Global biogeography of airborne viruses in public transit systems and their host interactions

Abstract

Background: There is a diverse assemblage of microbes in air in built environments (BEs), but our understanding of viruses and their interactions with hosts in BEs remains incomplete. To address this knowledge gap, this study analyzed 503 metagenomes isolated from air samples from public transit systems in six global cities, namely Denver, Hong Kong, London, New York City, Oslo, and Stockholm. Viral genomes were recovered from samples via metagenomic binning, and viruses' taxonomy, functional potential, and microbial hosts were determined. The study also investigated correlations between virus and host abundances, the coevolution of clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) systems and anti-CRISPR (Acr) proteins, and the potential impacts of auxiliary metabolic genes (AMGs) on hosts.

Results: Airborne viruses in global BEs exhibited biogeographical variations in diversity, composition, function, and virus-host interactions. Nearly half of the vOTUs analyzed were from the Caulimoviridae family, while 31.8% of them could not be taxonomically classified. Diverse functions were identified within the vOTUs, together with antimicrobial resistance genes with the potential to confer resistance to various antibiotics and antimicrobial agents. Strong correlations were observed between vOTU and host abundances, with clear distinctions between virulent and temperate viruses. However, there was limited co-evolution of CRISPR-Cas systems and Acr proteins, which was likely due to the oligotrophic and physical conditions in the BEs and the dominance of vOTUs with a virulent lifestyle. Phage-encoded AMGs appeared to have the potential to enhance host fitness. These findings highlight biogeographical variations in airborne viruses in BEs and that physical and oligotrophic conditions in BEs drive virus survival strategies and virus-host coevolution.

Conclusion: There are biogeographical variations in airborne viruses in BEs in global cities, as physical and oligotrophic conditions in BEs drive virus survival strategies and virus-host coevolution. Moreover, the characteristics of airborne viruses in BEs are distinct from those of viruses found in other, more nutrient-rich ecosystems. Video Abstract.

Keywords: Airborne viromes; Built environments; Metagenomics; Virus–host coevolution.

Fig. 1

Bioinformatics workflow and biogeography of viral operational taxonomic units (vOTUs) in air samples from built environments in global cities. A Bioinformatics workflow for viral bin identification used in this study. B Accumulation curves of vOTUs in relation to the number of samples analyzed. C Distribution of the number of vOTUs in samples from a given city and between samples from different cities. The inset figure shows the average total and city-specific numbers of vOTUs per sample from each city. D Shannon indices of the viral communities in samples from each city over a two-year period. Kruskal–Wallis tests were performed to examine differences between samples from cities regardless of the year, while Mann–Whitney tests were performed to examine differences between samples from each city in two different years. E Principal coordinate analysis of the Bray–Curtis dissimilarity matrix for the samples from the cities over a two-year period. Points are colored by city and shaped by year. Permutational multivariate analyses of variance (PERMANOVA) and permutational multivariate analyses of dispersion (PERMDISP) were performed without considering the year in which the samples were collected. The ellipses show the multivariate normal distribution at a 90% confidence interval for samples from each city. NS: not statistically significant

Fig. 2

Functional landscape of the viral operational taxonomic units (vOTUs) in air samples from built environments in global cities. A Percentage and number of viral genes identified and shared between the five reference databases. B Percentages of viral genes that could and could not be matched, respectively, to any of the five reference databases. C Percentages of viral gene clusters that could and could not be matched, respectively, to any of the five reference databases, and the size distribution of viral gene clusters based on the number of genes. D Accumulation curve of viral gene clusters. E Functional annotation of viral genes according to the Protein Families (Pfam) database. Only the 50 most common viral functions, based on the number of genes matched to the Pfam database, are shown. Gene functions associated with key viral signatures are highlighted in red

Fig. 3

In-situ virus–host links and correlations between the abundances of viral operational taxonomic units (vOTUs) and hosts in air samples from built environments in global cities. A Predicted hosts at the phylum and family levels for vOTUs at the family level in samples from all cities. The length of a bar indicates the number of virus–host links. B Network diagram showing the in-situ virus–host links in samples from each city at the host-family level. The central nodes represent the hosts, and the surrounding nodes represent the vOTUs linked to those hosts. The same virus–host links may appear in multiple cities, depending on the abundance of the virus and host. C Average relative abundances of vOTUs (right panel) and their linked hosts (left panel) at the phylum and family levels in samples from each city. The bars are colored according to the taxonomy of the linked hosts. D Pearson’s correlations between host abundances and virus-to-host abundance ratios (VHRs) for six families of hosts in samples from each city. The families shown were present in greater than or equal to 150 samples across all cities, and there were significant correlations between their abundances as hosts and VHRs in samples from at least one city. Only the coefficients and p-values that were significant in samples from a given city are shown. *p < 0.05, **p < 0.01, ***p < 0.001

Fig. 4

Associations between the viral lifestyle and host or microbe abundance in air samples from built environments in global cities. A Proportions of viral operational taxonomic units (vOTUs) with a predicted virulent or temperate lifestyle (left axis) and the average virus-to-host abundance ratios (VHRs) in different host families in samples from each city (right axis). The proportion of vOTUs with a virulent or temperate lifestyle was calculated by dividing the abundance of vOTUs with a virulent or temperate lifestyle by the total abundance of vOTUs with either a virulent or temperate lifestyle. The vOTUs without a classified lifestyle and their corresponding linked hosts were excluded. B Average relative abundance of vOTUs with a predicted virulent and temperate lifestyle in samples from each city. Mann–Whitney tests were performed to assess the difference in the average relative abundance of vOTUs with different lifestyles in samples from each city regardless of the year. C Pearson correlations between virus-to-microbe abundance ratios and the abundance (RPKM) of vOTUs predicted to have a temperate (upper) or virulent (lower) lifestyle across samples from each city. NS: not statistically significant; ***p < 0.001

Fig. 5

Evidence of virus–host interactions via the clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated (Cas) system, anti-CRISPR (Acr) proteins, and potential functions of putative auxiliary metabolic gene (AMG) in air samples from built environments in global cities. A Schematic diagram illustrating the linkage between a viral operational taxonomic unit (vOTU) and its host through CRISPR spacer matching. The Acr protein (pAcr059398) of the vOTU can evade the CRISPR–Cas system (subtype I-E) of its linked host. Three viral proteins were identified within the genome of the linked host. B Schematic diagram illustrating six putative AMGs encoded by a viral genome fragment (~ 33.7 kb). C High-confidence protein structures of the six putative AMGs indicated in panel B. The colors range from the N terminus to the C terminus in a rainbow pattern