Novel Plant-Associated Brevibacillus and Lysinibacillus Genomospecies Harbor a Rich Biosynthetic Potential of Antimicrobial Compounds

We have previously reported the draft genome sequences of 59 endospore-forming Gram-positive bacterial strains isolated from Vietnamese crop plants due to their ability to suppress plant pathogens. Based on their draft genome sequence, eleven of them were assigned to the Brevibacillus and one to the Lysinibacillus genus. Further analysis including full genome sequencing revealed that several of these strains represent novel genomospecies. In vitro and in vivo assays demonstrated their ability to promote plant growth, as well as the strong biocontrol potential of Brevibacilli directed against phytopathogenic bacteria, fungi, and nematodes. Genome mining identified 157 natural product biosynthesis gene clusters (BGCs), including 36 novel BGCs not present in the MIBiG data bank. Our findings indicate that plant-associated Brevibacilli are a rich source of putative antimicrobial compounds and might serve as a valuable starting point for the development of novel biocontrol agents.


Introduction
Coffee and pepper are valuable products of Vietnamese agriculture. Their coffee production worldwide ranks second, after Brazil. Black pepper is of similar importance in Vietnam, and has been exported to more than 110 countries. However, during the last years, there has been a tendency of harvest yield reduction in the case of coffee of about 20% compared to the yield of 2014. Harvest losses are caused by plant pathogens such as fungi (e.g., Fusarium oxysporum), oomycetes (Phytophthora palmivora) and root-knot nematodes (e.g., Meloidogyne incognita). In the past, chemical pesticides were used to control the phytopathogens, but their use is no longer permitted due to their toxic remnants. The epidemic occurrence of fast death disease, which is damaging black pepper plantations, has led to a rural exodus of farmers. For this reason, the sustainable development of agriculture, which includes the applying of highly efficient and reliable biocontrol agents, useful for preventing and suppressing pests on black pepper and coffee, is an urgent need.
in a shaker at 28 • C and 180 rpm for 24 h. Due to their slow growth, Brevibacillus spp. were found to be enriched when soil samples adherent to plant roots were incubated under shaking for two weeks, diluted to 10 −9 , and then plated either onto R2A agar (Oxoid Lim., Basingstoke, UK) or onto 1/5 diluted NPT agar consisting of 0.4 g/L nutrient broth, 1 g/L potato dextrose, 1.2 g/L tryptic soy broth, 2 g/L MES hydrate (2-morpholino-ethan sulfonic acid), and 15 g/L agar, pH 5.75. Colonies appeared after three to seven days incubation at 28 • C. The purified strains were maintained as glycerol stocks (20%, w/v) at −80 • C. The cultivating of the bacterial strains and DNA isolation have been described previously [1,2].
For assessment of genome similarity and phylogeny, the genome sequence data were uploaded to the Type (Strain) Genome Server (TYGS), available at https://tygs.dsmz.de, accessed on 3 December 2022 [19]. Information on nomenclature was provided by the List of Prokaryotic names with Standing in Nomenclature (LPSN, available at https://lpsn. dsmz.de, accessed on 3 December 2022) [20]. All user genomes were compared with all type strain genomes available in the TYGS database via the MASH algorithm [21], and the ten strains with the smallest MASH distances were chosen per user genome. Using the Genome BLAST Distance Phylogeny approach (GBDP), the ten closest type strain genomes for each of the user genomes were calculated. GTDB sp. were calculated with GTDB-Tk and the FASTANI calculator (https://gtdb.ecogenomic.org/, accessed on 3 December 2022). GTDB-Tk is a toolkit for assigning objective classifications to bacterial and archaeal genomes based on the Genome Database Taxonomy (GTDB). In silico DNA-DNA hybridization (dDDH) values were calculated in the TYGS platform using formula d 4 , which is the sum of all identities found in the high score segment pairs (HSPs) divided by the total length of all HSPs. A pan-genome analysis was performed using the Bacteria Pan Genome Analysis pipeline (BPGA) [22] with the amino acid sequences and default parameter settings (50% identity). The tree files provided by the BPGA pipeline and TYGS were visualized with iTOL (https://itol.embl.de/#, accessed on 3 December 2022).
In addition, the EDGAR3.0 pipeline [16] was used for elucidating taxonomic relationships based on genome sequences. To construct a phylogenetic tree for a project, the core genes of these genomes were computed. In a following step, the alignments of each core gene set are generated using MUSCLE, and the alignments are concatenated to one huge alignment. This alignment is the input for the FastTree software (http://www.microbesonline.org/fasttree/, accessed on 3 December 2022) to generate approximately-maximum-likelihood phylogenetic trees. The values at the branches of FastTree trees are _NOT_ bootstrapping values, but local support values computed by FastTree using the Shimodaira-Hasegawa test. FastANI [23] was used to calculate ANI heatmaps within the EDGAR pipeline for a selected set of genomes.

Biocontrol Activity against Plant Pathogens and Plant Growth Promotion
Antibacterial activity was examined as follows: 0.5 mL of a stationary culture (around 10 9 cells) of the phytopathogenic indicator bacteria (Clavibacter michiganensis, Dickeya solani, Erwinia amylovora) were mixed with 2 mL liquid soft agar (0.7%) at 40 • C, and then added to Petri dishes with 1.5% LB-Agar. The test bacteria (Brevibacillus and Lysinibacillus strains) were grown in liquid culture for 48 h. under continuous shaking. 10 µL of the culture was allowed to soak into filter paper discs (2 mm in diameter), and were then placed onto the surface of the soft agar containing the indicator bacteria. The cultures were examined every day and allowed to grow for six days at 27 • C.
The antifungal activity of the isolates was determined as follows: Plugs (5 mm in diameter) with the pathogenic fungi were placed onto potato dextrose agar (PDA). Paper discs with the test bacteria were then added 20 mm away from the fungi. The cultures were incubated for six days at 27 • C and examined daily.
A bioassay of nematicidal activity was performed with Caenorhabditis elegans N2 (Carolina, U.S.A., https://www.carolina.com, accessed on 3 December 2022) fed with Escherichia coli OP50 cells. The culture and synchronization of the worms was performed as previously described [24]. The L4 stage was used for two different bioassays performed as described previously [25]. In the slow killing assay, around 40 L4 C. elegans worms were added to the nematode growth medium (NGM) agar plate containing the test bacteria. The mixture was incubated for 3-5 days at 25 • C and inspected daily. In the liquid fast killing assay, the test bacteria were grown overnight under shaking (200 rpm) at 37 • C in 3 mL liquid assay medium. 100 µL of the bacterial culture was diluted with 500 µL M9 medium, and transferred into 12 well plates. Each well was seeded with 40-60 L4 stage N2 nematodes, and the assay was performed at 25 • C for 24 h. Mortalities of nematodes were defined as the ratio of dead nematodes over the tested nematodes.
Root-knot nematode Meloidogyne sp. was isolated from roots of infested pepper plants according to [26]. Tomato plantlets were grown in pots with natural soil under controlled conditions in a greenhouse. Test bacteria and second stage juvenile (J2) nematodes were added to the pots two weeks after transplanting. The number of knots in tomato plants was estimated ten weeks after infestation with the nematodes [27].
A plant growth promotion assay was performed with wild type Arabidopsis thaliana (EDVOTEK, USA https://www.edvotek.com/, accessed on 3 December 2022) according to [28]. The surface sterilized seeds were pre-germinated on Petri dishes containing halfstrength Murashige-Skoog medium semi-solidified with 0.6% agar and incubated at 22 • C under long daylight conditions (16 h light/8 h dark) for seven days. The roots of Arabidopsis seedlings were then dipped into a diluted spore suspension of the test bacteria (10 5 CFU/mL) for five min., and five seedlings were transferred into a square Petri dish containing half-strength MS-medium solidified with 1% agar. The square Petri dishes were incubated in a growth chamber at 22 • C at a daily photoperiod of 14 h. The fresh weight of the plants was measured 21 days after transplanting for estimation of the ability of bacterial strains for growth promotion. All experiments were performed in triplicate and the standard deviation SD was indicated as bars of the column diagrams.

Identification of Gene Clusters Involved in the Synthesis of Secondary Metabolites
Gene clusters for secondary metabolite synthesis were mined using antiSMASH pipeline version 6 [29] under settings of all features and BAGEL4 [30].

Data Analysis
The data obtained from biocontrol and plant growth promotion experiments were analyzed by Statistic Analysis Systems (SAS) software. The results of three replicates (n = 3) were expressed as SD (standard deviations). Significance was calculated with t-tests using the ANOVA procedure according to Duncan at LSD = 0.05. Every experiment was conducted using a completely random design. Excel software was used to create the graphical representations.

Resequencing of Selected Lysinibacillus and Brevibacillus Strains Revealed the Presence of Genomic Islands and Extrachromosomal Elements
Lysinibacillus sp. CD3-6, Brevibacillus parabrevis HD3.3A, Brevibacillus sp. DP1.3A, and Brevibacillus sp. M2.1A were sequenced using nanopore sequencing technology. The complete genome of Lysinibacillus sp. CD3-6 consisted of two DNA elements: a single circular chromosome with 4,810,966 bps (CP085880.1), and a small DNA with 2485 bps (CP085881.1). The total size of both DNA elements was 4,813,451 bps harboring 4833 coding sequences. No genes with similarity to plasmid replication proteins were detected in the small DNA element, excluding its definition as a plasmid. A circular plot of the CD3-6 chromosome (4,810,966 bps) computed against the most related genome (Lysinibacillus sp. JNUCC-52) is shown in Supplementary Figure S1. CD3-6 was used as reference for computing the core genome against a set of 15 Lysinibacillus genomes representing the most related taxonomic groups (clusters A1-4, see Section 3.2). The core genome consisted of 2733 coding sequences (CDS). The pan genome consisted of 9100 CDS. A total of 187 singletons were detected in CD3-6 when compared with the other 15 genomes (Table 1). Genomic island (GI) prediction using the IslandViewer 4 webserver [17] revealed that the CD3-6 genome was rich in putative genomic islands, mainly characterized by the presence of site-specific integrases, HNH endonucleases, and phage proteins ( Figure 1A). GI 1 (618,752-660,952) contained 31 genes including tnpA (IS200/IS605 family transposase) and a tyrosine-type recombinase/integrase encoding gene. GI 4 (1,776,323-1,805,601) contained 39 genes including several phage proteins and a gene encoding FAD dependent thymidylate synthase. A predicted class-ii lassopeptide was detected using the antiSMASH pipeline version 6 [29] in GI 5. The largest GI predicted in CD3-6 was GI 8 (2,649,327-2,738,124) harboring 119 genes including the type II toxin-antitoxin system (UED78437.1, UED78438.1) encoded by hicA and hicB. HicAB modules appear to be highly prone to horizontal gene transfer genes [31]. A complete list of the genes predicted in the CD3-6 GIs is given in Supplementary Table S1. (2,649,327-2,738,124) harboring 119 genes including the type II toxin-antitoxin system (UED78437.1, UED78438.1) encoded by hicA and hicB. HicAB modules appear to be highly prone to horizontal gene transfer genes [31]. A complete list of the genes predicted in the CD3-6 GIs is given in Supplementary Table S1.  The chromosome of the isolate B. parabrevis HD3.3A (6,154,192 bp) is presented in Supplementary Figure S2. HD3.3A was used as reference for computing the core genome against representatives of the Brevibacillus-A5 branch (clusters 25-27, see Section 3.2). The core genome consisted of 3044 CDS. The pan-genome consisted of 9438 CDS. A total of 166 singletons were detected in HD3.3A (Table 1). Some 16 GIs were found to be distributed within the HD3.3A chromosome ( Figure 1B). They were enriched with transposases of different families, such as IS3, IS21, IS110, IS256, Mu, TnsA, TnsD, Tns7, and site-specific integrases probably involved in horizontal gene transfer. An agrB gene probably involved in processing of cyclic lactone autoinducer (AIP) was detected in GI 5. Phage genes were detected in several GIs of this strain. GI 15 (5,552,579,031) harbored a number of flagellar proteins, possibly affecting the motility behavior of HD3.3A. (Supplementary Table S2).
The Brevibacillus sp. M2.1A chromosome (6300 kb) was found to be closely related to the Brevibacillus brevis strain 12B3 (Supplementary Figure S4). Computing with the other members of the Brevibacillus-A6 branch yielded 3295 core genes. The pan genome was formed by 13,428 genes. The M2.1A genome contained 113 singletons (Table 1). In contrast with DP1.3A, the GIs detected in the M2.1A chromosome ( Figure 1D) did not harbor genes involved in the synthesis of secondary metabolites. A complete thioredoxin system with thioredoxin, thioredoxin-disulfide reductase, and thiol peroxidase was detected in GI 13 (Supplementary Table S4). This system might enable M2.1A to respond efficiently against oxidative stress [35].
M2.1A harbored low-copy plasmid DNA (size: 19,434 bps, GC content: 42.0%), whose partition might be governed by the plasmid partition protein A [36] ( Table 1). The ParA family protein (MCC8438707.1) shared the highest similarity with the AAA family ATPase from Brevibacillus borstelensis (WP_251238174.1, 93.49% identity). In addition, the plasmid also shared partial sequence similarity (16% of their total length) with the low-copy plasmid from Brevibacillus sp. DP1.3A (CP085878).  [37]. Calculation of core and pan genomes was performed as described in the text (Section 3.1). The protein features estimated with RAST are presented in Supplementary Table S5.

Lysinibacillus CD3-6 Forms a Distinct Genomospecies Together with
Lysinibacillus JNUCC-52 The 16S rRNA gene sequence was extracted using the TYGS server from the whole genome sequence of CD3-6 and used for phylogenetic analysis. We have also directly sequenced the 16S rRNA of CD3-6, which was deposited in the NCBI data base as MW820197.1. Both sequences differ by only one nucleotide. The resulting tree indicated a single species cluster formed by CD3-6 with Lysinibacillus sphaericus as the closest related species (Supplementary Figure S5). In order to exclude the possibility that the type of strains without known genome sequences, but more related to CD3-6, escaped our analysis, we performed a BLASTN-supported search (https://blast.ncbi.nlm.nih.gov, accessed on 3 December 2022) for related 16S rRNA sequences. However, no 16S rDNA sequences with more similarity than L. sphaericus (99.61%) were detected.
Since16S rRNA sequences are often not sufficient for species discrimination, we used the genome sequences for taxonomical strain identification. We started our analysis with the CD3-6 genome (CP085880) and 112 Lysinibacillus genomes obtained from the NCBI data bank. The phylogenetic tree ( Figure 3) suggested that the members of the Lysinibacillus genus can be divided into two major groups, A and B. Group B members were not always representatives of the Lysinibacillus genus, but were often classified as representatives of different Ureibacillus species. Thus, group B seemed to be heterogenous, and to contain different genera. Group A contained Lysinibacillus genomes related to the type strain Lysinibacillus sphaericus DSM 28 [10], formerly Bacillus sphaericus. Based on the cut-off values for species delineation using ANI (96%) and dDDH (70%) [38], 24 clusters (A1-A24) were distinguished (Supplementary Table S6). Ten of the clusters contained type strains of recognized Lysinibacillus species, but most of the clusters (14) did not contain type strains, and represented according to the definition given by EZBioCloud (https://help.ezbiocloud.net/genomospecies/, accessed on 3 December 2022) for unnamed genomospecies consisting of so far unclassified or wrongly labelled Lysinibacillus genomes.
The type strain Lysinibacillus sphaericus DSM 28 was classified as being a member of cluster A4. Due to its genome sequence, Lysinibacillus sp. CD3-6 was assigned to belong to cluster A2 when using a 70% dDDH radius around each of the 13 Lysinibacillus type strains ( Figure 3B).
Microorganisms 2023, 11, x FOR PEER REVIEW 10 of 32 contain different genera. Group A contained Lysinibacillus genomes related to the type strain Lysinibacillus sphaericus DSM 28 [10], formerly Bacillus sphaericus. Based on the cutoff values for species delineation using ANI (96%) and dDDH (70%) [38], 24 clusters (A1-A24) were distinguished (Supplementary Table S6). Ten of the clusters contained type strains of recognized Lysinibacillus species, but most of the clusters (14) did not contain type strains, and represented according to the definition given by EZBioCloud (https://help.ezbiocloud.net/genomospecies/, accessed on 3 December 2022) for unnamed genomospecies consisting of so far unclassified or wrongly labelled Lysinibacillus genomes.
The type strain Lysinibacillus sphaericus DSM 28 was classified as being a member of cluster A4. Due to its genome sequence, Lysinibacillus sp. CD3-6 was assigned to belong to cluster A2 when using a 70% dDDH radius around each of the 13 Lysinibacillus type strains ( Figure 3B). . Approximately-maximum-likelihood phylogenetic tree for 113 Lysinibacillus genomes, calculated by EDGAR using the Fast Tree software (http://www.microbesonline.org/fasttree/, accessed on 3 December 2022) and drawn by iTOL. Type strains are labelled in blue letters. Strains, previously misidentified as Lysinibacillus and now reclassified as representatives of other species, are labelled in brown. The tree harbored two main branches. Branch A contains the species related to Lysinibacillus sphaericus, whilst branch B contains the species which are more remote from L. sphaericus. The tree was built out of a core of 156 genes per genome, for a total of 17628. The core has 58525 AA-residues/bp per genome, for a total of 6613325. (B). Lysinibacillus tree inferred with FastMe 2.1.6.1 from GBDP distances calculated from whole genome sequences using the Type (Strain) Genome Server TYGS (https://tygs.dsmz.de, accessed on 3 December 2022). The tree consisted of 26 species and 28 subspecies clusters. Analysis was performed using both Maximum Likelihood and Maximum Parsimony. The numbers above the branches are GBDP pseudo-bootstrap support values > 60% from replications, with an average branch support of 66.0 %. Genomospecies according to Figure 3A are indicated. The first two colored columns to the right of each name refer to the genomebased species and subspecies clusters, respectively, as determined by dDDH cut-off of 70 and 79%, respectively. The GTDB species are indicated at the right. The clustering yielded 16 species clusters and strain CD3-6 (labelled in red) was assigned together with JNUCC-52 as novel genomospecies (A2, Lysinibacillus sp002340205). Type strains are labelled in blue letters. The tree was rooted at the midpoint.
The type strains most related to CD3-6, Lysinibacillus sphaericus KCTCC 3346, Lysinibacillus tabacifolii K3514, and Lysinibacillus mangiferihumi, possessed dDDH values (d4) far below of the species cut off (dDDH < 70, Supplementary Table S6). We conclude that CD3-6 represents, together with Lysinibacillus JNUCC-52, a novel genomospecies distinguished from the L. sphaericus species cluster. It should be noted that L. mangiferihumi, L. tabacifolii, and L. varians were recently characterized as later heterotrophic synonyms of L. sphaericus [39], and do not represent valid species. . Approximately-maximum-likelihood phylogenetic tree for 113 Lysinibacillus genomes, calculated by EDGAR using the Fast Tree software (http://www.microbesonline.org/fasttree/, accessed on 3 December 2022) and drawn by iTOL. Type strains are labelled in blue letters. Strains, previously misidentified as Lysinibacillus and now reclassified as representatives of other species, are labelled in brown. The tree harbored two main branches. Branch A contains the species related to Lysinibacillus sphaericus, whilst branch B contains the species which are more remote from L. sphaericus. The tree was built out of a core of 156 genes per genome, for a total of 17,628. The core has 58,525 AAresidues/bp per genome, for a total of 6,613,325. (B). Lysinibacillus tree inferred with FastMe 2.1.6.1 from GBDP distances calculated from whole genome sequences using the Type (Strain) Genome Server TYGS (https://tygs.dsmz.de, accessed on 3 December 2022). The tree consisted of 26 species and 28 subspecies clusters. Analysis was performed using both Maximum Likelihood and Maximum Parsimony. The numbers above the branches are GBDP pseudo-bootstrap support values >60% from replications, with an average branch support of 66.0%. Genomospecies according to (A) are indicated. The first two colored columns to the right of each name refer to the genome-based species and subspecies clusters, respectively, as determined by dDDH cut-off of 70 and 79%, respectively. The GTDB species are indicated at the right. The clustering yielded 16 species clusters and strain CD3-6 (labelled in red) was assigned together with JNUCC-52 as novel genomospecies (A2, Lysinibacillus sp002340205). Type strains are labelled in blue letters. The tree was rooted at the midpoint.

Novel GS Were Assigned for Plant-Associated Brevibacillus Isolates
The phylogenomic tree constructed with 134 Brevibacillus genomes, including the eleven Brevibacillus isolates obtained from Vietnamese crop plants, demonstrated that three different major clusters (A, B, C) can be distinguished (Figure 4). Group B consisted of Brevibacillus laterosporus strains, whilst group A was formed by representatives of Brevibacillus brevis and their closest relatives. Within group A, six major branches with a total of 23 clusters were distinguished. Only 14 of those clusters were covered by type strains, whilst nine clusters represent novel genomospecies without validly recognized type strains. The majority of strains obtained from Vietnamese crop plants clustered within the Brevibacillus A group at branch 6. Only the putative B. parabrevis strains HD3.3A and HD1.4A clustered within branch 5 (Supplementary Table S7). A third cluster, group C, appeared as a heterogenous group of Brevibacilli, displaying a distant relationship to each other and to the other representatives of the Brevibacillus taxon ( Figure 4A).
According to the analysis routine performed by the Type (Strain) Genome Server (TYGS), five of the Vietnamese Brevibacillus isolates clustered together with valid recognized type strains. The Brevibacillus strains HD1.4A, and HD3.3A were assigned as B. parabrevis, and HB1.1, HB1.2, HB1.4B formed a cluster with the B. porteri type strain. The remaining six isolates did not cluster together with validly published type strains and most likely represent four novel GS. Brevibacillus sp. RS1.1, HB2.2, and HB1.3 clustered together within one putative novel species cluster, but split into two different subspecies. Two distinguished novel species cluster were formed by Brevibacillus sp.: DP1.3A and M2.1A. Another novel GS was formed by Brevibacillus MS2.2 together with Brevibacillus sp. Leaf182, isolated from Arabidopsis phyllosphere [9] ( Figure 4A).
The ANI was proposed to replace classical DNA-DNA hybridization (DDH) as the method for prokaryotic species circumscription in 2009 [45]. The FastANI heatmap constructed with the Brevibacillus group A genomes (Supplementary Figure S8) corroborated that the clusters presented in the phylogenomic tree (Supplementary Figure S9) correctly reflected their taxonomic relationship down to species level when using the cut-off level (95-96%) recommended for interspecies identity [43]. In addition, we have aligned our data with the most recent release of the Genome Taxonomy Database (GTDB) [40] (Release 07-RS207 (8 April 2022), and also found appropriate designations used in Supplementary Figure S8 in case of clusters not covered by recognized type strains. According to the classification given by ANI, dDDH, and GTDB, the query Brevibacillus isolate genomes were clustered as follows: Brevibacillus sp. M2.1A formed the unique GS A6-34 as a single strain, which was found to be most related to the Brevibacillus formosus cluster (GS A6-35). However, ANIb values estimated when compared with this cluster were found to be below the species cut-off (<96%). The GTDB classification of this cluster was s_Brevibacillus sp. 013284355.   Approximately-maximum-likelihood phylogenetic tree for 134 Brevibacillus genomes using the FastTree software accessible within the EDGAR package 3.0 [16]). Brevibacilli investigated in this study are labelled in red letters. The tree was built out of a core of 495 genes per genome, for 66,330 in total. The core has 167683 AA-residues/ bp per genome, 15461980 in total. To construct a phylogenetic tree for a project, the core genes of these genomes are computed. In a following step, alignments of each core gene set are generated using MUSCLE, and the alignments are concatenated to one huge alignment. The ANI was proposed to replace classical DNA-DNA hybridization (DDH) as the method for prokaryotic species circumscription in 2009 [45]. The FastANI heatmap constructed with the Brevibacillus group A genomes (Supplementary Figure S8) corroborated that the clusters presented in the phylogenomic tree (Supplementary Figure S9) correctly reflected their taxonomic relationship down to species level when using the cut-off level (95-96%) recommended for interspecies identity [43]. In addition, we have aligned our . Approximately-maximum-likelihood phylogenetic tree for 134 Brevibacillus genomes using the FastTree software accessible within the EDGAR package 3.0 [16]). Brevibacilli investigated in this study are labelled in red letters. The tree was built out of a core of 495 genes per genome, for 66,330 in total. The core has 167,683 AA-residues/ bp per genome, 15,461,980 in total. To construct a phylogenetic tree for a project, the core genes of these genomes are computed. In a following step, alignments of each core gene set are generated using MUSCLE, and the alignments are concatenated to one huge alignment. (B). Brevibacillus tree inferred with FastMe 2.1.6.1 [44] from GBDP distances calculated from whole genome sequences using the Type (Strain) Genome Server TYGS (https://tygs.dsmz.de, accessed on 3 December 2022). Analysis was performed using both Maximum Likelihood and Maximum Parsimony, with 14 type strains (labelled in blue letters) and 48 additional genome sequences including the Brevibacillus strains isolated from Vietnamese crop plants (labelled by red letters). The numbers above branches are GBDP pseudo-bootstrap support values >60% from replications, with an average branch support of 85.1%. The first column to the right of each name refers to the genome-based species according to the nomenclature given in (A). Species and Subspecies cluster (columns 2 and 3) are characterized by dDDH cut-off values of 70 and 79%, and ANI values of 96 and 98%, respectively. A total of 15 GTDB species clusters could be distinguished. The tree was rooted at the midpoint.
The strains including their corresponding GS are summarized in Supplementary Table S7 and Supplementary Figure S9, and were used for bioassays of their ability to control plant pathogens and to promote plant growth, as described in the next section.

Plant-Associated Brevibacilli Promote Plant Growth and Suppress Plant Pathogens
Our in vitro bioassays demonstrated that the Brevibacillus strains did efficiently inhibit the growth of phytopathogenic bacteria, fungi, and nematodes (Supplementary Table S8).

Antagonistic Activity against Gram-Positive and Gram-Negative Bacteria
Antagonistic activity against bacteria was indicated by inhibition zones around filter discs containing the test bacteria. The filter discs were placed onto soft agar mixed with the Gram-positive bacterium Clavibacter michiganensis, the causative agent of ring-rot in potato tubers. Antibiosis was demonstrated with all tested representatives of the Brevibacillus taxon independent of whether cells or supernatants were used. Inhibition zones appeared after one day and increased steadily during the whole period of observation. Brevibacillus sp. HB2.2, Brevibacillus sp. RS1.1, B. porteri HB1.4B, and Brevibacillus sp. DP1.3A were among the strains with the highest antagonistic activity against Clavibacter michiganensis. However, the antagonistic activity of Lysinibacillus CD3-6 was hard to detect ( Figure 5).  Table S8).
The Gram-negative plant pathogens Dickeya solani and Erwinia amylovora (causative agent of fire blight disease at orchard trees) were also used as indicators for antagonistic activity. This ruled out the possibility that the Gram-negative bacteria were less inhibited by the Brevibacillus strains as the Gram-positive bacterium C. michiganensis. Brevibacillus sp. RS1.1, Brevibacillus sp. MS2.2, B. porteri HB1.4B, and Brevibacillus DP1.3A inhibited the  Table S8).
The Gram-negative plant pathogens Dickeya solani and Erwinia amylovora (causative agent of fire blight disease at orchard trees) were also used as indicators for antagonistic activity. This ruled out the possibility that the Gram-negative bacteria were less inhibited by the Brevibacillus strains as the Gram-positive bacterium C. michiganensis. Brevibacillus sp. RS1.1, Brevibacillus sp. MS2.2, B. porteri HB1.4B, and Brevibacillus DP1.3A inhibited the growth of Dickeya solani and Erwinia amylovora after an incubation period of two or three days, but their inhibition zones were relatively small. Brevibacillus sp. HB2.2, B. parabrevis HB2.2, and HD1.4A did not apparently inhibit the growth of both Gramnegative pathogens, suggesting that Brevibacilli were more efficient against Gram-positive bacteria (Supplementary Figure S10).

Antifungal Activity
Antifungal activity was examined in vitro using four different Fusarium species (F. oxysporum, F. culmorum, F. poae, F.graminearum) known for causing fusarium wilt disease [46], and Aspergillus niger ('black mold'). Lysinibacillus sp. CD3-6 was found to be inefficient against most of the Fusarium strains, but the Brevibacillus strains did suppress the growth of all phytopathogenic fungi and the oomycete P. palmivora, one of the most detrimental plant pathogens in Vietnam [47] Figure S11).

Nematicidal Activity
Root-knot nematodes, such as Meloidogyne spp., are one of the most important plant pathogens in tropical and temperate agriculture, and are responsible for significant harvest losses of main Vietnamese crops, such as coffee and black pepper [48]. In order to analyze the antagonistic activity of the Brevibacillus strains and Lysinibacillus CD3-6, we first tested their suppressing effect against the model nematode Caenorhabditis elegans. Fast and slow death rates were estimated in a bioassay under laboratory conditions. This ruled out the possibility that the Brevibacillus strains were much more efficient than Lysinibacillus sp. CD3-6. Brevibacillus sp. M2.1A displayed the highest killing effect against C. elegans ( Figure 6A). In order to examine the suppressing effects against phytopathogenic nematodes more directly, we isolated a representative strain of Meloidogyne sp. directly from the galls of infested black pepper plant roots according to the hypochlorite procedure [49]. The suppressing effect exerted by the test bacteria on disease development was examined in a greenhouse experiment. Ten weeks after the transplanting of the tomato plantlets in soil, the formation of root knots was visually registered and used as a measure for calculating the disease index according to [27]. The Brevibacillus strains were found to be efficient in reducing the disease severity to around 50% and less compared to the untreated control. Again, Brevibacillus M2.1A performed the best, whilst Lysinibacilllus sp. CD3-6 was found to be less efficient than all tested Brevibacillus strains ( Figure 6B).
Whilst reports about the biocontrol action of Bacillus ssp., such as B. firmus [50], B. velezensis former B. amyloliquefaciens [25], B. cereus, B. thuringiensis, and B. subtilis [51] are increasing, to the best of our knowledge we present here the first report on the nematicidal activity of Brevibacilli against root-knot nematodes. Whilst reports about the biocontrol action of Bacillus ssp., such as B. firm velezensis former B. amyloliquefaciens [25], B. cereus, B. thuringiensis, and B. subt increasing, to the best of our knowledge we present here the first report on the n activity of Brevibacilli against root-knot nematodes.

Plant Growth Promotion
We examined the effect of the Brevibacillus strains and Lysinibacillus CD3-6 abidopsis thaliana biotest system [28]. Several Brevibacillus strains enhanced the the Arabidopsis seedlings in a considerable manner (Figure 7, Supplementary T

Plant Growth Promotion
We examined the effect of the Brevibacillus strains and Lysinibacillus CD3-6 in the Arabidopsis thaliana biotest system [28]. Several Brevibacillus strains enhanced the growth of the Arabidopsis seedlings in a considerable manner (Figure 7, Supplementary Table S9).
The above phenotypic experiment proved that plant-associated Brevibacillus strains isolated from Vietnamese crop plants are able to positively interact with Arabidopsis plants, and to stimulate their growth in the range previously reported for the prototype of the Gram-positive plant-growth-promoting rhizobacteria, Bacillus velezensis FZB42 [52]. The above phenotypic experiment proved that plant-associated Brevibacillus strains isolated from Vietnamese crop plants are able to positively interact with Arabidopsis plants, and to stimulate their growth in the range previously reported for the prototype of the Gram-positive plant-growth-promoting rhizobacteria, Bacillus velezensis FZB42 [52].

Genome Mining for Putative Natural Product Biosynthesis Gene Clusters
The identification of biocontrol actions directed against phytopathogenic bacteria, fungi and nematodes prompted us to investigate the biosynthetic potential of the isolates using either their draft or their full genome sequences. Antimicrobial compounds belong to structurally diverse groups of molecules, such as nonribosomal peptides (NRP) and polyketides (PK), and ribosomally synthesized and posttranslationally modified peptides [53]. The antiSMASH 6.0 version [29] was used for the prediction and annotation of the biosynthetic gene clusters (BGCs). The antiSMASH results were subsequently compared against the MIBiG database [54] in order to identify characterized and uncharacterized BGCs. In total, 151 BGCs in the 11 Brevibacillus isolates, and 6 BGCs in Lysinibacillus sp. CD3-6 were identified and separated into nine distinct classes (Supplementary Table S10). A total of 36 of them were not detected in the MIBiG database and might encode the biosynthesis of uncharacterized secondary metabolites. The edeine gene cluster, previously detected in Brevibacillus brevis Vm4 and X23 [55], but not listed in the MIBiG databank, was found to be widely distributed in the Brevibacillus spp. strains. A variant of another modular PK-NRP hybrid, Paenilipoheptin, recently described in Paenibacillus polymyxa E681 [56], was detected for the first time in the genus Brevibacillus.

Gene Clusters Encoding Modular and Nonmodular Polyketides
Genes with more than 90% similarity to the gene cluster encoding the non-ribosomal synthesized polyketide macrobrevin [9] were detected in five of the Vietnam Brevibacillus strains (Supplementary Table S11). Generally, polyketides are synthesized by giant polyketide synthases in a non-ribosomal manner [57]. Similar to the gene cluster in Brevibacillus Leaf182 (BGC0001470), Brevibacillus sp. strains DP1.3A and HB1.3 contained the

Genome Mining for Putative Natural Product Biosynthesis Gene Clusters
The identification of biocontrol actions directed against phytopathogenic bacteria, fungi and nematodes prompted us to investigate the biosynthetic potential of the isolates using either their draft or their full genome sequences. Antimicrobial compounds belong to structurally diverse groups of molecules, such as nonribosomal peptides (NRP) and polyketides (PK), and ribosomally synthesized and posttranslationally modified peptides [53]. The antiSMASH 6.0 version [29] was used for the prediction and annotation of the biosynthetic gene clusters (BGCs). The antiSMASH results were subsequently compared against the MIBiG database [54] in order to identify characterized and uncharacterized BGCs. In total, 151 BGCs in the 11 Brevibacillus isolates, and 6 BGCs in Lysinibacillus sp. CD3-6 were identified and separated into nine distinct classes (Supplementary Table S10). A total of 36 of them were not detected in the MIBiG database and might encode the biosynthesis of uncharacterized secondary metabolites. The edeine gene cluster, previously detected in Brevibacillus brevis Vm4 and X23 [55], but not listed in the MIBiG databank, was found to be widely distributed in the Brevibacillus spp. strains. A variant of another modular PK-NRP hybrid, Paenilipoheptin, recently described in Paenibacillus polymyxa E681 [56], was detected for the first time in the genus Brevibacillus.

Gene Clusters Encoding Modular and Nonmodular Polyketides
Genes with more than 90% similarity to the gene cluster encoding the non-ribosomal synthesized polyketide macrobrevin [9] were detected in five of the Vietnam Brevibacillus strains (Supplementary Table S11). Generally, polyketides are synthesized by giant polyketide synthases in a non-ribosomal manner [57]. Similar to the gene cluster in Brevibacillus Leaf182 (BGC0001470), Brevibacillus sp. strains DP1.3A and HB1.3 contained the complete set of 15 modules, whilst strains MS2.2, HB2.2, and RS1.1, harboring only 12 modules, did probably encode truncated versions of the polyketide (Supplementary Figure S12). Macrobrevin displays a unique polyketide structure, and was shown to be active against leaf colonizing Bacillus strains [9].
Gene clusters with weak similarity to the nonmodular Type III PKS gene cluster (BGC0001964) were detected in most Brevibacilllus strains, and also in Lysinibacillus sp. CD3-6 (Supplementary Table S11). Type III PKSs produce a large number of aromatic natural compounds in plants, fungi and bacteria. They catalyze the condensation of coenzyme activated starter units with (methyl)malonyl-CoA extender units by decarboxylative Claisen condensations [58]. A type III PKS chalcone synthase has been described in Streptomyces griseus [59] and other Gram-positive bacteria such as B. subtilis [60].
The three genes tycA, tycB, and tycC encoding the cyclic ß-sheet decapeptide tyrocidine (C 66 H 87 N 13 O 13 ) were detected in the 11 Brevibacillus strains representing the species B. parabrevis, B.porteri and four novel genomospecies (Supplementary Table S11). The gene clusters widely resembled the tyrocidine gene cluster from Brevibacillus brevis ATCC8185 [62]. Except for the adenylation domains present in modules 3 and 4, the other eight modules were found highly conserved in all Brevibacillus strains (Supplementary Figure S13). This corresponds to the structures reported for tyrocidine A, B and C bearing residues Phe 3/4 , Phe 3 /Trp 4 , and Trp 3/4 , respectively. The membrane effective antibiotics kill Gram-positive bacteria such as Bacillus subtilis and Staphylococcus aureus [63].
Surprisingly, we detected a novel variant of the gramicidin gene cluster in Brevibacillus sp. DP1.3A (GS A6-30) containing two additional modules in the 3 -region of the lgrB gene ( Figure 8). Their adenylation domains were coding for Ala and D-Val, and the modules seem to be caused by a partial duplication in the 3 -region. We hypothesize that the 18module gene cluster is coding for a linear heptadecapeptide (N-formyl-heptadecapeptideethanolamine) containing two additional amino acid residues, Ala and D-Val, in position 7 and 8. Other Brevibacillus strains harbored truncated forms of both gramicidin gene cluster variants. A truncated pentadecapeptide variant was detected in Brevibacillus sp. RS1.1, whilst gene clusters encoding truncated heptadecapeptides were detected in Brevibacillus sp. HB2.2, and MS2.2, belonging to genomospecies A6-33, and A6-34, respectively (Supplementary Table S11). Since the truncated gramicidin gene clusters were predicted from draft genomes (WGS), we cannot exclude the possibility that the deletions are due to incomplete sequences. The gramicidin gene cluster present in the MIBiG data bank, BGC0000367 (AP008955.1), was identified as a 13-module gene cluster bearing a 3 deletion in the lgrD gene, probably encoding a truncated variant of the heptadecapeptide (Supplementary Table S11, Supplementary Figure S14). We propose to replace BGC0000367 by the gramicidin gene clusters from B. brevis ATCC8185, encoding the complete grami-cidin pentadecapeptide, and Brevibacillus sp. DP1.3A, encoding the putative gramicidin heptadecapeptide.
Brevibacillus sp. MS2.2. harbored a gene cluster with high similarity to marthiapeptide (BGC0001469), a gene cluster recently detected in Brevibacillus Leaf182 [9] (Supplementary Table S11, Supplementary Figure S15). The polythiazole cyclopeptide marthiapeptide was first described in Marinactinospora thermotolerans and shown to inhibit Gram-positive bacteria and cancer cells [65]. Surprisingly, we detected a novel variant of the gramicidin gene cluster in Brevibacillus sp. DP1.3A (GS A6-30) containing two additional modules in the 3′-region of the lgrB gene ( Figure 8). Their adenylation domains were coding for Ala and D-Val, and the modules seem to be caused by a partial duplication in the 3′-region. We hypothesize that the 18-module gene cluster is coding for a linear heptadecapeptide (N-formyl-heptadecapeptide-ethanolamine) containing two additional amino acid residues, Ala and D-Val, in position 7 and 8. Other Brevibacillus strains harbored truncated forms of both gramicidin gene cluster variants. A truncated pentadecapeptide variant was detected in Brevibacillus sp. RS1.1, whilst gene clusters encoding truncated heptadecapeptides were detected in Brevibacillus sp. HB2.2, and MS2.2, belonging to genomospecies A6-33, and A6-34, respectively (Supplementary Table S11). Since the truncated gramicidin gene clusters were predicted from draft genomes (WGS), we cannot exclude the possibility that the deletions are due to incomplete sequences. The gramicidin gene cluster present in the MIBiG data bank,  Table S11). The six genes edeP, edeN, edeL, edeK, edeJ, and edeI encode NRPS and PKS-NRPS hybrids involved in the non-ribosomal synthesis of the atypical cationic edeine peptides (Supplementary Figure S16A). Edeines contain a ß-Tyr or a ß-Phe residue at the N-terminus and a spermidine-polyamine structure at the C-terminus, flanking five non-proteinogenic amino acids in the central part (Supplementary Figure S14A). Edeine is a broad-spectrum antimicrobial agent acting against bacteria and fungi. DNA synthesis is inhibited at low concentrations <15 µg/mL [66]. Despite the fact that edeines have been discovered as early as 1959 [67], corresponding gene clusters have not been deposited in the MIBiG data bank until now.
Interestingly, we found a gene cluster in Brevibacillus sp. DP1.3A, which strongly resembled the phn gene cluster in Paenibacillus polymyxa E681 (BGC0001728). The phn gene cluster is possibly involved in the non-ribosomal synthesis of the cyclic lipoheptapeptide paenilipoheptin [56]. Both gene clusters contain eight modules (Supplementary Table S11). The first module is involved in fatty acid synthesis, while the remaining modules, containing seven adenylation domains, were predicted to be responsible for the synthesis of a seven-member peptide chain containing D-ser-(1), D-phe-(5), tyr-(6), and glu-residues (7) ( Figure 8B). The structure of paenilipoheptin synthesized by P. polymyxa was elucidated by MALDI-LIFT TOF/TOF MS as being C13-ß-NH2-FA-Ser (1)-Dab (2)-Trp (3)-Val (4)-Phe (5)-Tyr (6)-Glu (7) [56], which is compatible with the predicted structure of the gene cluster detected in Brevibacillus DP1.3A (Supplementary Figure S17). A thioesterase domain at the 3 end of the phnE gene was missing in both gene clusters, indicating that cyclization might be accomplished by a free-standing TE enzyme as proposed by Vater et al. [56].

Gene Clusters Representing RiPPs, and Bacteriocins
In contrast to polyketides and peptides, which are synthesized independently from ribosomes, numerous secondary metabolites with antimicrobial activity such as RiPPs (ribosomally synthesized and posttranslationally modified peptides) and (unmodified) bacteriocins are synthesized by a ribosome-dependent mechanism. Several groups are distinguished [68]. Some of them, such as lassopeptides, LAPs, lanthipeptides, UviB peptides, and sactipeptides, were detected applying the antiSMASH and BAGEL4 [30] toolkits in the Brevibacillus isolates and/or the Lysinibacillus sp. CD3-6 strain ( Figure 9).
Many RiPP biosynthetic proteins recognize and bind their cognate precursor peptide through a domain known as RiPP recognition element (RRE) [69]. The detection of RRE domains can be helpful in identifying gene clusters involved in the synthesis of novel classes of RiPPs [70], but does not necessarily identify a specific category of RiPPs. RRE containing domains with weak similarity to Pantocin-Microcin RRE (BGC 0000585) were found in nearly all Brevibacillus strains, but they were not associated with other genes involved in RiPP synthesis. In contrast, the gene for the lassopeptide RRE domain protein in Lysinibacillus sp. CD3-6 was associated with the genes encoding the lassopeptide class ii core (leader) peptide with a putative macrolactam sequence, and the lasso peptide biosynthesis B2 protein (Supplementary Table S12).
Gene clusters encoding linear azol(in)e-containing peptides (LAPs) were detected in all 11 Brevibacillus isolates, and Lysinibacillus CD3-6 (Supplementary Table S12). Most of them were identified as being members of the TOMM class (thiazole/oxazole-modified microcins) characterized by a gene cluster consisting of a cyclodehydratase gene and associated genes encoding dehydrogenase and a maturation protein. A gene encoding a TOMM precursor leader peptide was detected upstream of these three genes (Supplementary Figure S16). Typically, the TOMM precursor leader peptides were characterized by a homologous leader region and then a region enriched with Cys residues, a feature of the hetero-cycloanthracin/sonorensin family [68]. This type of thiopeptide encoding gene was detected in representatives of the Brevibacillus genomospecies 25, 29, 31, and 33 (Supplementary Table S12 Table S12). The LAP gene cluster in Lysinibacillus CD3-6 encoded a SagB/ThcOx family dehydrogenase, and a YcaOlike family protein, but did not possess genes encoding maturase and precursor peptide proteins. Many RiPP biosynthetic proteins recognize and bind their cognate precursor peptide through a domain known as RiPP recognition element (RRE) [69]. The detection of RRE domains can be helpful in identifying gene clusters involved in the synthesis of novel classes of RiPPs [70], but does not necessarily identify a specific category of RiPPs. RRE containing domains with weak similarity to Pantocin-Microcin RRE (BGC 0000585) were found in nearly all Brevibacillus strains, but they were not associated with other genes involved in RiPP synthesis. In contrast, the gene for the lassopeptide RRE domain protein in Lysinibacillus sp. CD3-6 was associated with the genes encoding the lassopeptide class Ripps similar to Linocin M18 were found in all Brevibacillus GS (Supplementary Table S12). The Linocin_M18 bacteriocin, first isolated from Brevibacterium linens M18 by Valdez-Stauber and Scherer [71], inhibits Gram-positive bacteria. These widely distributed proteins, referred to as encapsulins, form nano-compartments within the bacterium which contain ferritin-like proteins or peroxidaseenzymes. Lanthipeptides are defined by the presence of ß-thioether cross-links, which are generated by the posttranslational modification of Ser/Thr and Cys residues [72]. Best studied are class I and class II lanthipeptides, which are modified by different dehydratases (LanB or LanM) and cyclases. Gene clusters encoding class I lanthipeptides were detected in two representatives of Brevibacillus A representative of UV inducible peptides (UviB) was detected in Brevibacillus parabrevis using BAGEL4 supported genome mining (Supplementary Table S12, Supplementary Figure S22). The Phage_holin_BhlA family is a family of holin-like proteins from both bacteriophages and bacterial chromosomes. BhlA, a putative holin-like protein of Bacillus licheniformis AnBa9 [73], and Bacillus pumilus WAPB4 [74], showed antibacterial activity against several Gram-positive bacteria.

Gene Cluster Involved in Synthesis of Siderophores and Other BGCs
The asbABCDEF gene cluster (BGC0000942) from Bacillus anthracis, the causative agent of anthrax, is responsible for the biosynthesis of petrobactin, a catecholate siderophore that functions in both iron acquisition and virulence [75]. It has been argued that the iron-siderophore petrobactin contributes to B. anthracis pathogenesis, which requires maintaining sufficient iron concentration during growth in the host tissue [76]. However, it has been documented that petrobactin synthesis is also common in non-pathogenic members of the B. cereus senso latu group [77].  Table S12). By contrast, the gene clusters encoding non-ribosomal synthesis of the second siderophore bacillibactin, common in the B. subtilis, and B. cereus group, were not detected in the investigated Brevibacillus strains.
Other BGCs such as gene clusters involved in synthesis of terpenes, and the cyclic lactone autoinducer with a putative role in quorum sensing (AgrB precursorpeptides) were detected in most Brevibacillus isolates. Lysinibacillus sp. CD3-6 also harbored terpene encoding genes, and a gene cluster possibly involved in the synthesis of a betalactone containing protease inhibitor (Supplementary Table S12).

Uncharacterized NRP and PK-NRP Hybrid Gene Clusters in Brevibacilli
In addition to the characterized BGCs mentioned above, we detected 35 hitherto unknown NRPs and NRP-PK hybrid scaffolds in the Brevibacillus genomes, and one partial NRPS gene cluster containing a tyrosine module in Lysinibacillus CD3-6 (Supplementary Table S13).
A putative PK-NRP hybrid consisting of four modules (PK-orn-x-phe) occurred in all nine Brevibacillus strains belonging to the Brevibacillus A6-branch. In contrast, a three-module hybrid (asn-gly-pk) occurred only in Brevibacillus M2.2 (Supplementary Figure S23).
Giant PK-NRP hybrids consisting of either 16, 20, or 25 modules (Supplementary Figure S24) were detected in several Brevibacillus strains. The 16 M scaffold consisted of 14 PK modules (mal, ccmal, ohmal), and two modules involved in the non-ribosomal synthesis of amino acids (M6:X, M16: Ser). The three B. porteri strains, and Brevibacillus HB1.3, RS1.1, HB2.2, and M2.1.A harbored all 16 modules of the hybrid, whilst Brevibacillus DP1.3A harbored a deleted gene cluster, in which the modules 9-16 were missing. The largest gene cluster harboring 25 modules was detected in B. parabrevis HD3.3A. Some 14 modules were responsible for the non-ribosomal synthesis of amino acids, whilst the 11 pk-modules were involved in the synthesis of mal, ohmal, and ccmal. A deleted form of this scaffold harboring only 20 modules was detected in B. parabrevis HD1.4A.
Decapeptide scaffolds were detected adjacent to the B. parabrevis giant M25/20 PK-NRP hybrid scaffolds (Supplementary Table S13). Experimental proof is needed to corroborate their stand-alone state. Alternatively, it is also possible that they are directly connected with the giant hybrids.
Slightly modified NRP scaffolds consisting of either six or seven modules, and always starting with Glu-Ser at their N-terminus, were present in nearly all representatives of the A6 branch (Supplementary Figure S25) Preliminary experiments did not reveal hexa-or heptapeptides in the Brevibacillus strains, but pentapeptides with partial corresponding sequences were detected. At present, we cannot exclude the possibility that the scaffolds are responsible for the synthesis of the pentapeptides, but one or two of the modules might be overread or not expressed.
An interesting NRP gene cluster encoding for three cysteine residues was detected in B. parabrevis HD1.4A, HD3.3A, Brevibacillus DP1.3A, and B. porteri HB1.1, HB1.2, HB1.4B (Supplementary Table S13). Due to the presence of predicted methyltransferase domains, it can be assumed that the cysteine residues are methylated (Supplementary Figure S26). The cryptic nrs gene cluster of the model biocontrol bacterium Bacillus velezensis FZB42 also harbors three modules encoding for cysteine. Recently, trithiazole was identified as the product of the nrs gene cluster, and NrsB as the oxidizing enzyme of the thiazoline precursor [78]. However, a more careful comparison of the domain structure of both gene clusters revealed significant differences between both BGCs, excluding the possibility that the final product can be a polythiazole ( Figure 10). Although no o Ox domain was detected, two genes involved in dihydroxybenzoate (dhb) synthesis were ( Figure 10A). The 2,3-dhb-AMP ligase and an oxido-reductase, predicted to be involved in dhb synthesis, possess counterparts in the dhb gene cluster of FZB42 known to be responsible for the non-ribosomal synthesis of the siderophore bacillibactin (BGC0000616.1, Figure 10C). For this reason, we assume that the unknown gene cluster might be involved in the biosynthesis of a novel siderophore a with function in iron acquisition. reason, we assume that the unknown gene cluster might be involved in the biosynthesis of a novel siderophore a with function in iron acquisition.

Conclusions
Recently a total of 59 endospore-forming Gram-positive bacteria were isolated from healthy crop plants within fields in Vietnam that were infested with plant-pathogenic nematodes and fungi. According to their draft genome sequences, the majority of the strains were classified as being members of the B. subtilis group, mainly B. velezensis, and the B.cereus group. The remaining 12 strains were provisionally classified as Brevibacillus sp. and Lysinibacillus sp., respectively [1]. In this study we have focused on the members of the Brevibacillus and Lysinibacillus taxon, and a special procedure for enrichment of

Conclusions
Recently a total of 59 endospore-forming Gram-positive bacteria were isolated from healthy crop plants within fields in Vietnam that were infested with plant-pathogenic nematodes and fungi. According to their draft genome sequences, the majority of the strains were classified as being members of the B. subtilis group, mainly B. velezensis, and the B.cereus group. The remaining 12 strains were provisionally classified as Brevibacillus sp. and Lysinibacillus sp., respectively [1]. In this study we have focused on the members of the Brevibacillus and Lysinibacillus taxon, and a special procedure for enrichment of Brevibacillus strains was developed. The sequencing of the whole genomes of Lysinibacillus sp. CD3-6 and of three Brevibacillus strains allowed for a detailed genome analysis of the selected strains, and the identification of several extrachromosomal elements. The novel isolate Brevibacillus sp. DP1.3A, for example, harbored three different plasmids, representing either the low-copy type, whose segregation was directed by ParM, or the high copy type, replicating according to the rolling circle model. The high plasticity of the genomes was also indicated by the presence of numerous genomic islands within the Lysinibacillus and Brevibacillus chromosomes, and the high number of genes present in the pan-genomes.
Furthermore, we have elucidated the taxonomic position of the eleven Brevibacillus strains, and of the Lysinibacillus strain on the base of their complete or draft genome sequences. Five of the Brevibacillus strains were classified as being members of the known species, B. parabrevis (2), and B. porteri (3). The other six Brevibacillus strains, and Lysinibacillus CD3-6 were, according to their ANI and dDDH values, classified as being members of five novel genomospecies.
The main outcome of this work was characterizing the Brevibacillus isolates as potent antagonists of important plant pathogens. Brevibacillus strains were found to be efficient in directly inhibiting the growth of bacterial phytopathogens, such as Clavibacter michiganensis, Xanthomonas campestris, Erwinia amylovora, Dickeya solani, and fungal pathogens, such as Fusarium oxysporum, and other representatives of the Fusarium genus. The direct antagonistic action of the Brevibacillus strains against the phytopathogenic oomycete Phytophthora palmivora was also observed. Brevibacilli displayed high nematicidal activity, and significantly suppressed the formation of root-knots in tomato plants infested with Meloidogyne sp. in greenhouse experiments under controlled conditions.
The Brevibacillus genomes harbored a rich arsenal of known and hitherto uncharacterized BGCs predicted to synthesize ribosomally and non-ribosomally diverse classes of secondary metabolites with potential antagonistic action against plant pathogens. A total of 151 BGCs including 35 uncharacterized BGCs were identified by genome mining in the Brevibacillus genomes, whilst Lysinibacillus sp. CD3-6 harbored only six BGCs. The occurrence of important BGCs in the representatives of different genomospecies is shown in Figure 11. In several cases, it seems that the distribution of BGCs in Brevibacilli is dependent on their taxonomical position. Gene clusters for the synthesis of edeine and petrobactin, for example, were not detected in B. parabrevis, but were present in the species cluster A29, A30, A31, A33, and A34. The gene clusters devoted to tyrocidine synthesis were identified in all isolates. In contrast, gene clusters involved in the synthesis of marthiapeptide and paenilipoheptin were only detected in single strains representing the species cluster A29 and A30, respectively. The paenilipoheptin gene cluster exhibited striking similarity to a corresponding gene cluster in Paenibacillus polymyxa [56] Our results are in line with the previous finding that Brevibacillus sp. Leaf182 is the most potent antagonist within the Arabidopsis phyllosphere-microbiome consisting of 224 strains [9]. The BGC responsible for non-ribosomal synthesis of the antibacterial polyketide macrobrevin, originally described in Leaf182, was detected in three novel genomospecies (A29, A30, and A33). Its presence on the DP1.3A genomic islands suggested that the cluster could be transferred by horizontal gene transfer. Two types of gramicidin gene clusters were detected. One of them which harbored 18 modules has been not described previously. In summary, our genome mining results indicated that Brevibacilli are a treasure box of widely unexplored AMPs and other secondary metabolites. Biocontrol agents developed from plant-associated Brevibacillus strains should extend our present arsenal of bio-based plant protection agents that are necessary for a more sustainable agriculture.  Figure S4, Circular plot of the genome of Brevibacillus sp. M2.1A generated with BioCircos; Figure  S5, GBDP tree (16S rDNA gene sequence based) with Lysinibacillus sp. CD3-6; Figure S6, Lysinibacillus tree inferred with FastME 2.1.6.1 from GBDP distances calculated from 37 entries; Figure S7, Heat Map of the FastANI-Matrix [23] of group A Lysinibacillus genomes; Figure S8, Heat Map of the FastANI-Matrix [23] of Brevibacillus strains isolated from Vietnamese crop plants, Figure S9, Phylogenetic tree constructed with the Vietnamese Brevibacillus isolates; Figure S10, In vitro assay of biococontrol action of Lysinibacillus and Brevibacillus strains against Dickeya solani and Erwinia amylovora; Figure Figure S22, UviB and sactipeptides were detected in Brevibacillus HD1.4A by BA-GEL4; Figure S23, Unknown PK-NRP hybrids pk-orn-x-phe and asn-gly-pk; Figure S24, Unknown PK-NRP hybrids: 16M PK-NRP and 20/25M PK-NRP; Figure S25, Unknown NRPS-like: glu-ser-ilex-phe-orn/ glu-ser-val-val-x-phe-orn/ glu-ser-val-X-X-phe-D-orn (Brevipentin); Figure S26, Unknown NRPS-like: cys-cys-cys (Brevitriazol); Table S1, Genomic islands and their genes in Lysinibacillus sp. CD3-6 predicted with IslandViewer;  M2.1A generated with BioCircos; Figure S5, GBDP tree (16S rDNA gene sequence based) with Lysinibacillus sp. CD3-6; Figure S6, Lysinibacillus tree inferred with FastME 2.1.6.1 from GBDP distances calculated from 37 entries; Figure S7, Heat Map of the FastANI-Matrix [23] of group A Lysinibacillus genomes; Figure S8, Heat Map of the FastANI-Matrix [23] of Brevibacillus strains isolated from Vietnamese crop plants; Figure S9, Phylogenetic tree constructed with the Vietnamese Brevibacillus isolates; Figure S10, In vitro assay of biococontrol action of Lysinibacillus and Brevibacillus strains against Dickeya solani and Erwinia amylovora; Figure Figure S22, UviB and sactipeptides were detected in Brevibacillus HD1.4A by BAGEL4; Figure S23, Unknown PK-NRP hybrids pk-orn-x-phe and asn-gly-pk; Figure S24, Unknown PK-NRP hybrids: 16M PK-NRP and 20/25M PK-NRP; Figure S25, Unknown NRPS-like: glu-ser-ile-x-phe-orn/glu-ser-val-val-x-phe-orn/glu-ser-val-X-X-phe-D-orn (Brevipentin); Figure S26, Unknown NRPS-like: cys-cys-cys (Brevitriazol); Table S1, Genomic islands and their genes in Lysinibacillus sp. CD3-6 predicted with IslandViewer;