Comparative genome analysis of the Lactobacillus brevis species

Background Lactobacillus brevis is a member of the lactic acid bacteria (LAB), and strains of L. brevis have been isolated from silage, as well as from fermented cabbage and other fermented foods. However, this bacterium is also commonly associated with bacterial spoilage of beer. Results In the current study, complete genome sequences of six isolated L. brevis strains were determined. Five of these L. brevis strains were isolated from beer (three isolates) or the brewing environment (two isolates), and were characterized as beer-spoilers or non-beer spoilers, respectively, while the sixth isolate had previously been isolated from silage. The genomic features of 19 L. brevis strains, encompassing the six L. brevis strains described in this study and thirteen L. brevis strains for which complete genome sequences were available in public databases, were analyzed with particular attention to evolutionary aspects and adaptation to beer. Conclusions Comparative genomic analysis highlighted evolution of the taxon allowing niche colonization, notably adaptation to the beer environment, with approximately 50 chromosomal genes acquired by L. brevis beer-spoiler strains representing approximately 2% of their total chromosomal genetic content. These genes primarily encode proteins that are putatively involved in oxidation-reduction reactions, transcription regulation or membrane transport, functions that may be crucial to survive the harsh conditions associated with beer. The study emphasized the role of plasmids in beer spoilage with a number of unique genes identified among L. brevis beer-spoiler strains.


Background
Lactobacillus brevis is a member of the lactic acid bacteria (LAB), which are catalase-negative, non-sporulating, non-motile, rod or coccus-shaped Gram-positive bacteria. L. brevis grows optimally at 30°C and within a pH range of 4 to 6 [1 -3]. It is an obligatory hetero-fermentative bacterium producing lactic acid, carbon dioxide and ethanol and/or acetic acid [1][2][3]. Using phylogenomic and comparative genomic analysis Duar et al. studied the relatedness within the Lactobacillus genus in light of their natural habitat in order to understand their evolutionary history [4]. They assigned lactobacilli species into three main lifestyle categories: free living (environmental and plant isolates), host adapted or as "nomadic" [4].
L. brevis has been isolated from silage, as well as from fermented cabbage and other fermented foods [5,6], and is assigned to the free-living lifestyle group of lactobacilli [4]. L. brevis strains, among other lactobacilli, are of particular interest as they have been granted Qualified Presumption of Safety (QPS) status and consequently have been widely used in the production of fermented foods [1,7]. In addition to their application in food fermentations they are purported to have potential as health-promoting or probiotic bacteria [1,7]. In contrast to these positive attributes, L. brevis strains have also been reported as the causative agent of food or beverage spoilage, in particular of beer [8,9]. LAB species are reported to cause approximately 70% of microbial beer-spoilage incidents, and among this group L. brevis isolates are particularly problematic [10][11][12]. They are associated with the production of malodorous compounds, acidity and/or turbidity with negative impacts on the organoleptic properties of the final product. Bacterial spoilage of beer may result in product withdrawal or recall with concomitant economic losses for the brewing industry [10][11][12]. Beer spoilage by Gram-positive bacteria has been studied previously and the main mechanism of hop resistance known so far involves an active extrusion of the toxic compound using transporters identified as: (a) HorA which functions as an ABC-type multidrug transporter to expel hop compounds, in particular iso-α-acids, from the bacterial cytoplasm, (b) HorC a proton motive force-dependent hop excretion transporter, and (c) the H + -ATPase which increases the pumping of protons released from the hop compounds [13][14][15]. The transmembrane protein HitA is also thought to play a role in the transport of divalent cations, where iso-α-acids exchange protons for cellular divalent cations such as Mn 2+ [16].
To date a number of comparative genome studies of the Lactobacillus genus have been described [1, [17][18][19], some of which have provided insights into the taxonomy of the Lactobacillus genus [3,7], or its fermentation capabilities [3]. Carbohydrate metabolism has been assessed in several Lactobacillus species LAB such as L. casei or L. plantarum [2]. However, a broad comparative genome analysis of the L. brevis species has as yet not been undertaken. Recent advances in next generation sequencing technologies has facilitated a rapid surge in the number of bacterial genomes now available for comparative analysis within a genus or a species.
In the current study, Single-Molecule-Real-Time (SMRT) sequencing technology [20,21] was employed to generate the complete genome sequence of an additional six L. brevis strains isolated from silage and the brewery environment. Using the dataset of 19 complete chromosomal sequences, a comparative genome analysis of the L. brevis taxon was undertaken through an assessment of the phylogeny, pan-and core-genome, and niche adaptation with particular emphasis on adaptation to the brewing environment. The importance of plasmids was also investigated in relation to beer spoilage ability.

Results and discussion
Isolation of L. brevis strains Six L. brevis strains were isolated and included as part of the study, with the aim of expanding the collection of L. brevis genome sequences currently available, as well as studying their ability to grow and colonize harsh environments such as beer. Three L. brevis strains were isolated from beer and characterized as beer-spoilers based on their ability to grow in beer. Two additional L. brevis strains were isolated from the brewing environment, yet lack the ability to grow in beer and are thus not classified as beer-spoilers (Table 1 and Fig. 1). In addition, the sixth L. brevis strain sequenced as part of this study originates from silage [22] and was included as a non-brewing environmental isolate (Table 1). The different L. brevis isolates showed different colony morphology ranging from a dry irregular colony type for L. brevis UCCLBBS449 and UCCLB95 to a slimy and ropy colony type for UCCLBBS124 (Table 1). Plasmid profiling of the different isolates revealed a distinct plasmid content for each isolate. Growth curves in MRS broth demonstrated the unique growth profiles of the individual isolates confirming that the isolates were distinct from each other. Furthermore, only L. brevis UCCLB95, UCCLBBS124 and UCCLBBS449 were characterized as beer-spoilers having the ability to survive and grow in beer, while L. brevis SA-C12, UCCLB521 and UCCLB556 were defined as non-beer-spoilers (Table 1 and Fig. 1).

General genome features
The complete chromosomal sequences of nineteen L. brevis strains were selected for analysis, thirteen of which available at that time were obtained from the NCBI database, while the remaining six were sequenced as part of this study using SMRT sequencing technology ( Table 2). These 19 selected L. brevis strains had been isolated from different ecological niches: silage, fermented food, animal's gut and the brewery environment ( Table 2). The general features of the 19 L. brevis genomes are indicated in Table 3 and include an average chromosome length of 2.49 Mbp (ranging from 2.27 to 2.79 Mbp) and a G + C content of 46%. An average of 2338 predicted CDSs (Coding Sequences) per chromosome were identified to which approximately 78.3% could be assigned a function based on in silico predictions using BLAST (Basic Alignment Search Tool), while the remaining 21.7% were annotated as hypothetical proteins (Table 3). A type II CRISPR-Cas (Clustered Regularly Interspaced Short Palindromic Repeats) locus was found in the chromosome of L. brevis BGP6, L. brevis NPS-QW145 and L. brevis SRCM101106 where variability was observed in the spacer region, distinct spacers were observed in each of these three L. brevis strains suggesting an active system acquiring unique and various spacers for protection against invading DNA over time. Conversely, in the chromosome of L. brevis TMW1.2112 and L. brevis TMW1.2113 ten identical spacers were detected suggesting that these two strains are clonal or that this CRISPR-Cas system is inactive, and that these common spacers originate from a common ancestor that acquired genetic material from viruses/plasmids that it encountered in the past [23]. The L. brevis strain ZLB004 chromosome revealed the presence of four CRISPR locus, one was associated to a type I-E CRISPR-Cas system, a second one was associated to a type II CRISPR-Cas system both potentially active systems. The two other CRISPR locus were not associated to any CRISPR-Cas proteins suggesting inactive system.
PacBio SMRT sequencing was used to determine the diversity and frequency of methylated motifs recognized by Restriction Modification (R/M) systems within the six newly sequenced and annotated L. brevis strains as part of this study. R/M systems constitute one of the biological barriers exerted by a strain against foreign DNA [24]. This analysis revealed the presence of various m6A motifs and allowed the identification of three motifs assignable to Type I R/M system and six motifs assignable to Type II R/M system ( Table 4). The presence of specific methylated motifs was linked to the presence of specific R/M systems in the corresponding L. brevis strains (Table 4). Somewhat surprisingly, L. brevis UCCLB95 does not appear to encode any R/M systems.
The predicted mobilome of L. brevis All complete chromosome sequences were analyzed for the presence of mobile elements such as IS elements and genes specified as encoding transposases. This analysis indicates that L. brevis strains UCCLBBS449 and UCCLB95 contain the highest number of insertion sequence (IS) elements/transposases, 114 and 132, respectively ( Table 3).
The genome sequences were also investigated for prophages, revealing various predicted intact or partial prophage regions (Table 3), displaying in most cases similarity to the published L. brevis temperate bacteriophage LBR48 [25]. The plasmid content of the L. brevis strains is detailed below.

Phylogenetic analysis
The phylogenetic relationship between the genomes of the nineteen L. brevis strains were investigated by a comparative analysis of their 16S rRNA sequences (Fig. 2a). The resulting phylogenetic tree distinguishes five clades (clades A through to E). Clade A represents two L. brevis beer-spoiling strains UCCLB95 and UCCLBBS449 both isolated from spoiled beer, displaying slow growth in nutritive media or beer (. Table 1 and Fig. 1). Clade B encompasses three L. brevis strains: L. brevis SRCM101106 isolated from food L. brevis BDGP6 isolated from the gut of a drosophila and L. brevis NCTC13768 from an unknown isolation source. Clade C is represented by six L. brevis strains, of which one was isolated from food (L. brevis SRCM101174), three from silage (L. brevis SA-C12, ATCC 367 and 100D8) and two strains (L. brevis UCCLB521 and UCCLB556), both isolated from the     (Table 1). These latter two strains may have been introduced into the brewery through raw materials such as cereal grains thus explaining the observed phylogenetic relation to the silage L. brevis isolate SA-C12. Clade D includes five L. brevis strains, all retrieved as beer-spoiler strains from the brewing environment, and all exhibiting a slimy, ropy phenotype (Table 1) [9]. Clade E gathers three L. brevis strains, two isolated from fermented food (L. brevis KB290 and NPS-QW-145) as well as L. brevis ZLB004 isolated from pig's feces.
In order to obtain a more refined view of the phylogeny of the 19 analyzed strains, a so-called phylogenetic supertree was constructed based on 631 conserved orthologous proteins that had been identified as single-copy genes conserved across all investigated chromosomal sequences (19 L. brevis strains and Enterococcus faecalis V583 as an outgroup) [26,27]. This supertree does not display distinct clades separating the L. brevis strains in different groups as was observed with the 16S rRNA phylogenetic tree, suggesting a close relatedness within the species (Fig. 2b). However, upon close inspection of this phylogenetic tree, it appears that L. brevis strains isolated from food and silage cluster on one branch of the tree, while L. brevis brewery isolates cluster on another. The L. brevis strains isolated from gut microbiota BDGP6 and ZLB004 as well as L. brevis strains NCTC13768 and 100D8 gather in a smaller clade.

Pan/core-genome analysis
A pan-genome analysis was performed in order to determine the total number of distinct genes present on the combined chromosomal sequences of the analyzed L. brevis strains. The pan-genome curve displays an asymptotic trend, growing with an average rate of 136 genes per genome in the first nine iterations, then the number of new genes decreased leading to a total pan-genome content of 3968 genes (Fig. 3). The mathematical function displayed on the graph reveals an exponential value lower than 0.5 indicating that the pan-genome is in a closed state. The core genome was determined to   (Fig. 3). As an overall result, both analyses indicate a closed pan-genome for L. brevis species, while also indicating that a sufficient number of strains had been included to adequately describe the genetic repertoire of the L. brevis species.

Comparative analysis of orthologous genes
The comparative analysis used in this study was based on chromosomal sequences. The core genome of 1428 genes is divided in 1170 orthologous gene families (single copy) and 258 paralogous gene families (multi-copy). Unique gene families to each chromosome were also recorded and 246 unique gene families were identified across the nineteen L. brevis strains (Fig. 4a). Functional assignment efforts revealed that 75.2% of the unique gene families encoded proteins of unknown function (hypothetical proteins), while 4.5% encoded (pro)phage-related proteins. The remaining unique gene families encode proteins that could benefit the fitness of the strain such as CRISPR-Cas system (e.g. Type I-E CRISPR Cas system in L. brevis ZLB004), restriction-modification systems (e.g. Type I R/M system in L. brevis strain UCCLBBS124), or cell wall polysaccharide synthesis (e.g. genes predicted to encode glycosyltransferases and a polysaccharide polymerase only found in the L. brevis strain NPS-QW-145).
In order to further investigate the functionality and diversity encoded by the core and dispensable genomes, a Cluster of Orthologous Group (COG) analysis was employed. The genome content of the 19 selected L. brevis strains was classified into different groups depending on their function. More than 75% were predicted to be involved in housekeeping functions, vital for the strain to grow such as those participating in transcription or translation. Approximately 16% of the genes were assigned to COG groups with only a general function predicted or of unknown function (Fig. 4b).

Evolution and adaptation to beer environment
When plotting the number of CDSs as a function of genome size for the different L. brevis strains, the group exhibiting the largest genome size as well as the highest number of CDS are L. brevis strains isolated from beer and characterized as beer-spoilers as well as the L. brevis strain BDGP6 displaying the biggest CDS number. L. brevis strains known to be beer-spoilers possess an average of 2385 CDS, while those isolated from food, silage, animal's gut and non-beer spoiling brewery isolates display an average of 2311 CDSs (Fig. 5). This observation suggests a link to adaptation to a new environment, i.e. the beer or brewery environment, which may have necessitated the acquisition of novel genes and corresponding functions in order to survive in this harsh environment. To understand if the beer-spoiling strains had acquired a specific set of genes or associated functions, genes that may putatively be associated to beer adaptation were first predicted to be those that would be present in the genomes of at least four beer spoiling strains (Table 5). From this analysis, 58 genes of interest were highlighted as well as 26 hypothetical proteins. Out of these 58 genes, approximately 21% encode proteins related to oxido-reduction reactions (Flavodoxin, oxidoreductases and short-chain dehydrogenases), 22% are linked to transcription (transcriptional regulators, RNA polymerase sigma-24 subunit ECF subfamily), 21% encode membrane and cell surface proteins and 14% are related to membrane transport (MFS transporter, permease, ABC transporters) ( Table 5).
When exposed to beer, bacteria are subjected to various stresses, among them a low pH (3.8-4.7) and hop compounds [10]. When iso-α-acids enter the cell cytoplasm, they dissociate into hop anions and protons decreasing the intracellular pH [10]. Therefore, bacteria would have to adapt in order to regulate their internal pH in order to survive [28][29][30]. Furthermore, in beer the presence of ethanol (0.5-10% w/w) causes oxidative stress in bacteria, this results in the production of Reactive Oxygen Species (ROS) such as hydrogen peroxide and free radicals leading to cell damage [31,32]. Despite the stress and harsh environment imposed by the beer environment, some bacteria have evolved and acclimatized to this new medium. It may thus be possible that some of the L. brevis strains acquired additional functions which allow them to grow and survive in beer and which has led to an increased genome size. The fact that 21% of these chromosomal genes encode proteins related to redox reactions is of interest and suggests a link between L. brevis beer-spoiler strains and oxidative stress response. Six of the 12 genes that encode functions relating to oxido-reduction reactions present in at least four beer-spoiler L. brevis strains are predicted to encode NADH oxidoreductases and short-chain dehydrogenases/reductases (SDRs). These proteins are part of the large family of NAD(P)(H)-dependent oxidoreductases and are believed to behave as scaffold proteins for an NAD(P)(H) redox sensor system [33]. In previous studies, the role of SDRs during oxidative stress was highlighted in species such as Bacillus subtilis where they are required for survival in severe ethanol stress [34], or in Burkholderia pseudomallei during salt stress [35].
Furthermore, 22% of the chromosomal genes that seem to be specifically associated with beer-spoiling L. brevis strains are linked to transcriptional regulation, suggesting that these regulators act on specific genes to control their expression and confer an advantage when present in beer. It would be interesting to study which genes are affected by these transcriptional regulators to assess the mechanisms employed to survive in this harsh environment. Of the L. brevis beer-spoiler specific chromosomal genes 21% encode membrane and cell surface proteins suggesting an adaptation to survive in the harsh beer environment. 14% encode proteins associated with membrane transport such as permeases and ABC transporters suggesting exchange between the strain and its environment and possibly a role in extrusion where the L. brevis isolate would expel protons or iso-α-acids in order to survive and thrive in beer, as has been described previously [10,13,15].
Interestingly some of the chromosomal genes identified among L. brevis beer-spoiler strains in this analysis had also been highlighted in a previous study as beer-spoilage diagnostic marker genes (DMG) [36]. These genes are predicted to code for an oligogalacturonide transporter, a short chain dehydrogenase and a RNA polymerase sigma factor ECF subfamily, which reinforces the hypothesis for their involvement in beer spoilage adaptation.

The role of plasmids in adaptation to beer environment
Different proteins involved in beer spoilage have been identified on plasmids indicating the importance of  plasmids for bacterial strains in beer spoilage. This might suggest a role for plasmid mobilization and transfer between bacterial strains throughout evolution to adapt to a new environment such as beer.
The nineteen analyzed L. brevis strains were predicted to harbour up to nine plasmids with strains L. brevis KB290 and L. brevis UCCLBBS449 exhibiting the largest plasmid complements of the assessed strains. The plasmid size ranges from 2.8 Kb to 107.0 Kb ( Table 3). The number of plasmids and their size do not appear to be linked to the isolation source of the L. brevis strains (e.g. four plasmids for L. brevis SRCM101106 versus nine plasmids for L. brevis KB290, both isolated from fermented food) or to the beer spoilage ability of the isolate (two plasmids for L. brevis UCCLB95 versus nine plasmids for L. brevis UCCLBBS449 both characterised as beer-spoilers).
Investigating analogies between plasmids among L. brevis beer-spoiler strains revealed that the plasmid content of L. brevis TMW1.2108 and L. brevis TMW1.2111 were very similar. Indeed, the six plasmids of L. brevis TMW1.2111 show at least 90% identity to seven of the eight plasmids contained by strain L. brevis TMW1.2108, with the exception of plasmid TMW1.2108-5. Similarly, L. brevis strains TMW1.2112 and TMW1.2113 present a close plasmid composition as the four plasmids of L. brevis TMW1.2113 are at least 90% identical to four out of five plasmids of L. brevis TMW1.2112 with the exception of plasmid TMW1.2112-1.
Out of the 38 plasmids shared between L. brevis beer-spoiler strains, only three plasmids seem to be unique, sharing less than 10% similarity with any other plasmid. These three plasmids were found in L. brevis Table 5 List of genes identified in the chromosome sequence of at least four L. brevis beer spoiler strains. 26 genes coding for hypothetical proteins were also identified (Continued) UCCLBBS449 (UCCLBBS449_pF, UCCLBBS449_pH and UCCLBBS449_pI) and contain mostly genes coding for hypothetical proteins, replication proteins as well as genes coding for proteins involved in conjugation such as mobilization proteins and a relaxase. Refined analysis of specific genes shared only between at least three L. brevis beer-spoiler strains, identified only twenty-five genes ( Table 6). In this list of unique genes shared only between L. brevis beer-spoiler strains, the gene coding for the membrane protein HorC is noteworthy, as it is known to be involved in beer spoilage [14] and is present in all L. brevis beer-spoiler strains with the exception of L. brevis TMW1.2113.
Interestingly, the gene encoding the ABC transporter HorA [13] and present in L. brevis beer-spoiler strains TMW1.2108, TMW1.2111, TMW1.2113, UCCLBBS124 and UCCLBBS449 does not figure in this list as a similar protein can be found in plasmid sequences of the L. brevis strains KB290, SRCM101106 isolated from fermented food and L. brevis UCCLB556 isolated from the brewery and characterized as a non-beer spoiler. Moreover, the transmembrane protein HitA [16] has been identified only in two of the L. brevis beer-spoiler strains UCCLBBS449 and TMW1.2112. These observations reinforce the statement that involvement of these genes in beer survival and spoilage is not always verified as they are not consistently present in beer-spoiler organisms nor are always corresponding to beer spoilage ability if present in a strain [15]. The list of genes present only in L. brevis beer-spoiler strains shows that strains L. brevis TMW1.2108 and TMW1.2111 possess more than 90% of these genes whereas L. brevis UCCLB95 only possesses one gene coding for the membrane transporter HorC ( Table 6). The remainder of the L. brevis beer-spoiler strains carry approximately 50% of these particular genes ( Table 6).
Out of these 25 unique genes shared among L. brevis beer-spoiler strains approximately 25% code for hypothetical proteins of unknown function. Meanwhile,~30% of these genes appear to encode cell wall-associated proteins either as membrane transporters (MFS transporter, HorC) or as cell wall biosynthesis (lipopolysaccharide glycosyltransferases, acyltransferases). As mentioned above, a beer-spoiling strain would need to extrude toxic compounds using transporters and adapt its cell wall composition to survive the harsh beer environment. A smaller portion of these unique genes are linked to transcription regulation, replication or mobilome.
Interestingly, some of the plasmid-associated genes identified among L. brevis beer-spoiler strains in this analysis have also been highlighted previously as unique attributes of beer-spoiling strain plasmids [36]. The gene coding for the CrcB like-protein involved in ion transport was found on plasmid BSO 464-2 of the L. brevis beer-spoiler strain BSO 464 as well as a gene coding for enolase involved in glucose metabolism. A gene coding for cytosine deaminase is present in five out of the seven L. brevis beer-spoiler strains used in this analysis ( Table  6) and was identified as a unique attribute on the plasmid pPECL-8 of the beer-spoiler Pediococcus claussenii ATCC BAA-344 [36]. Moreover, a plasmid-associated gene coding for a glycosyltransferase family 2 was highlighted in the analysis (Table 6), this protein was associated with excess β-glucan formation leading to a slimy ropy phenotype in the L. brevis beer-spoiler TMW1.2112 [37]. This slimy phenotype was observed in the L. brevis strain UCCLBBS124 first described in this study (Table 1), and the gene coding for the glycosyltransferase family 2 was identified on one of its plasmid UCCLBBS124_pB.
This overall examination of plasmid-associated genes shows the importance of extrachromosomal DNA in beer spoilage adaptation and opens new possibilities for understanding the beer spoilage process with an updated list of potential proteins of interest only present in L. brevis beer-spoiler strains.

Conclusions
The isolation and genome sequencing of six L. brevis strains combined with thirteen additional, publicly available L. brevis genomes allowed a comparative genome analysis of the L. brevis species. The deduced pan-genome of these L. brevis isolates appears to be in a closed state, indicating that the representatives used in this study are sufficient to describe the genetic diversity of the taxon. Throughout evolution, it appears that L. brevis strains specified and differentiated one from another by acquiring plasmids and prophages, despite for the presence of CRISPR-Cas and R/M systems which may have limited such foreign DNA invasion events. These latter systems are of relevance for future functional investigations that may necessitate the development of DNA transfer and/or mutagenesis tools. L. brevis strains represent a significant threat for the brewing industry being the most common cause of beer spoilage; however, this spoiling ability is strain specific. The comparative genome analysis performed here highlights that L. brevis strains with the ability to grow in beer possess a higher number of CDSs in their overall chromosomal sequences. This observation suggests a link to evolution and adaptation to beer in which the strain would have acquired novel genes and functions in order to adapt and survive in the harsh environment that beer represents. The role(s) of the "acquired" or beer-specific CDSs revealed that almost a quarter of these are linked to oxido-reduction reactions, possibly playing a role in the response to oxidative stress. Another 22% are linked to transcription regulation, 21% encode cell surface proteins while 14% are encoding membrane transport related proteins and possibly associated to harmful compound extrusion encountered by the L. brevis strains when surviving and growing in beer. Additional genetic diversification of these L. brevis strains is expected to have occurred through plasmid acquisition that also likely contributes to beer adaptation. The plasmid content analysis of the different L. brevis beer-spoiler strains highlighted the presence of unique proteins shared among these strains. These proteins are mostly hypothetical proteins while approximately 30% are linked to membrane transport, and cell-wall synthesis. These observations demonstrate the complexity of microorganisms' beer spoilage ability and suggests that adaptation of the L. brevis strain to beer is a complex process, not due to the action of only one specific gene, but more likely the intervention of a complex, multi-factorial response.

Isolation of L. brevis strains
Five distinct L. brevis strains (UCCLB521, UCCLB556, UCCLB95, UCCLBBS124 and UCCLBBS449) were isolated from the brewing environment, while SA-C12 had previously been isolated from silage [22]. The strains were characterized by evaluating their plasmid content as well as growth curve profiles in MRS broth or in beer at 30°C. Plasmids were isolated after overnight growth of the L. brevis strains in MRS broth at 30°C, cells were pelleted by centrifugation for 10 min at 5000 rpm followed by lysozyme treatment (30 mg/mL lysozyme in TE + 25% sucrose) at 37°C for 30 min. Plasmids were extracted using the GeneJET Plasmid Miniprep Kit (Thermo Scientific™). Plasmid profiles of the different L. brevis isolates were analyzed using a 1% agarose gel. Growth curve profiles in MRS broth or in beer were performed at 30°C by hourly OD 600nm measurements for a period of 55 h. Moreover, colony morphology was recorded following growth on MRS agar plate at 30°C.

Sequencing and annotation
L. brevis strains were streaked on MRS agar plates and grown at 30°C for 24 h. For each L. brevis strain, a single colony was inoculated into MRS broth and grown overnight at 30°C. Cells were pelleted by centrifugation at 5000 rpm for 10 min. The supernatant was removed and the pelleted cells were frozen at − 20°C prior sending for sequencing. Sequencing was performed using the PacBio SMRT next generation sequencing technology (performed by GATC Biotech, Germany). De novo genome assemblies were performed using the Pacific Biosciences SMRT Portal analysis platform. Open Reading Frame (ORF) or coding sequence (CDS) prediction was performed using Prodigal prediction software [38] and confirmed using BLASTX alignments [39]. Automatic annotations were refined using Artemis v16.0.0 where ORF predictions were manually checked, start codons adjusted and pseudogenes identified. Transfer RNA (tRNA) genes were predicted using tRNA-scan-SE v2.0 [40], while ribosomal RNA (rRNA) genes were identified using RNAmmer v1.2 [41]. The sixteen L. brevis genomes obtained from NCBI were re-annotated as described above in order to treat identically all sequenced genomes used in this study.

Methylome analysis
Following de novo genome assembly, the RS_Modificatio-n_and_Motif_Analysis.1 protocol of the SMRT Analysis portal was employed for base modification and methylated motif detection. This analysis was performed on L. brevis strains sequenced, assembled and annotated as part of this study. Methylation motifs with a score equal or higher than 40 (corresponding to a P-value of < 0.0001) were considered specific and used for further analysis. ORFs of genomes were investigated for the presence of restriction/modification systems using the BLASTP alignment function of the REBASE database [42] (cut-off E-value of 0.0001; with at least 30% similarity over at least 80% of the sequence length). A comparative genome analysis was employed to associate the presence of R/M system-encoding genes with the presence of methylation motif(s).

Comparative genomics
All protein sequence comparisons were performed using all-against-all, bi-directional BLAST alignments [39]. An alignment cut-off value of E-value 0.0001, and a similarity cut-off level of at least 30% amino acid identity across 80% of the sequence length was used. Results were analyzed with the Markov Clustering Algorithm (MCL) [43] and proteins encoded were categorized in predicted functional groups based on COG (Clusters of Orthologous Groups) assignments [44].

Phylogenetic analysis
The supertree was prepared using the BLAST-based comparative approach described above in order to identify chromosomal orthologous proteins. The set of chromosomal orthologous proteins was concatenated for each strain and an ungapped alignment was performed using MUSCLE v3.8.31 [45]. The phylogenetic tree was computed using the maximum-likelihood method in PhyML v3.0 and bootstrapped employing 1000 replicates [46]. The final tree was visualized using MEGA7. A tree based on 16 S rRNA genes was constructed using clustalw and visualized via ITOL (Interactive Tree Of Life) [47]. The chromosome sequence of Enterococcus faecalis V583 (Accession: AE016830) was included as an outgroup.