Les virus interagissent avec des hôtes qui couvrent des domaines microbiens éloignés dans des tapis hydrothermaux denses

Nature Microbiology volume 8, pages 946-957 (2023)Citer cet article

9124 Accès

165 Altmétrique

Détails des métriques

Dans la nature, de nombreux microbes résident dans des communautés denses et métaboliquement interdépendantes. Nous avons étudié la nature et l'étendue des interactions microbe-virus en relation avec la densité et la syntrophie microbiennes en examinant les interactions microbe-virus dans un tapis hydrothermal profond en biomasse dense. En utilisant le séquençage métagénomique, nous trouvons de nombreux cas où des microbes phylogénétiquement distants (jusqu'au niveau du domaine) codent pour une immunité basée sur CRISPR contre les mêmes virus présents dans le tapis. Les preuves d'interactions virales avec les domaines microbiens transversaux des hôtes sont particulièrement frappantes entre les partenaires syntrophiques connus, par exemple ceux engagés dans la méthanotrophie anaérobie. Ces modèles sont corroborés par une inférence basée sur la ligature de proximité (Hi-C). Des enquêtes sur des ensembles de données publiques révèlent que des virus supplémentaires interagissent avec des hôtes dans différents domaines de divers écosystèmes connus pour héberger des biofilms syntrophiques. Nous proposons que l'entrée de particules virales et/ou d'ADN dans des cellules hôtes non primaires puisse être un phénomène courant dans les écosystèmes densément peuplés, avec des implications éco-évolutives pour les microbes syntrophiques et une augmentation inter-population médiée par CRISPR de la résilience contre les virus.

Dans la nature, la plupart des bactéries et archées se trouvent sous forme d’agrégats ou de biofilms1. Ces agrégats microbiens sont souvent constitués d'organismes phylogénétiquement distants engagés dans des métabolismes interdépendants (par exemple, la syntrophie)2. Cependant, la plupart des interactions hôte-virus sont étudiées dans des cultures liquides homogènes et de nombreuses lacunes subsistent dans notre compréhension des interactions hôte-virus dans des biofilms denses, liés au substrat et hétérogènes3. En particulier, des questions majeures existent en ce qui concerne la gamme d'hôtes, le cycle de vie viral, les modes de dispersion et la coévolution hôte-virus dans des communautés microbiennes complexes où des microbes génétiquement divers et phylogénétiquement éloignés coexistent à proximité et s'engagent dans des métabolismes hautement imbriqués.

On pense généralement que les virus infectent une gamme restreinte d’hôtes. Des études récentes ont toutefois suggéré que les virus à large gamme d’hôtes pourraient être plus courants dans la nature et avoir été négligés en raison de biais de culture4. Jusqu’à présent, il existe des rapports faisant état de virus infectant plusieurs espèces bactériennes5, ordres6 et éventuellement phylums7,8,9. De plus, il a été démontré que les gammes d’hôtes viraux constituent un trait dynamique10. Notamment, une étude récente11 a rapporté que l’adsorption des phages et leur entrée dans les cellules n’équivalent pas à l’achèvement complet du cycle lytique, ce qui indique que les virus peuvent interagir avec un ensemble plus diversifié de cellules dans lesquelles un cycle d’infection complet peut être effectué.

Nous avons émis l’hypothèse qu’une gamme d’hôtes plus large pourrait être répandue dans les biofilms dominés par des métabolismes syntrophiques en raison d’un contact prolongé avec des microbes phylogénétiquement divers et d’une dispersion limitée des hôtes et des virus et/ou d’une gamme d’habitats causée par des substances polymères extracellulaires (EPS) et une hétérogénéité spatiale. Pour répondre à cette hypothèse, nous avons caractérisé les génomes viraux et toutes les interactions virales avec des bactéries ou des archées (ci-après appelées interactions hôte-virus) dans un tapis microbien hydrothermal des grands fonds, ces tapis étant des biofilms chimioautotrophes omniprésents autour des évents hydrothermaux. Ces tapis sont constitués de communautés très denses et métaboliquement couplées de bactéries et d'archées12, et présentent des gradients spatiaux marqués et une variabilité temporelle de la température et de la géochimie13. Nous montrons que les microbes phylogénétiquement distants (c'est-à-dire des taxons de différents phyla et même domaines) dotés de capacités métaboliques putativement syntrophiques codent souvent pour une immunité basée sur des répétitions palindromiques courtes groupées régulièrement espacées (CRISPR) contre les mêmes virus dans le tapis. Cette tendance n'est pas détectée dans les échantillons de panache hydrothermal physiquement adjacents présentant des biomasses plus faibles de communautés métaboliquement similaires. De plus, ces génomes microbiens présentent des colocalisations avec les mêmes génomes viraux sur la base du séquençage par ligature de proximité Hi-C. En examinant les métagénomes accessibles au public, nous avons également découvert des virus interagissant avec des taxons bactériens et archéens dans d’autres écosystèmes connus pour héberger des biofilms syntrophiques. Nous avons en outre étudié les implications éco-évolutives de ces interactions hôte-virus en examinant les gènes métaboliques auxiliaires (AMG) dans les génomes viraux, ainsi qu'en identifiant les gènes viraux et microbiens en cours de sélection. Enfin, nous proposons quatre modèles d'interactions virales polyvalentes avec des hôtes syntrophiques et discutons de leurs implications sur l'évolution microbienne, notamment en ce qui concerne le transfert horizontal de gènes, la diversification génétique et la mémoire immunologique à l'échelle communautaire médiée par CRISPR.

0.05) between microbial and viral compositions was identified. The rep_vMAGs recovered from this study exhibited very high taxonomic and gene content diversity relative to the genetic diversity space occupied by the reference viral genomes (Extended Data Fig. 2). Only 4 rep_vMAGs could be clustered at the ‘genus’ level20 with reference viral genomes, and could be classified as two (previously designated) Podoviridae, one Myoviridae and one Siphoviridae (Supplementary Table 4). Notably, a taxonomic cluster consisting of 7 rep_vMAGs was distantly associated with Flavobacterium phages, and 3 of the rep_vMAGs formed a novel genus-level cluster that shared no similar genes with any of the characterized reference viral sequences. A majority (29 out of 49) did not share high similarity in gene content with the reference or with each other. Many of the viral genomes contained novel auxiliary metabolic genes (AMGs) such as Rubisco large domain-containing protein, aldolase II domain-containing protein, nitroreductase domain-containing protein, phosphate starvation-inducible protein PhoH and terillium resistance protein TerD (Extended Data Fig. 3a,b). We also detected evidence of host-virus arms race, with some viral genomes encoding defence machinery such as RelE/StbE family toxin, HigA family antidote and a putative abortive infection protein (Extended Data Fig. 3c). A complete list of the annotated AMGs and other notable viral genes is provided in Supplementary Table 5./p>95% nucleotide identity (ID)) into 102 clusters. Most (91%) of the detected CRISPRs were specific to a population and 80% of the CRISPR-encoding populations were associated with at most 2 unique CRISPRs. However, we observed identical or near identical (>95% ID) CRISPR repeats shared among phylogenetically distant populations. It is possible that these CRISPR loci were horizontally transferred21, but we cannot rule out the possibility of binning errors resulting from their repetitive and divergent nature. Such CRISPR repeats detected across taxa were excluded from spacer-based host-virus matching due to the ambiguity in assigning a specific host taxon to a repeat. Additionally, we identified populations (Gammaproteobacteria_17_1, Desulfobacteria_193_1, Desulfobacteria_189_1) encoding as many as 6 distinct CRISPR repeats, probably representing within-population diversity of CRISPR loci. No correlation was found between the number of unique CRISPRs and the rep_mMAG size, relative abundance or habitat range./p> 0.05). We binned 168 mid- to high-quality rep_mMAGs (see Supplementary Table 11 for the full description) across the 10 HW assemblies, and although taxonomically distinct from the rep_mMAGs recovered from the mat assemblies, the two datasets featured similar metabolic capabilities (Supplementary Table 12 and Extended Data Fig. 8a) and similar levels of species evenness (Extended Data Fig. 8b, Welch’s t-test, two-sided, n = 20, P > 0.05). The microbial communities of the HW samples were more homogeneous than the mat samples (Extended Data Fig. 8c) despite the larger physical distances between the HW samples. Similar to the mat samples, HW samples were dominated by two sulfur oxidizing Gammaproteobacteria (HW_Gammaproteobacteria_164_1, HW_Gammaproteobacteria_163_1; Extended Data Fig. 9a). Interestingly, we observed an order of magnitude less frequent detection of CRISPR loci in the HW assemblies compared with the mat assemblies (Supplementary Table 13, Welch’s t-test, two-sided, n = 20, P = 0.001). Furthermore, only 12 of the CRISPRs in the HW assemblies could be associated with medium- to high-quality MAGs (Supplementary Table 14), resulting in a much sparser and less robust CRISPR-based immunity network (Extended Data Fig. 5b and Supplementary Table 15), with only one confident interaction between an SOB (HW_Gammaproteobacterira_162_1) and a virus. The similarities between the plume and mat samples, such as geographical proximity, community metabolic capabilities and sequencing depth, provide a rationale and opportunity for comparison. Lower abundances of the CRISPRs in the plume samples indicate that the plume communities are less reliant on CRISPR-based adaptive immunity. The transferability and specificity of CRISPR-based immunity confer ecological significance to this observation, raising the question of how such immunological memory is selected for in different environments. While this comparison illuminates key differences in the nature and extent of host-virus interactions between the mat and the plume, there are some caveats to consider for further interpretation: first, the sparseness in the plume CRISPR-based immunity network is likely due in part to the lower abundance and diversity of recovered viral contigs (Supplementary Table 16 and Extended Data Fig. 9b), where only the fraction of viruses that were infecting microbes and/or were attached to particles larger than 0.022 µm were recovered. Second, differences in the CRISPR-based immunity do not necessarily reflect the patterns of the underlying networks of in situ host and virus interactions./p> 2.5); however, most could not be annotated with a function. Interestingly, 3 of the 4 annotated genes undergoing diversifying selection were involved in DNA and RNA metabolism, such as genes encoding DNA-directed RNA polymerase (RNAP) beta and beta prime (rep_vMAG_21), DNA ligase (rep_vMAG_31) and Superfamily II DNA/RNA helicase (rep_vMAG_6). We also detected a LamG domain-containing protein (vMAG_4), possibly involved in signalling and cell adhesion, to be undergoing diversifying selection. The gene encoding RNAP in rep_vMAG_21 (RNAP1; Extended Data Fig. 10a) featured the highest pN/pS ratio of 4.9, with 8 non-synonymous mutations scattered throughout the protein (Extended Data Fig. 10b). Notably, rep_vMAG_21 featured a second RNAP gene fragment encoding the beta subunit (RNAP2) (Extended Data Fig. 10a) that is not homologous to RNAP1 and not seemingly undergoing selection, possibly contributing to the relaxation of purifying selection on RNAP1. RNAP1 was highly divergent from the previously characterized RNAP sequences and was rooted at the base of the Caudoviricetes multimeric RNAP clade38 (Extended Data Fig. 10c). This example of diversifying selection on RNAP1 suggests that these viruses may play an important role in expediting the evolution of housekeeping proteins that typically undergo purifying selection in cellular organisms. Microbial genes undergoing diversifying selection (pN/pS > 2) included genes encoding products involved in various defence systems, such as type II toxin-antitoxin system RelE/ParE toxin, HindIII family type II restriction endonuclease, Type III-B CRISPR module RAMP protein Cmr1, as well as genes involved in more recently characterized PARIS and Septu anti-phage arsenal39./p>70% completeness and <10% contamination) MAGs were used for subsequent analysis. Mid- to high-quality MAGs were dereplicated at 97% ANI using dRep v3.0.1 (ref. 54) and were designated as representative MAGs (rep_mMAGs). rep_mMAGs were taxonomically classified using GTDB-Tk v1.7.0 (ref. 55). Genes were predicted using Prodigal v2.6.3 (ref. 56) and annotated by aligning them using Diamond v2.0.7.145 (ref. 57) against the UniRef100 database58 with an e-value cut-off 1 × 10−5. Additionally, METABOLIC v4 (ref. 59) and DefenseFinder v1 (ref. 60) were used to identify potential metabolic and antiviral genes, respectively./p>

5× and breadth (fraction of the rep_mMAG covered by at least one read) of >0.7, relative abundances in each sample were determined using the genome-wide average read-mapping coverage. For rep_vMAGs with an average coverage of >5× and breadth >0.7, normalized abundances in each sample were calculated by normalizing the average coverage of viral scaffolds in each rep_vMAG by the number of reads in each sample./p>

20 bp) than in Fig. 3 (spacer length >25 bp and each edge representing two distinct matches). Only interaction with spacer length >25 bp is highlighted with the red edge. Viral nodes are scaled to the rep_vMAG length, and rep_mMAGs with genomic capacity to carry out sulfur oxidation are colored in blue./p>5, breadth >0.7). Viral nodes (circular) are labeled according to the corresponding rep_vMAG ID. microbial nodes are colored according to the taxon, using the same color scheme as the main Fig. 3. Node sizes correspond to the sample-specific read-mapping coverages. Thickness of the edges represent the number of contig-to-contig linkages, while the darkness of the edges correlates to the maximal normalized strength of the Hi-C contacts between any two contigs in a host-virus pair. Host-virus pairs that were previously detected using CRISPR-spacer matches are colored in red. Identified Hi-C linkages between viruses are noted with blue edges./p> 0.05). Box plot shows the quartiles (25, 50, 75 percentiles) with the upper and lower whiskers showing the max and min value within 1.5 times the interquartile respectively. (C) Principal coordinate analyses of the rep_mMAGs in the two datasets; hydrothermal mat samples are colored in red and hydrothermal water samples are colored in blue. The percentage of variance explained by each axis is shown in the axis label./p>5 coverage and >0.7 breadth using read mapping are shown and proviruses are excluded./p>25 with at most two mismatches. Supplementary Table 9. Hi-C library statistics. Supplementary Table 10. Hi-C normalized linkage between rep_vMAG and rep_mMAG. Contig-to-contig linkage information was consolidated by count of linkage, average residual (normalized ‘strength’) of the linkage and maximum residual of the linkage. Supplementary Table 11. Statistics of the rep_mMAGs binned across the ten hydrothermal water samples. Supplementary Table 12. Genome-based metabolic capabilities and other genetic features of rep_mMAGs in the hydrothermal water samples. Supplementary Table 13. Number of high-confidence (evidence level = 4) CRISPR repeats binned across environments. For hydrothermal mat samples and hydrothermal mat samples from this study, we include information on the total number of evidence level 4 CRISPRs detected across environments as well as those that were binned in mid- to high-quality MAGs. Supplementary Table 14. Binned CRISPR repeats in hydorthermal water samples. Supplementary Table 15. All CRISPR-spacer to protospacer matches in hydrothermal water samples with spacer length >20 with at most two mismatches. Supplementary Table 16. Information of the high-quality and complete rep_vMAGs binned across the ten hydorthermal water samples. Supplementary Table 17. List of UViG ID and their putative hosts and host-prediction methods./p>