Abstract
Viruses in built environments (BEs) raise public health concerns, yet they are generally less studied than bacteria. To better understand viral dynamics in BEs, this study assesses viromes from 11 habitats across four types of BEs with low to high occupancy. The diversity, composition, metabolic functions, and lifestyles of the viromes are found to be habitat dependent. Caudoviricetes species are ubiquitous on surface habitats in the BEs, and some of them are distinct from those present in other environments. Antimicrobial resistance genes are identified in viruses inhabiting surfaces frequently touched by occupants and in viruses inhabiting occupants’ skin. Diverse CRISPR/Cas immunity systems and anti-CRISPR proteins are found in bacterial hosts and viruses, respectively, consistent with the strongly coupled virus–host links. Evidence of viruses potentially aiding host adaptation in a specific-habitat manner is identified through a unique gene insertion. This work illustrates that virus–host interactions occur frequently in BEs and that viruses are integral members of BE microbiomes.
Introduction
Viruses warrant our attention because they have potentially detrimental impacts on human health1 but also play crucial roles in many ecosystems2,3,4. Built environments (BEs), where people typically spend most of their lives, harbor a rich diversity of microorganisms5, but most studies of BEs have largely focused on bacteria and fungi while overlooking viruses6,7. The total concentration of the viruses in BEs is estimated to be ~105 particles/cubic meter8. Although the environmental conditions of most BEs are oligotrophic and considered poorly suited for microbial life9, a conspicuous diversity of viruses, including epidemic-associated viruses (e.g., SARS-CoV-210 and yellow fever virus11), have been found in microbial communities in air and on surfaces in BEs. A few studies on viromes in public buildings (e.g., daycare centers and restrooms) have mainly focused on a small spatial scale and limited sample types and have not investigated the bacterial hosts of the viruses12,13,14. A recent global-scale study that applied bulk metagenomic sequencing without virus enrichment provided evidence that viruses are ubiquitous on public surfaces in BEs15.
Virus–host interactions are central to the ecology and evolution of microbiomes in diverse ecosystems4,16,17. Recent advances in bioinformatic tools have enabled accurate prediction of the association between metagenome-derived viruses and their potential bacterial hosts, including exact matches of molecular signals (namely clustered regularly interspaced short palindromic repeat [CRISPR] spacer, integrated genome, and tRNA) and consistent k-mer frequency18. Phages have evolved diverse lifestyle and transmission strategies, such as temperate–lytic life cycle switching, transduction, and host gene disruption, to exploit the hosts’ cellular machinery for reproduction19. In most marine and soil environments, phages are often highly diverse and abundant, thereby routinely infecting a significant fraction of their microbial hosts, which, together with the expression of virus-encoded auxiliary metabolic genes (AMGs) in host genomes, plays a key role in global nutrient cycling4,20,21. From an ecological perspective, phages in a microbial community can mediate the competition among bacterial species by establishing lytic infections through several well-established ecological models, including the “kill-the-winner” and “piggyback-the-winner” models22.
While phages can drive rapid genetic and phenotypic changes in bacteria, bacterial hosts can also readily evolve defense mechanisms to counter phage attacks through de novo mutation and other molecular mechanisms23. Recently, various functional CRISPR/CRISPR-associated (Cas) systems in bacteria have been identified in a body-wide human metagenomic study24. However, to antagonize the host immune system, phages have evolved anti-CRISPR (Acr) proteins to inactivate bacterial Cas nucleases during infection25. Long‐term inactivation of CRISPR/Cas by inhibitor phages can lead to the loss and even absence of CRISPR/Cas in some bacterial lineages26.
CRISPR/Cas systems have been reported in surface microbiomes across urban environments worldwide15; however, the immune mechanisms of infection and the virus–host interactions (e.g., the extent of virus–host links, the prevalent viral life cycle, and the novelty of Acr proteins) that occur in BEs are poorly understood. To fill this knowledge gap and explore the diversity and ecosystem functions of viruses in BEs, 738 bulk metagenomes from diverse habitats across different BEs in Hong Kong (HK) were investigated in this study. The highly coupled virus–host interactions identified in this study support the notion that viruses aid the adaptation of bacterial hosts to the specific environmental conditions of BEs and that the abundance of most bacterial populations in BEs is strongly correlated with their resident viruses. This study provides evidence that viruses are integral members of BE microbiomes.
Results
Habitat-dependent diversity and distribution of the BE viromes
From the 738 bulk metagenomes collected from rural and urban BEs in HK, including piers, public facilities, residences, and subways (Fig. S1a, Supplementary Data 1), ~4.5 million assembled contigs were generated with MetaWRAP (see “Methods”), with lengths mostly between 1 and 3 kb were obtained (Fig. S1b). Viral contigs were identified using Visorter227 and DeepVirFinder28; the latter showed a better performance for shorter contigs (1–3 kb; Fig. S1c). In total, 594,851 unique viral contigs with lengths ≥1 kb were recovered from all samples (Fig. 1a). After quality filtering, 1174 viral genomes with completeness ≥50% (98 complete, 346 high-quality, and 730 medium-quality genomes) were identified (Fig. S1d, Supplementary Data 2). These genomes were well represented across the four types of BEs (Fig. 1b), with 66% of them detected on surfaces in residences (Fig. 1c). Despite analyzing the bulk metagenomes, only 28% of the viral genomes showed evidence of host integration based on an assessment of the provirus integration sites (i.e., the host region was predicted on both ends of viral genomes) (Fig. 1c). Of the 471 viral operational taxonomic units (vOTUs) identified, 355 were found in at least two samples (Fig. S2a); among the types of BE, the largest number of vOTUs were found in residences. At a higher taxonomic ranking, the viral genomes were clustered into 332 and 220 genus- and family-level vOTUs, respectively. The rarefaction curves of the vOTUs did not reach a plateau, suggesting that additional samples are required to capture the virome diversity in BEs within a city (Fig. S2a).
Given that the samples were collected from different habitats in terms of sources (i.e., air and surface) and materials (i.e., concrete, metal, and wood), the habitat-dependent features of the viromes were further investigated. Viral genomes were recovered most frequently from occupants’ skin (21% of all samples) and doorknobs (15%) and least frequently from air (2%) (Fig. S2b). The virome composition mostly differed between habitats (analysis of similarity R = 0.355, p < 0.001), and permutational multivariate analysis of variance confirmed that habitat was the main driver of variation (R2 = 0.148, p < 0.001) (Fig. 1d). The airborne virome was distinct from the surface-borne viromes, with a low within-habitat variance according to the Bray–Curtis dissimilarity distance in the principal coordinate analysis and the homogeneity of multivariate dispersions (permutational analysis of multivariate dispersions F = 53.29, p < 0.001) (Fig. 1d, S2c). Additionally, the vOTUs on handrails, poles, and ticket kiosks exhibited a significantly lower richness and Shannon diversity index than the vOTUs on occupants’ skin and frequently touched indoor surfaces (e.g., doorknobs) (analysis of variance [ANOVA], p < 0.05; Fig. S2d, Supplementary Data 3). The species evenness varied significantly between habitats (ANOVA, p < 0.05), but the average evenness values of all habitats were > 0.85 (Fig. S2d), suggesting that no habitat had dominant vOTUs.
Next, the vOTUs were assigned to taxonomic ranks based on comparison with known viral sequences from the Integrated Microbial Genome and Viral (IMG/VR) database29. Most of the vOTUs (92.4%) could not be taxonomically classified into a known viral genus or family, similar to the reported rate of novelty for the viromes collected from other ecosystems2,30, and could only be resolved as unclassified members of the class Caudoviricetes (Fig. 1c). Among the annotated vOTUs, 1.7%, 1.7%, 1.3%, and 0.6% belonged to the dsDNA viral genus Pahexavirus, ssRNA-RT viral family Metaviridae, ssDNA viral family Genomoviridae, and ssDNA viral family Inoviridae, respectively (Fig. S3). Specifically, the members of the family Autographiviridae, an extensively studied family of virulent phage31, were enriched and dominant in subway air (Fig. S3, Supplementary Data 4); in contrast, the members of the genus Pahexavirus were abundant on doorknobs and skin surfaces (Fig. S3, Supplementary Data 4), which is not surprising because these have been shown to infect skin bacteria (e.g., Propionibacterium32). Furthermore, the members of the order Orthopolintovirales, an emerging group of viruses known as virophages33, were also enriched on human skin-associated surfaces (Fig. S3). Notably, the human-associated papillomaviruses in the family Papillomaviridae, which can be transmitted directly or indirectly via skin contact34, were mainly found in frequently touched habitats (i.e., doorknobs, ticket kiosks, and handrails in public facilities) (Fig. S3a)…
