA. Environmental and counts: Wind speed, light, temperature, and salinity recorded. Flow cytometry quantified virus-like particles (VLPs), prokaryotes, and small phototrophic eukaryotes; VLP counts validated by epifluorescence microscopy. Enrichment factors (EF) computed as SML/SSW ratios. Ice nucleating particles (INPs) measured using immersion freezing assays of 5 µm prefilters. Air mass trajectories: 5-day HYSPLIT backward trajectories ending at 700 m above site computed hourly using ERA5 reanalysis to evaluate time over sea vs land and loading conditions (within mixing layer and surface wind >3 m s−1) for two rain events (Event 1: Feb 7–9; Event 2: Feb 14–16 and 20–22). DNA extraction and sequencing: DNA extracted from size fractions (>0.2 µm, <0.2 µm flocculated viral fraction) of seawater, rain, snow; aerosol filters processed with PowerMax kit. Sequenced shotgun metagenomes (55 total). Negative/handling controls sequenced; reads mapping to controls removed. Metagenomic pipeline: Reads quality-trimmed (bbduk, Sickle). Assemblies via metaviralSPAdes (viral-enriched) then metaSPAdes; combined for downstream. Viral identification with VIBRANT, VirSorter (categories 1,2,4,5), ViralVerify; host contamination removed with CheckV. Viral scaffolds >10 kb dereplicated at 95% similarity (VIRIDIC). Read mapping to viral scaffolds with Bowtie2; presence called if ≥75% genome breadth covered with reads ≥90% identity. Mean coverage computed and sum-normalized. Viral genes annotated with Prodigal (meta) and DRAM-v; gene-sharing networks with vConTACT2 vs RefSeq (Dec 2021). Viral clusters (VCs) used for diversity and enrichment analyses. G/C content calculated per viral scaffold detected in samples. Prokaryotic community: rpS3 genes identified (DIAMOND to FunTaxDB, clustering at 99% identity) for taxonomic profiles and diversity analyses (phyloseq, Bray–Curtis NMDS, PERMANOVA). MAGs binned with MetaBAT2 and MaxBin2, aggregated with DAS Tool, curated with uBin; quality checked (CheckM) and classified (GTDB-Tk). Read mapping assessed MAG presence and breadth. Virus–host links inferred by (i) k-mer frequency matching (VirusHostMatcher, d2* ≤ 0.3) to MAGs and (ii) CRISPR spacer–protospacer matches: CRISPR arrays (evidence ≥3) detected with CRISPRCasFinder; DR-guided spacer extraction with MetaCRAST from read sets; spacers filtered, clustered (99%), BLASTn-short against viral scaffolds (≥80% nt identity). SNP/variant analysis performed on select circular viral genomes (Geneious). Statistical analyses included Pearson/Spearman correlations, linear models (AIC), Kruskal–Wallis with Dunn’s tests, PERMANOVA, and betadispersion.

• Abundances and enrichment:
  • VLPs: Sea foams 5.0×10^7–1.8×10^8 mL−1; SML 1.3×10^7–3.4×10^7; SSW 1.4×10^7–2.0×10^7, indicating a gradient toward the atmosphere. Rainwater VLPs: 3.7×10^3–3.4×10^5 mL−1. SML VLP EF (SML/SSW) ranged 0.7–1.9.
  • Prokaryotes: Foams 1.3×10^6–3.8×10^6 mL−1; SML 7.0×10^5–1.1×10^6; SSW 7.0×10^5–8.7×10^5. EFs 0.9–1.3. Rain: 2.7×10^3–1.8×10^4 mL−1. Small phototrophic eukaryotes: foams 3.4×10^3–1.8×10^4; SML 1.8×10^3–6.2×10^3; SSW 2.1×10^3–5.0×10^3 mL−1.
  • Virus-to-prokaryote ratios highest in foams (25.3–48.4), then SML (15.5–34.2), SSW (19.3–26.7); rain varied widely (7.1–127.8).
  • Significant correlations in SML between VLPs and hosts: VLP vs prokaryotes (Pearson r=0.70, p=0.025), VLP vs small phototrophic eukaryotes (r=0.74, p=0.014). EFs of VLPs vs prokaryotes correlated (Spearman ρ=0.83, p=0.0056). In SSW, VLP–host correlations not significant.
  • Environmental links: In SSW, prokaryotes and small phototrophic eukaryotes correlated with salinity (r=0.71, p=0.022 and r=0.68, p=0.032). A linear model with wind×salinity explained 59.6% variance in SML eukaryote EF (F=5.43, p=0.038), outperforming single-parameter models.
• INP activity: Highest in sea foams, then SML, then SSW; freezing onset at −4 to −6 °C.
• Prokaryotic diversity and composition: Aerosol and rain had lower alpha-diversity than marine samples; aerosols significantly less diverse than SSW (adjusted p=0.0136). Beta-diversity differed among ecosystems (PERMANOVA p<0.001). Shared rpS3 genes across marine, aerosol, and precipitation indicated aerosolization of specific taxa (e.g., Oceanospirillum, Loktanella, Pelagibacter, Crocinitomix, Pirellula); some taxa in aerosols/rain absent locally in seawater.
• MAGs and host predictions: 116 MAGs (median completeness 86.3%, contamination 3.9%); mostly Gamma-/Alphaproteobacteria, Bacteroidia, Planctomycetes; one MAG from rain (Pedobacter), one from aerosol (Pirellulales). K-mer host links: 120 marine viruses matched Ca. Pelagibacter MAGs; rain-only viruses linked to Porticoccaceae and Flavobacteriaceae MAGs.
• Viral diversity and enrichment: Viral alpha-diversity differed between aerosols and SSW (p<0.0001) and between SML (>0.2 µm) and SSW viromes (p=0.029). SML >0.2 µm viral communities were distinct (NMDS p=0.001). Several viral clusters (VCs) were enriched in rain relative to SML/foam (max EF up to 15.8), including clusters related to Pelagibacter and Flavobacterium phages; many enriched VCs were unrelated to known references (singletons/outliers).
• Case study of aerosolization: Two circular viruses examined. Virus_1 (39.7 kb, 46.7% G/C) was abundant in foams and detected in aerosols and rain; Virus_2 (35.1 kb, 35.1% G/C) abundant in seawater but absent in aerosols/rain. Virus_1 linked to a Porticoccus MAG; SNP overlaps across foam, aerosol, and rain supported transfer.
• Cross-ecosystem sharing: Of 1813 viral scaffolds (>10 kb), foams, SML, and SSW shared 837; 15 viruses detected across all ecosystems. Precipitation had 109 viruses found exclusively in rain; overall, 6.2% (112) of viruses were shared between precipitation and seawater (including foam). Some aerosol–rain overlaps may be affected by minor rain contact with filters.
• Air mass trajectories and marine signal: Rain Event 2 (Feb 14–22) air masses spent 72% of prior 4 days over sea with loading conditions met 35% of the time, yielding more marine viruses (85) and MAGs (79) detected in rain than Event 1 (Feb 7–9; 64% over sea; loading 10%; 22 marine viruses and 38 marine MAGs in rain).
• Genomic adaptations: Viruses exclusively detected in rain had significantly higher G/C content than total rain viruses (p=0.0002). Marine viruses had significantly lower G/C than aerosol, total rain, and rain-only viruses (p<0.0001), suggesting adaptation or source differences.
• CRISPR evidence of encounters: CRISPR spacers from marine samples matched protospacers of seawater viruses and rain-only viruses, indicating virus–host interactions across ecosystem boundaries; dominant arrays showed preferential targeting of rain-derived vs marine viruses.

The results demonstrate that aerosolization of marine viruses from the sea surface occurs naturally, with selective enrichment in the SML and especially sea foams facilitating transfer into the boundary layer and subsequent deposition via precipitation. Significant correlations between VLPs and host cell abundances and EFs in the SML indicate tight virus–host coupling at the interface, consistent with host-dependent viral enrichment. Lower diversity and distinct community structures in aerosols and rain relative to seawater reflect environmental filtering and source mixing during aerosolization and scavenging. The detection of overlapping viral populations, including 6.2% shared between rain and seawater, together with SNP continuity of an individual virus from foam to aerosol to rain, provides direct evidence for cross-boundary transport. Air mass trajectory analysis links enhanced marine viral and MAG deposition in rain to greater ocean exposure and favorable loading conditions, underscoring the importance of meteorology in shaping airborne microbial and viral biogeography at the ocean–atmosphere interface. Higher G/C content in rain-only and aerosol viruses relative to marine viruses suggests selection for traits that may confer atmospheric persistence (e.g., resistance to UV/thymine-specific damage) or reflects different source pools with distinct nucleotide compositions; potential host–virus G/C matching may further influence these patterns. CRISPR spacer matches from marine prokaryotes to rain-exclusive viruses indicate regular encounters and potential infectivity upon deposition, implying that atmospheric cycling can introduce novel viral genotypes into marine surface communities and influence microbial evolution and biogeochemical processes.

This study provides field-based metagenomic evidence that marine viruses are aerosolized from the sea surface (especially from SML and foams), transported within the boundary layer, and returned via precipitation, establishing bidirectional dispersal along the natural water cycle. Approximately 6.2% of viruses were shared between precipitation and seawater; viral transfer was supported by SNP continuity and CRISPR spacer records indicating cross-ecosystem encounters. Rain and aerosol viruses showed elevated genomic G/C content compared to marine viruses, suggesting adaptation to atmospheric conditions or different source contributions. Air mass history strongly influenced the extent of marine viral/MAG signatures in rain. Future research should: (i) conduct culture-based infection experiments to test infectivity post-aerosolization and post-deposition; (ii) refine source apportionment and quantify spatial–temporal dynamics across diverse regions and meteorological conditions; (iii) assess functional roles of airborne marine viruses, including potential as ice nucleating particles and impacts on cloud processes; (iv) integrate standardized atmospheric sampling (varied altitudes) to bridge ocean, boundary layer, and cloud microphysical regimes.

• Aerosol sampling height (~2 m above ground) characterizes seaborne boundary-layer aerosols but is not representative of broader atmospheric aerosol populations (e.g., CCN/INPs aloft).
• Potential cross-contamination: small amounts of rainwater may have contacted aerosol filters during sampling/filter exchange; handling controls were applied, but residual effects cannot be fully excluded.
• VLP quantification caveats: particle-rich foams required pre-removal of >50 µm particles before flow cytometry, possibly underestimating VLPs; fluorescence-based methods can generate false positives (fake VLPs).
• Enrichment factor interpretation does not account for residence times in the SML and dynamic interface turnover.
• G/C content interpretations are inferential; causality (e.g., UV protection, nutrient limitation, temperature) requires culture-based validation and controlled experiments.
• Some analyses had limited sample sizes (e.g., snow n=1) and pooled rain samples, potentially affecting diversity estimates and detection sensitivity.