Bradymonabacteria, a novel bacterial predator group with versatile survival strategies in saline environments

Background Bacterial predation is an important selective force in microbial community structure and dynamics. However, only a limited number of predatory bacteria have been reported, and their predatory strategies and evolutionary adaptations remain elusive. We recently isolated a novel group of bacterial predators, Bradymonabacteria, representative of the novel order Bradymonadales in δ-Proteobacteria. Compared with those of other bacterial predators (e.g., Myxococcales and Bdellovibrionales), the predatory and living strategies of Bradymonadales are still largely unknown. Results Based on individual coculture of Bradymonabacteria with 281 prey bacteria, Bradymonabacteria preyed on diverse bacteria but had a high preference for Bacteroidetes. Genomic analysis of 13 recently sequenced Bradymonabacteria indicated that these bacteria had conspicuous metabolic deficiencies, but they could synthesize many polymers, such as polyphosphate and polyhydroxyalkanoates. Dual transcriptome analysis of cocultures of Bradymonabacteria and prey suggested a potential contact-dependent predation mechanism. Comparative genomic analysis with 24 other bacterial predators indicated that Bradymonabacteria had different predatory and living strategies. Furthermore, we identified Bradymonadales from 1552 publicly available 16S rRNA amplicon sequencing samples, indicating that Bradymonadales was widely distributed and highly abundant in saline environments. Phylogenetic analysis showed that there may be six subgroups in this order; each subgroup occupied a different habitat. Conclusions Bradymonabacteria have unique living strategies that are transitional between the “obligate” and the so-called facultative predators. Thus, we propose a framework to categorize the current bacterial predators into 3 groups: (i) obligate predators (completely prey-dependent), (ii) facultative predators (facultatively prey-dependent), and (iii) opportunistic predators (prey-independent). Our findings provide an ecological and evolutionary framework for Bradymonadales and highlight their potential ecological roles in saline environments. Video abstract.

Bradymonabacteria are representative of the novel order Bradymonadales, which are phylogenetically located in the δ-Proteobacteria [17]. The first type species of Bradymonadales, Bradymonas sediminis FA350 T , was isolated in 2015 [17]. To date, 9 strains within the Bradymonadales have been isolated and found to belong to 7 candidate novel species; these Bradymonabacteria are bacterial predators [18]. Interestingly, the phylum Proteobacteria contains three orders of predatory bacteria. Among them, Myxococcales and Bradymonadales belong to δ-Proteobacteria, while Bdellovibrionales were classified as Oligoflexia in 2017 [19]. Myxococcales and Bdellovibrionales are facultative and obligate predators, respectively. Additionally, they have different distribution patterns in the environment. Myxococcales are mainly found in soil and sediment niches [20,21], while Bdellovibrionales are aquatic. However, how Bradymonadales adapt to predatory lifestyles and whether they have specific living strategies or ecological importance remain largely unknown.
Here, we analyzed the predation range of Bradymonadales on diverse bacteria and their predatory morphological and physiological characteristics. By using comparative genomic analysis of Bradymonadales and other predatory bacteria, we revealed the genetic and metabolic potential of this group. To assess the diversity and frequency of occurrence of the various ribotypes of known predators (Bradymonadales, Myxococcales and In total, 9 strains of bacteria in the novel order Bradymonadales were isolated using the enrichment culture method [22]. Among these strains, FA350 T [17,18] and B210 T [23] were the two type strains for different genera in Bradymonadales. Both these type strains were used to investigate the predator-prey range of Bradymonabacteria. A total of 281 isolated bacteria were cocultured with Bradymonabacteria FA350 T [17,18] or B210 T [23] as lawns in individual Petri dishes (Fig. 1a, Table S1). Zones of predation were measured ( Fig. 1b), and the results showed that the Bradymonabacteria preyed on diverse bacteria but showed a strong preference for Bacteroidetes (90% of tested bacteria could be preyed on) and Proteobacteria (71% of tested bacteria could be preyed on) (Fig. 1c). Predation on bacteria in the orders Flavobacteriales, Caulobacterales, Propionibacteriales, and Pseudomonadales was broadly distributed, with a mean predation percentage greater than 90%, while predation of Micrococcales and Enterobacteriales was less efficient.
Transmission electronic microscopy (TEM) and scanning electronic microscopy (SEM) analyses were performed to understand the mechanism of predation of strain FA350 T on the subcellular level. Lysis of the prey cells was detected near strain FA350 T in both the TEM and SEM analyses (Fig. 2). Strain FA350 T was found to have pili (Figs. 2b and 2g) and outer membrane vesicle (OMV)-like structures (Figs. 2d, 2e, 2f, and 2h). In addition, FA350 T cells contained intracellular particles with low electron density (Figs. 2b, 2c, 2d, and 2f), which were shown to contain polyhydroxyalkanoates (PHAs) by Nile blue A staining. FA350 T cells also contained several electron-dense (black) spots (Figs. 2b, 2c, 2d, and 2f), which indicated the presence of intracellular polyphosphate granules [24].
Both of these particle types significantly accumulated during predation (Fig. 2).

Bradymonabacteria are polyauxotrophs
To explore the metabolic capabilities and predation mechanism of this novel group, we analyzed 13 genomes of Bradymonadales (9 high-quality genomes sequenced from Almost all strains (except FA350 T ) possessed a minimal pentose phosphate pathway, which lacked key steps for the synthesis of ribose 5-phosphate (Fig. 3, Table S2) [27].
Most of the bradymonabacterial genomes lacked key enzymes for pyrimidine synthesis, such as aspartate carbamoyltransferase, which catalyzes the first step in the pyrimidine biosynthetic pathway. All genomes lacked the complete purine de novo pathway; they were missing the phosphoribosylaminoimidazole carboxylase catalytic subunit or even the whole pathway.
In addition to this auxotrophy in the synthesis of pentose and nucleotides, all the genomes lacked complete pathways for the synthesis of many amino acids, such as serine, methionine, valine, leucine, isoleucine, histidine, tryptophan, tyrosine, and phenylalanine ( Fig. 3). For example, all the genomes encoded a potential D-3-phosphoglycerate dehydrogenase for the conversion of glycerate-3P into 3-phosphonooxypyruvate for aminoacid synthesis (Fig. 3). However, in all members of Bradymonabacteria, this pathway appeared to be blocked at the subsequent step because of the absence of phosphoserine aminotransferase, although Bradymonabacteria could continue with subsequent pathways to complete the biosynthesis of cysteine and glycine. Additionally, many cofactors and vitamins that promote bacterial growth [22], such as biotin, thiamin, ubiquinone, VB 12 , and VB 6, could not be synthesized by the de novo pathway in almost all the genomes.

Dual-transcriptome analysis of the potential predation mechanism of Bradymonabacteria
To further determine the genes involved in predation, we performed dual-transcriptome analysis of Bradymonas sediminis FA350 T with and without preying on Algoriphagus marinus am2 (Fig. S2). As with obligate predators, one way that Bradymonabacteria kill their prey bacteria is likely by using contact-dependent mechanisms. Here, the bradymonabacterial genomes possessed complete Type IV pili (T4P) (Fig. 3), and the attached areas showed more type IV pili than the unattached areas (SEM, Figs. 2g and 2h). The dual-transcriptome analysis showed that genes encoding the type IV pili twitching motility protein PilT (DN745_17255) were significantly upregulated during predation (Fig.   S3), suggesting that these genes may be involved in predation. Bradymonabacteria also had T4b pilins showing homology to those in Bdellovibrio bacteriovorus HD100, in which T4b pilins are necessary for predation [28,29] (Fig. S4), so T4b pilins may also participate in regulating predation in Bradymonabacteria. In addition, this group of bacteria had type II and type III secretion systems (the YscRSTUV proteins that form a membrane-embedded complex known as the ''export apparatus'' [30]). The dual-transcriptome analysis also supported the prediction that genes encoding the type III secretion system innermembrane protein complex (DN745_01900, DN745_10315, DN745_17280, DN745_03325, and DN745_00480) were significantly upregulated during predation (Fig. S3), implying that these genes may also be involved in predation.
Another way that Bradymonabacteria kill their prey bacteria is likely by secreting antimicrobial substances into the surrounding environment. As in most facultative bacterial predators, a few potential antimicrobial clusters of secondary metabolite synthesis, such as Lasso-peptide [31], were found in almost all genomes of Bradymonabacteria (Fig. 3). Genes involved in OMV-like biosynthesis were also detected in most genomes, such as ompA (cell envelope biogenesis protein), envC (Murein hydrolase activator) and tolR (envelope stability) [32]. Vesicle membrane-related genes (DN745_03865, DN745_02930, and DN745_07125) were significantly upregulated during predation (Table S4, Fig. S3).

Bradymonabacteria are novel predators different from obligate or facultative predators
Comparative genomic analysis with other bacterial predators was performed to explore whether Bradymonabacteria have a unique living strategy. Two-way cluster analysis showed that bradymonabacterial genomes contained features different from those of either obligate or facultative predators, which were phylogenetically located in a different branch (Fig. 4). The specific multiple metabolic deficiencies of Bradymonabacteria had some similarities to those of most obligate predators. For example, both Bradymonabacteria and obligate predators possessed a minimal pentose phosphate pathway, lacked key enzymes for pyrimidine synthesis, and lacked complete pathways for the synthesis of many amino acids, cofactors, and vitamins ( Fig. 4). However, Bradymonabacteria with multiple auxotrophies could grow on common media (such as marine agar medium), though at a low growth rate [33], unlike obligate predators.
Unlike most obligate predators, the polyphosphate accumulation pathway, containing a pair of genes (Polyphosphate kinase and Exopolyphosphatase) associated with both polyphosphate formation and degradation [34], was present in most Bradymonabacteria ( Fig. 4). Polyphosphate accumulation was also detected in FA350 T cells during predation ( Fig. 2). In contrast to most of the other predator genomes, potential PHA synthesis from β-oxidation of fatty acids [35] was observed in most bradymonabacterial genomes (Fig. 3).
In this study, TEM analysis showed that strain FA350 T could significantly accumulate PHAs during predation compared with pure culture (Fig. 2). Despite their incomplete fatty acid biosynthetic pathway, all Bradymonabacteria had a high copy number of long-chain fatty acid transporters (fadL) compared to those of other predators, allowing them to gather fatty acids from the environment (Fig. 4). In addition, genes associated with alkane synthesis, which is important for maintaining cell membrane integrity and adapting to cold environments [36], were present in most genomes of Bradymonabacteria (Figs. 3 and 4).
Thus, we proposed that Bradymonabacteria could be categorized as novel predators different from so-called obligate or facultative predators (Table 1).

Bradymonadales are mainly distributed in saline environments with high diversity
To evaluate the global prevalence of the Bradymonadales order, we surveyed recently published 16S rRNA gene amplicon studies that provided high taxonomic resolution along with relative sequence abundances. The 16S rRNA gene amplicons from 1552 samples were grouped into eight types of environments ( Fig. 5a and Table S5). A total of 811 samples were from inland environments, while others were from marine environments, with each biotope showing a somewhat different microbial community (Figs. 5b and S5).
Bradymonabacteria was detected in 348 of 741 marine samples (relative abundance>0.01%) but only 20 of 544 soil samples (Fig. 5a). All samples were sorted into an ordination diagram based on the similarity of communities (Fig. 5b). Saline biotopes were clearly separated from nonsaline biotopes (Fig. S6), suggesting that salinity was a significant factor in shaping microbial communities. For each biotope, the relative abundance of Bradymonadales in the saline environments (i.e., seawater and saline lake sediment) was significantly higher than that in the nonsaline environments (i.e., nonsaline soil and nonsaline water) (P<=0.0001, Fig. 5c). The distribution analysis was consistent with the genomic feature analysis (Fig. 2), in which several genes encoding sodium symporters and Na + /H + antiporters were found in the genomes, suggesting a beneficial effect of salinity on Bradymonabacteria.
In addition, we compared the relative abundance of Bradymonadales with those of two orders of well-known predatory bacteria, Bdellovibrionales and Myxococcales [12,37,38].
We found that Myxococcales and Bdellovibrionales were also globally distributed ( To explore the diversity and distinct evolution of bradymonabacterial subgroups in different biotopes, we performed a phylogenetic analysis of nearly full-length 16S rRNA gene sequences of diverse origin by maximum likelihood inference (Table S6). A total of 187 OTUs were detected and found to form six sequence clusters (Fig. 6a). Almost 87.2% of the representative sequences originated from saline biotopes (such as seawater, marine sediments, salterns, corals, and saline lakes). Since bradymonabacterial subgroups may be selectively distributed in local biotopes, we investigated the relative abundance of each subgroup throughout the 127 representative samples in which the relative abundance of Bradymonadales was above 1% (Fig. 6b). Five of the 6 bradymonabacterial subgroups showed significantly higher abundance in saline environments. Cluster-2 and cluster-6 were mainly observed in seawater biotopes, whereas cluster-3 was mainly observed in marine sediment and saline lake sediment (Fig. 6b), consistent with the environments of the cultured strains. Cluster-5 lineages tended to occur in both freshwater and seawater biotopes (Fig. 6b).

Discussion
In all ecosystems, predation is an important interaction among living organisms. Bacterial predators are proposed to play an important role in controlling and shaping bacterial populations in diverse environments [3,39]. However, despite their ecological importance, only a few examples of predatory bacteria have been studied in depth. Recently, many predatory bacteria from various phyla have been isolated from different environments; however, most of their predatory lifestyle strategies and adaptations remain unclear. This study systematically analyzed the predatory lifestyle adaptations, global distribution, and diversity of Bradymonadales; highlighted the ecological role of Bradymonadales; and provided a framework for the categorization of the known predatory bacteria.
In our study, based on comparative genomic and physiological analyses, Bradymonabacteria were shown to be a novel group of bacterial predators with versatile survival strategies different from those of either so-called "obligate" predators or "facultative" predators (Table 1). Furthermore, Bradymonabacteria had multiple metabolic deficiencies. Their incomplete pathways might be important for prey-dependent growth, as the precursor compounds could be acquired from predation. In addition, the loss of genes in the fatty acid biosynthetic pathway was notable, because fatty acids are integral components of the cellular membrane, and their synthesis is considered to be a housekeeping function of cells [40]. Thus, these organisms may incorporate exogenous fatty acids from prey bacteria into their membrane phospholipids using their high copy number of long-chain fatty acid transport proteins [41] (Fig. 3). The gene loss in these organisms may render them dependent on prey for their lost metabolic functions and may also provide a selective advantage by conserving predators' limited resources [42].
However, the sequenced genomes of Bradymonabacteria were surprisingly large (5.0 Mb to 8.0 Mb, Fig. S1a), suggesting that Bradymonabacteria are far from obligate parasites, with seemingly none of the reductive evolution that results from a parasitic lifestyle in bacteria such as Mycobacterium leprae [43]. The large size of their genomes may be indicative of the vast range of genes required for Bradymonabacteria to both effectively tolerate the absence of prey and carry out predation.
In contrast to most predators, Bradymonabacteria can synthesize many nutrient polymers, such as polyphosphate, PHA, and alkane molecules. Exopolyphosphatase catalyzes the hydrolysis of terminal phosphate residues from polyphosphate chains, accompanying the production of ATP and thus playing a role in the production of energy [44].
Bradymonadales cells may accumulate polyphosphate in the phosphate-rich zone, using it as an energy source [45]. Meanwhile, PHA granules are synthesized as sinks of excess carbon and are used as carbon and energy reserves in starvation conditions [46]. Under nutrient starvation, maintenance energy and free amino acids can be provided by endogenous substrates such as PHAs and polyphosphate [47,48]; this ability may be an important feature for the survival of Bradymonabacteria during intervals without predation. This feature is interesting among bacterial predators, as it is commonly found in animal predation. For example, the bear can store fat in its body to ensure that it will survive the long winter. In addition, Bradymonabacteria may also synthesize alkanes to maintain cell membrane integrity [36] and complement its poor fatty acid synthesis ability. Thus, these multiple auxotrophies and nutrient synthesis polymers confer on Bradymonabacteria a versatile survival strategy for natural environments, in contrast to the currently known obligate or facultative predators.
As bacterial predators, Bradymonabacteria have developed a wide range of mechanisms to attack their prey. Contact-dependent predation mechanisms allow predators to attach to the prey and then carry out predation. This process has a relatively low energy cost and could prevent secretory virulence factors from being diluted by the surrounding environment [49]. Bradymonabacteria also has T4P, which could pull adherent bacteria into close association with other bacteria [50]. T4P could also transport bound substrates such as DNA [51] into the periplasm and export exoproteins across the outer membrane [52]. Contact-dependent Type III secretion systems have also been found in Bradymonabacteria and are reported to be capable of moving virulence factors across bacterial outer membranes and directly across the host cell membrane into the cytoplasm of a host cell [53]. However, no reports have indicated that the type III secretion system is involved in direct combat between bacteria. Whether type III secretory complexes could penetrate the bacterial cell wall is unknown. Further gene knockout experiments and systematic TEM analysis should be performed to identify whether and how the type III secretion system works during predation.
Our biogeographic analysis suggested that Bradymonabacteria are mainly distributed in saline environments, and some other studies have also detected Bradymonadales in hypersaline soda lake sediments [25], suggesting that saline environments could be enriched in these bacteria. Our genome analysis also showed that Bradymonadales had many genes encoding sodium symporters and Na + /H + antiporters to maintain osmotic pressure in saline environments. These findings supported the global analysis (Figs. 5 and S8c), suggesting that Bradymonadales might be a dominant bacterial predator in some specific saline environments compared with Bdellovibrionales and Myxococcales. The analysis of the complex intragroup phylogeny of the 6 subgroups of Bradymonabacteria revealed that distinct evolutionary bradymonabacterial subgroups had arisen in different biotopes, suggesting the occurrence of adaptive evolution specific to each habitat.
Patterns related to salinity status also suggest that most Bradymonadales are halophiles [17].

Conclusion
The unique metabolic pathways of Bradymonabacteria, which include conspicuous metabolic deficiencies similar to those of obligate predators but with a more effective starvation stress response mechanism, provide these bacteria with a unique survival strategy (different from the survival models of so-called "obligate" or "facultative" predators). We propose a framework to categorize the current bacterial predators into 3 groups: (i) highly prey-dependent predators, such as most of the BALOs; (ii) facultatively prey-dependent predators, such as Bradymonabacteria; and (iii) prey-independent predators, such as Myxobacteria and Lysobacter sp. (Table 1). This categorization is helpful for further study of the different ecological importances of each type of bacterial predator. The evolution of bacterial predation in these three groups of predators should also be studied in the future to better understand the significance of predation to biological evolution.

Predation experiments
To explore the predation of Bradymonabacteria, we used Bradymonas sediminis FA350 T and Lujinxingia litoralis B210 T as representative strains. All candidate prey strains were obtained from our laboratory. Cells were centrifuged, washed and concentrated in sea water to a final OD 600 of 3.0 for predator strains and 6.0 for candidate prey strains. Drops of 5.0 μl of the predator strain suspensions were deposited on the surfaces of agar plates and allowed to dry. Next, 20.0 μl drops of each different candidate prey strain suspension were placed near the predator spot. The plates were incubated at 33 °C, and images were taken after 48 h with a digital camera. To detect PHA accumulation, the granules were stained with the Nile red component of Nile blue A.

Genome sequencing and comparative genome analyses
To explore the potential metabolic capacity of bacterial predators, we sequenced 3 complete genomes and 6 draft genomes of all currently known Bradymonabacteria isolate strains using the methods reported in our previous studies [18,55]. We also retrieved 37 predator genomes from NCBI (including 4 metagenome assembled genomes). tRNA and gene prediction were performed using tRNAscan and prodigal, respectively. The genomebased metabolic potential of the bacterial predators was predicted by BlastKOALA (https://www.kegg.jp/blastkoala/). The average nucleic acid identities among the 9 cultured Bradymonabacteria strains were calculated using pyani (https://github.com/widdowquinn/pyani), and the percentage of conserved proteins (POCP) in each strain was calculated as described previously by Qin et al. [56]).

Electronic microscopy analyses
We selected Algoriphagus marinus am2, which is smaller than the predator Bradymonas sediminis FA350 T , as prey. Bradymonas sediminis FA350 T and Algoriphagus marinus am2 were cultured separately to the exponential growth phase, adjusted to the same OD value, mixed together and cocultured on marine agar medium at 33 °C for 68 h.

Dual transcriptomic analyses
To determine the gene expression profiles of the type strain FA350 T , a pure culture of FA350 T and a coculture of FA350 T with the prey Algoriphagus marinus am2 were cultured on marine agar medium at 33 °C for 0 h, 68 h, and 120 h, respectively. Each time point was collected in triplicate (n=3) for further transcriptomic analysis. For the transcriptomic analysis, the RNA extraction, library construction, sequencing and analysis were performed as described in our previous study [57]. Sequencing was carried out on a HiSeq sequencer at Novogene Co., Ltd. (Beijing, China). For transcriptomic analysis of mixed culture samples (dual transcriptomic analysis), total RNA sequences were mapped to the complete genome of FA350 T using the method reported by Westermann et al. [58]. Then, the completely mapped sequences were selected for further analysis.

Phylogenetic analysis of bradymonabacterial type IV pili
An unrooted, maximum likelihood phylogeny shows relationships between the type IVa, type IVb, and type IVc pili and the archaellum (archaeal flagellum) and the T2SS and T4SS extension ATPases. The protein amino-acid sequences were aligned with mafft and used to estimate a maximum likelihood phylogeny with RAxML under the JTT substitution model with gamma-distributed rate variation. The protein amino-acid sequences of Bradymonas were annotated by RAST (Rapid Annotation using Subsystem Technology) [59]. The other protein amino-acid sequences were obtained from other research supplementary materials [60].

Biogeographic distribution database construction
All the 16S rRNA gene sequences analyzed in this paper were downloaded from the  (Table S5).

Microbial community composition
The raw 16S rRNA gene reads were filtered with UCHIME. Quality filtering, chimera detection, dereplication, clustering into OTUs and assigning taxonomic information were performed using VSEARCH [61]. The SILVA database Ref_SSU release 132 was used as a reference taxonomic database (https://www.arb-silva.de/). Alpha diversity indices (Shannon, Simpson, Good's coverage and Ace) detailing the microbial community composition within each sample were calculated using scikit-bio (http://scikit-bio.org/) in Python, and alpha diversity indices (Chao1) were calculated using the package fossil (https://www.rdocumentation.org/packages/fossil) in R. For estimating community dissimilarities, Bray-Curtis distance was calculated by vegan in R based on the relative abundance of each taxon at the order level.

Phylogenetic analyses
Both RAxML [62] and FastTree [63] were employed to construct the Bradymonabacteria phylogenetic tree. Given both the topology of the phylogenetic tree and its good coverage of all Bradymonabacteria lineages, we established the phylogenetic tree using 187 representative Bradymonabacteria 16S rRNA gene sequences, which were all longer than 1200 bp (at 98.5% cutoff). These sequences were aligned using mafft. The Bradymonabacteria subgroup designations were confirmed when one subgroup with > 10 representative sequences was monophyletic by two phylogenetic trees constructed by different programs using the maximum likelihood approach [64]. The environmental type (i.e., saline and nonsaline) of each Bradymonabacteria sequence in the tree was collected from GenBank. A genome-based phylogeny of bacterial predators and 9 cultured Bradymonabacteria strains was constructed using core genes [65], and trees were constructed using RAxML [62]. All phylogenetic trees were drawn using ggtree [66] in R.

Quantitative real-time PCR
The environmental DNA samples extracted in the previous step were used for qPCR experiments in order to detect the abundance of bacteria and Bradymonadales in each sample. The primer pair composed of 341F (5′-CCTACGGGAGGCAGCAG-3′) and 534R (5′-ATTACCGCGGCTGCTGGCA-3′) was used for quantification of bacteria [22]. A Bradymonadales-specific primer set composed of qBRA1295F (5′-CTCAGTTCGGATYGYAGTCTG-3′) and qBRA1420R (5′-GTCACCGACTTCTGGAGCAARCG-3′), which was designed in this study and generated an amplicon of 148 bases, was used for Plasmid Purification Kit (TaKaRa). DNA copy number was determined by the concentration and relative molecular weight of the Plasmid DNA. For each QPCR assay, the plasmid aliquot was serially diluted to produce concentrations ranging from 10 9 to 10 3 DNA copies/ μl to generate calibration curves. Each sample was measured in triplicate, and negative controls (no template NTC) were included.

Consent for publications
Not applicable Bradymonabacteral isolate is available upon request.

Competing interests
No conflict of interest exists in the submission of this manuscript, and the manuscript has been approved by all authors for publication. The authors declare that they have no competing interests.

Authors' contributions
DSM, GJC, JZ, and ZJD designed the study. SW carried out TEM, SEM, and transcriptome analyses. ZZD carried out FISH and real-time PCR analysis. DSM, QYL, SW, XPW, and RT performed bioinformatic analyses. DSM and ZJD analyzed data and wrote the paper. JYN, AZ and YY improved the paper writing. All authors read and approved the manuscript   TEM and SEM micrographs of Bradymonas sediminis FA350T (predator) and Algoriphagus marinus am2 (prey). We selected a prey Algoriphagus marinus am2, which was smaller than predator FA350T. a, The free-living prey Algoriphagus   Table S2.

Figure 4
Gene abundance in facultative and obligate bacterial predators. the heatmap is based on two-way cluster analysis of genomic abundance of genes encoding for KEGG protein groups which were specific to either facultative predators or obligate predators. Groups in blue background indicate the so-called facultative predators, groups in yellow background indicate the so-called obligate predators, and groups in red background indicate Bradymonabacteria. Two-way cluster analysis was clustered using ward.D2 method based on euclidean distances. Gene abundance matrix is available in Table S3.