Genomic, metabolic and phenotypic variability shapes ecological differentiation and intraspecies interactions of Alteromonas macleodii

Ecological differentiation between strains of bacterial species is shaped by genomic and metabolic variability. However, connecting genotypes to ecological niches remains a major challenge. Here, we linked bacterial geno- and phenotypes by contextualizing pangenomic, exometabolomic and physiological evidence in twelve strains of the marine bacterium Alteromonas macleodii, illuminating adaptive strategies of carbon metabolism, microbial interactions, cellular communication and iron acquisition. In A. macleodii strain MIT1002, secretion of amino acids and the unique capacity for phenol degradation may promote associations with Prochlorococcus cyanobacteria. Strain 83-1 and three novel Pacific isolates, featuring clonal genomes despite originating from distant locations, have profound abilities for algal polysaccharide utilization but without detrimental implications for Ecklonia macroalgae. Degradation of toluene and xylene, mediated via a plasmid syntenic to terrestrial Pseudomonas, was unique to strain EZ55. Benzoate degradation by strain EC673 related to a chromosomal gene cluster shared with the plasmid of A. mediterranea EC615, underlining that mobile genetic elements drive adaptations. Furthermore, we revealed strain-specific production of siderophores and homoserine lactones, with implications for nutrient acquisition and cellular communication. Phenotypic variability corresponded to different competitiveness in co-culture and geographic distribution, indicating linkages between intraspecific diversity, microbial interactions and biogeography. The finding of “ecological microdiversity” helps understanding the widespread occurrence of A. macleodii and contributes to the interpretation of bacterial niche specialization, population ecology and biogeochemical roles.


Results and Discussion
This study combines genomic and phenotypic evidence to illuminate mechanisms of ecological differentiation within Alteromonas macleodii, a bacterium with widespread distribution and biogeochemical importance in the oceans 24 . The study focused on twelve A. macleodii strains with closed genomes, featuring average nucleotide identities (ANI) of 96.5-99.9% and 16S rRNA gene similarities of >99% ( Fig. 1; Table S1). Despite this clear association to a single genospecies 37 , underlined by 3002 core genes, we detected considerable strain-level diversity related to 1662 accessory and 1659 unique gene clusters (Table S2). This is consistent with the pronounced diversity of the flexible genome in A. macleodii and the "sister species" A. mediterranea, as described previously 30 . Intraspecific differences were highlighted by a diverse pan-exometabolome of 138 core, 1796 accessory and 2096 unique molecular masses secreted during late exponential growth (Table S3). In the following, we contextualize (pan)genomic and phenotypic evidence to characterize how genome plasticity shapes interactions with cyanobacteria and macroalgae, degradation of aromatics and polysaccharides, chemical communication, iron acquisition, and intraspecific competition. These insights expand structural-genomic and evolutionary aspects of the Alteromonas pangenome 30,32,34,38,39 by ecological perspectives on niche specialization, competitive abilities and biogeography. plasmids and genomic rearrangements. As niche specialization is often mediated by mobile genetic elements 40 , we first characterized occurrence and function of plasmids. Eight out of twelve A. macleodii strains, including MIT1002 and EZ55 whose genomes were re-sequenced and closed herein, were found to contain a plasmid (Figs. 1, S1). Synteny of plasmids from A. macleodii Te101 and A. mediterranea DE1 corroborates the role of plasmids for niche specialization within and across species boundaries 13,41 .
The plasmids of six strains display a similar functional profile, harboring metal resistance and [NiFe] hydrogenase cassettes ( Fig. 2A) that have been described in Alteromonas before 42,43 and provide increased resistance compared to strains lacking these cassettes 43 . As homologous cassettes in A. mediterranea are encoded in a chromosomal genomic island 30,32,44 , plasmids possibly mediate their transfer between Alteromonadales 13 . Notably, number and arrangement of cassettes differed between strains ( Fig. 2A), which may result in varying expression levels and hence different resistance profiles 45 . In strain MIT1002, hydrogenase and resistance cassettes have been inserted into the chromosome, and a unique chemotaxis-related plasmid has been acquired (Fig. 2B). This event may enhance chemosensory abilities and provide a competitive advantage to access nutrient patches 46 .
The plasmid of strain EZ55 harbors a unique 20 Kb insert, enabling aerobic degradation of the aromatic hydrocarbons toluene and xylene (Figs. 2; S2) as rarely described in marine microbes to date 47,48 . The insert is overall homologous to the TOL plasmid from Pseudomonas putida ( Fig. 2A), a hydrocarbon-degrading Gammaproteobacterium from soil 49 . However, closer examination using MultiGeneBlast 50 suggests assembly during separate horizontal transfer events. Specifically, the downstream section (locus tags 04282-04290) has highest similarity to TOL plasmids from Pseudomonas strains, with amino acid identities between 70 to 86% (Fig. 3A). In contrast, the upstream section including the catechol meta-cleavage pathway (locus tags 04248-04260) has highest similarity to homologous clusters in Marinobacter followed by Pseudomonas spp., with amino acid identities between 52 and 98% (Fig. 3A). Considering multiple adjacent transposases and recombinases (locus tags 04244, 04264, 04266, 04267, 04270, 04273, 04279, 04291) and the fact that Alteromonas, Pseudomonas and Marinobacter co-occur during oil spills where toluene and xylene are present 51 , we hypothesize exchange of these clusters at contaminated sites. Alternatively, Marinobacter might constitute a "vehicle" between soil and seawater due to its occurrence in saline lakes und intertidal areas 52 and known acquisition of aromatic-degrading genes from Pseudomonas 53 . Considering the common association of Marinobacter spp. with phototrophs 54,55 , the cluster might likewise enable degradation of ecologically more relevant aromatics from cyanobacteria, e.g. derivatives of benzoate or cinnamate 56 . Alteromonas and Prochlorococcus. In addition to plasmids, ecological differentiation also relates to varying abilities for microbial interactions 57 . In this context, strains MIT1002 and EZ55 are naturally associated with Prochlorococcus cyanobacteria, to whom they establish mutualistic relationships by alleviating oxidative stress or nutrient limitation during extended periods of darkness [58][59][60][61] . Here, we demonstrate additional features that may support their co-existence. Specifically, only MIT1002 harbors a gene cluster encoding the potential for phenol metabolization (Figs. 4A; S2). This ability appears ecologically relevant considering upregulation of phenol hydroxylases in co-culture with Prochlorococcus (Table S4 with data from 62 ), the common production of phenolics by cyanobacteria 63 , and presence of a homologous gene cluster in Marinobacter algicola with comparable association to phototrophs 54 . The Alteromonas-Prochlorococcus interplay may be further strengthened by metabolic interrelations, as FT-ICR-MS revealed that MIT1002 and EZ55 secrete ecologically relevant exometabolites ( Table 1). Secretion of methyl-tryptophan and methyl-indolepyruvate may explain the differential regulation of tryptophan biosynthesis in Prochlorococcus when co-cultured with A. macleodii 64,65 , especially under restricted photosynthesis 59 . Secretion of asparagine and glutamine ( Table 1) indicates exchange of further amino acids, coincident with upregulation of related importers in Prochlorococcus when co-cultured 64 . Possible cross-feeding is supported by the potential for mixotrophy 66 and considerable usage of exogenous amino acids 67 in environmental Prochlorococcus assemblages. Hence, these compounds are possible drivers of varied prokaryotic 68,69 but also interkingdom interactions, as A. macleodii can likewise counteract amino acid deficiency in microalgae 25 .
Comparison with prior transcriptomic data 59 showed that interactions of MIT1002 with Prochlorococcus involve several unique genes (Table S4). For instance, differential regulation of unique chemotaxis-, motility-and biofilm-related genes in co-culture may strengthen physical associations 70 whereas upregulation of a phytase gene might enhance phosphorus acquisition 71 . Overall, the array of interactive features suggests that MIT1002 and EZ55 are adapted to a mutualistic niche with Prochlorococcus, a relevant notion considering the cyanobacterium's reduced metabolic repertoire and importance for biogeochemical cycles 72,73 . Alteromonas, macroalgae and algal polysaccharides. We herein isolated A. macleodii strains BGP6, BGP9 and BGP14 from alginate-supplemented microcosms in the south, equatorial and north Pacific Ocean (Table S1) using analogous procedures that yielded the alginolytic strain A. macleodii 83-1 from the Atlantic 33 . Strikingly, the new isolates and strain 83-1 are clonal, featuring only four polymorphisms in 4,801,369 core sites despite being isolated over wide geographic and temporal scales. These observations resemble the isolation of A. mediterranea strains with less than 100 polymorphisms from distant locations and years apart 30,38 . In addition, two A. australica strains with 99% ANI have been retrieved from opposite global locations 44 , illustrating that highly similar Alteromonas spp. are widely distributed over time and space.
The four clonal A. macleodii strains encode numerous carbohydrate-active enzymes (CAZymes) and other enzymes involved in carbohydrate-related KEGG categories ( Fig. S3A; Table S2), enabling the degradation of various algal polysaccharides 74 and indicating association with plants 75 . To examine whether these features trigger direct interactions with algae, A. macleodii 83-1 was incubated with tissue from the marine macroalga Ecklonia radiata, which contains >50% alginate and hence a preferred substrate 33 . However, no significant tissue degradation was observed (Fig. S3B) although epibiotic bacteria cause visible digestion of Ecklonia and other macroalgae [76][77][78] . These observations suggest that A. macleodii has limited abilities to attack macroalgal tissue, and potentially utilizes polysaccharide exudates released directly by the macroalga 74 or by co-metabolizing bacteria 18 . This proposed lifestyle is supported by low Alteromonas abundances on wild macroalgae 79 . Alternatively, colonization might occur in a neutral manner, comparable to other Alteromonas spp. with a similar CAZyme profile 80 . Considering nucleotide substitution rates of ca. 10 −8 per site/year in related Gammaproteobacteria 81 , the four clonal strains probably diverged only recently followed by rapid geographic spread, comparable to Phaeobacter strains from the same Pacific transect 82 . However, some features illustrate the beginning of differentiation. In BGP9, a 91 Kb region harboring a TonB/ExbBD membrane system and vitamin B 6 synthesis genes was translocated from chromosome to plasmid (Fig. 2C), which may influence iron and vitamin metabolism 83,84 . The transposed region also harbors the strain's sole cytochrome bc 1 complex, although essential genes are uncommon on plasmids 85 . At an estimated plasmid loss of ~10 −3 per cell and generation 86 , this event may pose a considerable risk for survival.  www.nature.com/scientificreports www.nature.com/scientificreports/ Specific adaptations to algal polysaccharide degradation were also found in strains MIT1002 and AD45, mediated by unique polysaccharide utilization loci (PUL) 87 . Specifically, only MIT1002 harbors a PUL encoding PL22 and PL26 polysaccharide lyases, a GH88 rhamnogalacturonyl hydrolase and several rhamnose-processing genes, allowing growth with rhamnogalacturonan as sole carbon source (Fig. 4B). A PL26-GH88 pair also occurs in the rhamnogalacturonan-degrading flavobacterium Gramella flava 88 , indicating co-functionality towards rhamnose-rich polysaccharides. As rhamnogalacturonan is present in widespread marine macroalgae 74 , degradative abilities may strengthen associations between MIT1002 and phototrophs. Homologous PUL in A. australica with 80% nucleotide identity (data not shown) demonstrates independent acquisition of these genes by other Alteromonas species, comparable to PUL targeting ulvan from green algae 89,90 . Strain AD45 harbors a unique PUL encoding GH85 and GH92 mannosidases and grows with alpha-mannan as sole carbon source (Fig. 4C), but comparable growth of MIT1002 indicates that mannosidase activity also occurs via other encoded GHs (Fig. S3A). Opposed to mannan-degrading marine flavobacteria 91 , strain AD45 does not encode sulfatases and may hence primarily target terrestrial mannans, corresponding to its coastal origin 92 and the lower degree of sulfatation in terrestrial polysaccharides 93 . A speculative link relates to the isolation of AD45 from the vicinity of aquaculture facilities, where mannan oligosaccharides are increasingly used as feed additive 94 . Overall, presence in diverse terrestrial and aquatic bacteria (Fig. S3C) suggests the PUL as a widespread niche-defining feature.
Finally, we found that adaptation towards algal polysaccharide degradation is also linked to numerical variation in CAZymes, in context of gene dosage effect and substrate affinity 18 . Specifically, A. macleodii strains that encode three PL1 pectate lyases grow significantly better on pectin than strains with only two lyases (Figs. 4D; S3A). Enrichment of the third lyase in the exoproteome of strain 83-1 74 suggests a role in extracellular substrate recognition and initial hydrolysis. Enhanced degradation through higher lyase numbers is consistent with observations in Zobellia galactanivorans, a common macroalgal associate and proficient polysaccharide degrader 78 . Overall, the patchy distribution of rhamnogalacturonan, mannan and pectin degradation discriminates A. macleodii into specific "polysaccharide utilization types" with distinct ecophysiological roles 95 . cellular communication. Ecological differentiation can also coincide with the potential to coordinate behavior at population level. In this context, we found that A. macleodii strains vary in their ability to synthesize homoserine lactones (HSL) for intraspecific communication via quorum sensing 96 . Two gene variants encoding N-acyl amino acid synthase occur in A. macleodii (Fig. 5A), but masses corresponding to C6-HSL, 3-oxo-C12-HSL and dodecanamide (the fatty acid moiety of 3-oxo-C12-HSL) were only detected in exometabolomes of strains 27126 T , HOT1A3 and MIT1002 ( Table 1). The restriction of HSL production to these strains is supported by antismash 97 , which only predicts their sequence variant as functional synthase (Table S1). Accordingly, the autoinducer domain of producers and non-producers has <80% sequence identity (data not shown). Synthase sequences of 27126 T , HOT1A3 and MIT1002 contain different substitutions (Fig. 5A), which potentially explains the observation of HSLs with differing chain lengths 98 . HSLs were only detectable using highly sensitive FT-ICR-MS but not standard bacterial monitor assays 99 , but HSLs can influence chemical interactions and surface attachment even at low concentrations 96,100 . Intraspecific HSL diversity has also been described among symbiotic Vibrio 101 , suggesting variable potential for chemical communication as common discriminator of closely related strains. iron acquisition. Successful niche colonization also depends on efficient acquisition of limiting micronutrients, including iron 102 . In this context, only strains 27126 T , EC673 and Te101 harbor a gene cluster for siderophore synthesis with demonstrated iron-scavenging activity (Fig. 5B), likely providing an advantage during iron limitation 103 . The gene cluster is homologous to the petrobactin operon of Bacillus spp. (Fig. 3B; 35% amino acid identity) and also occurs in other marine bacteria, suggesting broad ecological relevance 104 . In strain EC673 from the English Channel, the siderophore might support growth with benzoate as sole carbon source (Figs. 3C, S2) by counteracting iron limitation of benzoate breakdown 105 . This scenario could be advantageous considering the anthropogenic input of benzoate in its original habitat 106 . The benzoate cluster is located in a genomic island 32 and flanked by a transposase, underlining the importance of flexible loci for phenotypic variability. Notably, also A. mediterranea EC615 from the English Channel harbors the benzoate-related cluster (Fig. 3C), but encoded on a plasmid 38 . These observations indicate common occurrence and exchange of these genes via mobile genetic elements in habitats where certain chemicals may prevail.
Strain MIT1002 harbors a truncated siderophore cluster, where synthases have been separated by metal-resistance cassettes during the translocation from plasmid to genome (see above). This integration www.nature.com/scientificreports www.nature.com/scientificreports/ abolished iron-scavenging activity (Fig. 5B), showing that genetic exchange and restructuring of genomic islands can also be disadvantageous.

Implications for intraspecific interactions and biogeography.
To address broader eco-evolutionary implications, we asked whether strain-level variability affects population dynamics, competitive abilities and biogeographic distribution 107,108 . For instance, it is known that natural populations of A. macleodii can be dominated by specific strains through competitive exclusion 34,109 . To evaluate these aspects, three A. macleodii strains with comparable growth in monoculture (Fig. S4) were co-cultured with glucose as sole carbon source, and individual population sizes determined by quantitative PCR of unique genes (Table S5). The tripartite co-culture was dominated by strain MIT1002, which outcompeted both 83-1 and 27126 T over a period of 24 h (p < 0.01). Furthermore, strain 83-1 outcompeted 27126 T in late exponential phase (p < 0.001) (Fig. 6A). Comparable intraspecific differences were also observed in A. mediterranea, where greater competitive abilities coincided with higher motility 35 . The putative importance of motility in microbial interactions is supported by upregulation of related genes in MIT1002 when co-cultured with Prochlorococcus 62 .
Higher competitiveness of MIT1002 on glucose may provide an advantage in the environment, as glucose is one of the major marine carbohydrates 110 . Accordingly, MIT1002 showed a wider geographic distribution in TARA Ocean metagenomes (Fig. 6B, Table S6), indicating linkages between metabolic abilities and biogeography. Contact with diverse microbiota in different locations may also explain why unique genes of MIT1002 have been acquired from a wider taxonomic range ( Fig. 6C; Table S7). These patterns may be amplified by association with Prochlorococcus, considering the wide occurrence of the cyanobacterium and higher genetic exchange in host-associated niches 111,112 . In contrast, 27126 T has been isolated from oligotrophic waters with less biological activity and genetic exchange 113 , and lower growth efficiency on glucose may indicate a k-strategist lifestyle. Future co-culturing systems could address how co-existence or competitive exclusion proceed in more complex ecological scenarios, for instance pioneer-scavenger relationships during polysaccharide degradation 18 .  Table 1). Strain-specific amino acid substitutions (red) may explain differential HSL production (synthase locus tags in parentheses). (B) Gene cluster unique to strains 27126 T , EC673 and Te101 encoding a functional siderophore (locus tags from type strain), with iron-scavenging activity under iron-deplete (Fe−) but not iron-replete (Fe+) conditions in relation to deferoxamine mesylate (DFOM) standard. MIT1002 harbors a nonfunctional cluster after insertion of gene cassettes for cobalt-zinc-cadmium and mercury resistance. www.nature.com/scientificreports www.nature.com/scientificreports/ conclusions Here, we extend existing knowledge on (pan)genome evolution and structure in Alteromonas by functional perspectives on genome plasticity in twelve A. macleodii strains. The shown range of ecological strategies demonstrates that single genospecies can encompass considerable diversity of adaptive features, underlining the importance of polyphasic studies that link bacterial genotypes and phenotypes 114 . The "ecological microdiversity" among strains with >99% 16S rRNA gene identity should be emphasized in microbial diversity studies, which are only beginning to explore the extent of fine-scale variability in natural communities 36 . Notably, phylogenetic relationships only partially corresponded to ecological similarity, illustrated by the patchy distribution of niche-defining metabolic features. Hence, in line with common recombination and genetic exchange 30 , A. macleodii appears to perform constant "pathway sampling" that has not (yet) manifested in divergence of specific clades. Metabolic versatility probably facilitates flexible responses to environmental conditions, contributing to the feast-and-famine lifestyle and widespread occurrence of this marine bacterium 24,30 . Sequencing of additional genomes may reveal whether strain-specific abilities translate to the existence of phylogenetic clades with distinct ecological boundaries, corresponding to larger eco-evolutionary concepts 1,115,116 . Our functional-ecological interpretation of the A. macleodii pangenome, illustrating the extent of eco-genomic differentiation within bacterial species, has broader implications for niche specialization, microbial interactions and biochemical roles of marine bacteria.

Materials and Methods
isolation and sequencing of Alteromonas macleodii strains. Strains BGP6, BGP9 and BGP14 were isolated from alginate-enriched seawater from the south, equatorial and north Pacific Ocean on expedition SO248 with RV Sonne 117 . The genomes of BGP strains, MIT1002 and EZ55 were sequenced de novo using PacBio II technology (Supplementary Methods). In addition, a number of published closed genomes were analyzed (Table S1). pangenomic and phylogenetic analyses. Core, accessory and unique genes (Table S2) were identified using anvi' o v5.2 118 following the pangenome workflow of Delmont and coworkers 23 with minbit parameter 0.5, MCL inflation parameter 10, Euclidean distance and Ward linkage, and NCBI-BLASTp for sequence similarity analysis (see Supplementary Methods for details). For phylogenetic analysis, 92 single-copy core genes (https://help.ezbiocloud.net/ubcg-gene-set) were identified, extracted and aligned using the UBCG pipeline 119 with Alteromonas stellipolaris LMG21861 T as outgroup. The alignment was manually checked and submitted to  Table S6 for details). (C) Closest relatives of unique genes from strains MIT, 83-1 and 27126 T based on BLAST against NCBI RefSeq. Alt: Alteromonadaceae; Psalt: Pseudoalteromonadaceae; Vib: Vibrionaceae; Oce: Oceanospirillaceae; Pse: Pseudomonadaceae; CV: Cellvibrionales (see Table S7 for details).
exometabolomics. All cultivations were done in triplicate using SWM seawater minimal medium 136 . Each replicate was inoculated at 1% (v/v) with precultures grown in 10 mL SWM + 0.1% glucose for 24 h at 20 °C and 140 rpm (washed twice with sterile SWM and diluted to OD600 of 0.1 before inoculation). For exometabolomics, nine strains were inoculated in 50 mL SWM + 0.1% glucose at 0.5% (v/v) in triplicate. After incubation at 20 °C and 140 rpm until late exponential phase, a 20 mL subsample from each replicate was centrifuged for 20 min at 3500 g and 4 °C. In addition, three sterile media blanks were incubated and processed in the same manner. Exometabolites were purified from supernatants using solid phase cartridges 137 followed by ultrahigh-resolution mass spectrometry 138,139 on a 15 T Solarix Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR-MS) in negative mode (Supplementary Methods). Only peaks present in two biological replicates were considered, and only if detected in technical duplicates measured per replicate. Furthermore, spectra were calibrated and denoised using strict procedures to ensure that only bacterial metabolites were evaluated (Table S3). Tentative identification of masses was done using databases MetaCyc 140 and KEGG Compounds via R package KEGGREST 141,142 . Degradation of different substrates. Degradation of specific carbon sources was tested in SWM supplemented with phenol (final concentration 0.5 mM), toluene (1 mM), xylene (1 mM), sodium benzoate (2 mM), alpha-mannan (Carbosynth YM63069; 0.1% w/v), rhamnogalacturonan (Megazyme P-RHAM1; 0.1% w/v), or pectin (Fluka 76282; 0.1% w/v). Cultures were inoculated with precultures as described above and evaluated by photometry (OD600) or colony-forming units (log CFU mL −1 ) after plating serial dilutions on marine agar (cultures with aromatics subcultured twice before plating). In addition, strain 83-1 was tested for degradation of macroalgal tissue (Supplementary Methods). Briefly, healthy specimens of the brown macroalga Ecklonia radiata were incubated with strain 83-1 for 12 days and tissue degradation evaluated in comparison to a control without bacterial addition (n = 15).
Screening for bioactive secondary metabolites. Siderophore production was tested with sterile-filtered supernatants of overnight cultures in iron-deplete vs. iron-replete minimal medium using a modified CAS assay 143,144 with 50 µM deferoxamine mesylate (DFOM) and sterile medium as positive and negative controls, respectively. Activity was quantified against a seven-point DFOM standard curve (R² = 0.981). Production of HSLs was tested by streaking Alteromonas colony mass in parallel to the biosensor strains Chromobacterium violaceum CV026 and Agrobacterium tumefaciens A136 according to Ravn and coworkers 99 , with Phaeobacter inhibens DSM17395 as positive control.
co-culture and quantitative pcR of unique genes. Quantitative PCR (qPCR) was performed using a LightCycler 480 (Roche, Switzerland) according to Berger and coworkers 145 . For a unique gene of each A. macleodii strain, primers were designed using the Roche Universal Probe Library and ordered from TIB MolBiol Germany (Table S5). After confirmation of primer specificity against target and non-target strains, selected strains were grown as mono-and co-cultures in triplicate (inoculated with precultures as described above) in SWM + 0.1% glucose at 100 rpm and 20 °C. DNA was extracted using the Master Pure RNA Purification Kit (Epicentre, Madison, WI) and amplified in 15 µL qPCR reactions (each 10 µL of LightCycler 480 Probes Master, 3 µL PCR-H 2 O, 400 nM of each primer, 200 nM of the respective UPL probe and 5 µL template adjusted to 10 ng µL −1 ). Cycling conditions were 95 °C for 10 min, 45 cycles (95 °C for 10 s, 60 °C for 30 s, 72 °C for 1 s) and 40 °C for 30 s. For each biological replicate, three technical PCR replicates were run. Growth was expressed as DNA equivalents in relation to a five-point DNA standard curve for each strain (R 2 > 0.98).
Biogeography and taxonomic relatives of unique genes. Three genomic loci specific for strains MIT1002, 83-1 and 27126 T (Table S6) were searched against TARA Ocean metagenomes using the Sequenceserved-based web application at http://bioinfo.szn.it/tara-blast-server 146 . Detection was considered positive if at least one gene from two loci was detected with >99% identity and >70% query coverage. Furthermore, unique genes were searched against the NCBI RefSeq Protein database to identify the closest taxonomic relative.

Data availability
Complete genomes have been deposited at EMBL-EBI under study PRJEB32335 and are also available at IMG 147 under accession numbers 2738541260, 2738541261, 2738541262, 2738541267 and 2785510739, respectively.