Analyses of the probiotic property and stress resistance-related genes of Lactococcus lactis subsp. lactis NCDO 2118 through comparative genomics and in vitro assays

Lactococcus lactis subsp. lactis NCDO 2118 was recently reported to alleviate colitis symptoms via its anti-inflammatory and immunomodulatory activities, which are exerted by exported proteins that are not produced by L. lactis subsp. lactis IL1403. Here, we used in vitro and in silico approaches to characterize the genomic structure, the safety aspects, and the immunomodulatory activity of this strain. Through comparative genomics, we identified genomic islands, phage regions, bile salt and acid stress resistance genes, bacteriocins, adhesion-related and antibiotic resistance genes, and genes encoding proteins that are putatively secreted, expressed in vitro and absent from IL1403. The high degree of similarity between all Lactococcus suggests that the Symbiotic Islands commonly shared by both NCDO 2118 and KF147 may be responsible for their close relationship and their adaptation to plants. The predicted bacteriocins may play an important role against the invasion of competing strains. The genes related to the acid and bile salt stresses may play important roles in gastrointestinal tract survival, whereas the adhesion proteins are important for persistence in the gut, culminating in the competitive exclusion of other bacteria. Finally, the five secreted and expressed proteins may be important targets for studies of new anti-inflammatory and immunomodulatory proteins. Altogether, the analyses performed here highlight the potential use of this strain as a target for the future development of probiotic foods.

Introduction or competing for nutrients and epithelial adhesion sites. LAB produce different antimicrobial components, such as organic acids, hydrogen peroxide, carbon peroxide, diacetyl, low molecular weight antimicrobial substances, bacteriocins and adhesion inhibitors. The adhesiveness of LAB may involve passive forces, electrostatic interactions, hydrophobic steric forces, lipoteichoic acids, and lectins [21]. The hydrophobic nature of the outermost surface of microorganisms facilitates the adhesion of bacteria to the host epithelium, thereby conferring competitive advantages during the colonization of the GIT [22]. The antimicrobial susceptibility of intestinal microorganisms is an important criterion for the selection of probiotic strains, mainly due to the potential transfer of those genes to pathogenic or commensal bacteria that inhabit the GIT [23]. In the following sections, we present comparative genomic analyses of L. lactis NCDO 2118 and other Lactococcus species and predict genes that putatively code for acid stress resistance proteins, bacteriocins, adhesins and exported proteins.

General features, phylogenomics and synteny analyses
The general genomic features of all genomes used in this work are summarized in Table 1.
From the heatmap created with Gegenees (Fig 1), it is possible to visualize a high similarity between the subspecies of Lactococcus, with nucleotide similarities ranging from 40% to 100%. Additionally, the species and subspecies clustered separately, creating 3 green blocks of strains at the chart, represented by L. lactis subsp. lactis and L. lactis subsp. cremoris, with similarities ranging from 91% to 100%, and L. garvieae, in which the two strains of this species were 100% similar to each other.
On the phylogenetic tree created using 16S, the species and subspecies also clustered together, forming two main clades corresponding to the best similarity among L. lactis subsp. lactis and L. lactis subsp. cremoris (Fig 1). Additionally, L. garvieae strains appeared in an outside node compared to L. lactis species and are the two most distinct and distant species of Lactococcus on the heatmap and phylogenetic tree. Briefly, on the heatmap, the degree of intraspecies similarity varies from 91% to 100%, whereas interspecies similarity varies from 40% to 65%.
From the genome synteny analysis (S1 Fig), all strains from L. lactis subsp. lactis presented a high degree of synteny, where the most conserved genome compared to L. lactis NCDO 2118 (chosen as reference genome) was L. lactis KF147. Additionally, we performed a comparison with the plasmids of L. lactis NCDO 2118 and L. lactis KF147 strains. However, we verified a high degree of similarity from the beginning to the end of each plasmid sequence, meaning that they possibly harbored the same plasmid (data not shown).
(2) all strains from L. lactis subsp. lactis and L. lactis subsp. cremoris (non-pathogenic dataset), and (3) all strains from this study (including L. garviae). The number of metabolic pathways harbored by each genome varies from 148 to 206, with a general mean of~183 pathways. Both L. garvieae strains contained 148 metabolic pathways, L. lactis subsp. lactis showed an average of~192 metabolic pathways, and L. lactis subsp. cremoris showed~186 pathways.
The main differences were that the strain L. lactis NCDO 2118 contains more peptidoglycan biosynthesis pathways than L. lactis KF147 and L. lactis IL1403 strains. Other exclusive metabolic features of L. lactis NCDO 2118 in this context were complete anaerobic respiration pathways, fermentation of pyruvate to acetate, fermentation of fumarate, complete heterolactic fermentation, valine degradation, L-serine degradation, ammonia assimilation to glutamate, complete superpathway of acetate utilization and formation, protein degradation, initial pathway of sucrose degradation I, valine degradation, lysine degradation I and acyl-ACP thioesterase pathway (S1 Table).

Streptococcus thermophilus LMD-9**
In the GIPSy predictions, we identified 9 Genomic Islands (GEIs), 5 Metabolic Islands (MIs), 4 Symbiotic Islands (SIs) and 3 Miscellaneous Islands (MSIs), which were predicted as harboring both metabolic and symbiotic related factors. The GEIs are listed in S2 Table. All SIs were only partially present in the other strains, except for SI4, which was absent from all L. garviae strains, L. lactis subsp. cremoris strains and L. lactis IL 1403 (Fig 2). Additionally, all MIs presented regions of deletions in the pathogenic species L. garviae. The most prominent GEIs were MI3, which was only present in the two strains isolated from plants (L. lactis NCDO 2118 and L. lactis KF147), and MSI 2, which presented the biggest region of deletion in all Lactococcus, except for L. lactis NCDO 2118 and L. lactis KF147.

Antibiotic resistance
LAB that are widely used as probiotics or in starter cultures have the potential to host antibiotic resistance genes, thereby presenting a risk of transferring such genes to many lactic acid bacteria and other pathogenic bacteria [23]. In the antibiogram assay, L. lactis NCDO 2118 was susceptible to ceftriaxone, erythromycin, tetracycline, ampicillin, penicillin and chloramphenicol and resistant to vancomycin, oxacillin and amikacin antibiotics (Table 3). Additionally, we tried to correlate the antibiogram profile with the genome content of L. lactis NCDO 2118, which presented 22 antibiotic resistance-related genes putatively coding for a VanZ family protein (NCDO2218_1094), penicillin-binding proteins (NCDO2118_0402, NCDO2118_0445, NCDO2118_0526, NCDO2118_0880 and NCDO2118_2216), and multidrug efflux pump proteins (Table 4). Additionally, no antibiotic resistance related gene presented deviation in its genomic signature.

Identification of genes involved in acid stress and bile salt resistance
We searched the genome sequence of L. lactis NCDO 2118 for genes previously shown to be differentially expressed on cells cultivated under low and optimum pH (5.1 and 6.5, respectively) in L. cremoris MG1363 [38] (Table 5). Additionally, we also searched for genes differentially regulated by bile exposure in Bifidobacterium animalis and Bifidobacterium longum NCIMB 8809 [39; 40] and/or identified on the total proteome and surfome of Lactobacillus rhamnosus GG using proteomics analyses (Table 5). Here, we identified some genes in L. lactis NCDO 2118 that were previously reported to be involved in the acid stress response, including genes coding for chaperones (dnaK) and stringent response. Additionally, DnaK and Enolase are plasminogen receptors involved in bile modulation during intestinal colonization. Additionally, we assayed L. lactis NCDO 2118 to see how it responds to the challenges of acid pH and bile salt secretion in the gastrointestinal tract. When in contact with artificial gastric juice, 48% of the L. lactis NCDO 2118 was not inhibited and was able to grow after acid pH challenge, whereas the contact with bile salts inhibited 95% of the bacteria growth, showing a high sensibility, as a result of three independent experiments (S2 Fig). Identification of genes coding for adhesins and adhesion-related proteins Based on literature data, we predicted proteins involved in the adhesion mechanisms of L. lactis NCDO 2118, shown in Table 6. L. lactis NCDO 2118 harbors 19 genes putatively coding for adhesion-related proteins, such as the gene chiA (NCDO2118_2053) and the genes coding for the Chitin binding protein (CBP-NCDO2118_2054) and the laminin-binding protein (NCDO2118_1446).
To determine whether L. lactis NCDO 2118 exhibited adhesive ability, corroborating the in silico data, we performed microbial adhesion to solvents (MATS) experiments, which demonstrated a moderate cell surface hydrophobicity, as suggested by Nader-Macías et al., (2008) [41], with 52% association with xylene.
In L. lactis NCDO 2118, BAGEL predicted one bacteriocin for each of the three classes ( Fig  3): a lanthipeptide (class I), NCDO2118_1768 (putative Bacteriocin-lactococcin-A-class II) and a putative bacteriocin (class III), located between NCDO2118_2257 and NCDO2118_2258. The class III putative bacteriocin was not described in the L. lactis NCDO 2118 genome, possibly because the gene-finding methodology failed to predict it. The bacteriocin of class I is a lantibiotic Nisin coded by the nisZ gene (NCDO2118_1272), a natural variant of nisA [44]. Briefly, Nisin is commonly produced by strains of L. lactis, and the cluster of genes coding for the nisin biosynthesis proteins consists of 11 genes: nisABTCIP (biosynthesis and immunity), nisFEG Comparative genomics and in vitro assays of the probiotic Lactococcus lactis subsp. lactis NCDO 2118 (immunity) and the two-component regulatory system nisRK [45]. L. lactis NCDO 2118 harbors a nisBCIP operon (where nisP is a pseudogene), a nisRK two-component system and a nisFEG operon. Additionally, BAGEL has predicted the presence of another putative bacteriocin between NCDO2118_1258 and NCDO2118_1259 that is located close to the class I cluster of genes. However, the amino acid sequence predicted from this region only presents similarity to a hypothetical protein. Lactococcin A is a class IId, non-pediocin-like, single-peptide bacteriocin normally produced by strains of L. lactis. Four genes are responsible for the biosynthesis of lactococcin: the lactococcin-A coding gene, one immunity gene and the dedicated ABC transporter system along with its accessory protein. L. lactis NCDO 2118 harbors an immunity protein (NCDO2118_1767) and lactococcin-A (NCDO2118_1768). As for the class III prediction, the predicted putative bacteriocin is located upstream of two hypothetical proteins (NCDO2118_2258 and NCDO2118_2259); however, little is known about the organization of the gene cluster of class III bacteriocins [45], and the putative bacteriocin predicted by BAGEL only presents similarity to hypothetical proteins in GENBANK. Moreover, an additional bacteriocin-coding gene was harbored by GEI 9 (S2 Table), which was not predicted by BAGEL. Through blast analyses, we found a significant amino acid similarity, with identities varying from 76% to 98%, between this gene and a bacteriocin-coding gene from other L. lactis in the UNIPROT and NCBI BLAST databases. However, many of the genes were also described as hypothetical proteins. In addition, we also searched for other genes that could possibly play a role in the competitive exclusion of other bacteria. A Lyzozyme M1 and a Macrolide biosynthetic protein encoding genes were also included in S3 Table after manual curation in the L. lactis NCDO 2118 genome.
In the present study, a deferred agar spot assay was used for the initial determination of antagonistic activity via diffusible compound(s) produced by L. lactis NCDO 2118. To assay whether L. lactis NCDO 2118 could affect the growth of pathogenic bacteria, we used an approach to measure its antagonistic activity against the strains Salmonella enterica ATCC 14028, Escherichia coli ATCC 25723, Staphylococcus aureus 29213, Bacillus cereus ATCC 11778, Listeria monocytogenes ATCC 15313, Enterococcus faecalis ATCC 19433 and Pseudomonas aeruginosa ATCC 5853. L. lactis NCDO 2118 showed no effect on the growth of the abovementioned pathogenic strains.

In silico identification of putatively secreted proteins
Here, we strove to predict genes encoding secreted proteins from L. lactis NCDO 2118 that are absent from the genomes of the strains L. lactis IL1403 and L. cremoris MG1363, as the secreted proteins of L. lactis NCDO 2118 are possibly responsible for the immunomodulatory effects of this transient bacterium inside the host [18].
To predict the secreted proteins, we used the software SurfG+, which classifies the proteins using an identification approach based on the presence/absence of signal peptides, signal retention and transmembrane helix [46], which are correlated with the cell wall thickness of the bacteria. To determine the cell wall thickness, we made photomicrographs of L. lactis NCDO 2118 (Fig 4); the cell wall was measured more than 270 times, showing an average size of~20 nm, and this value was used to determine the motifs. If none of the motifs were found in the protein sequence, SurfG+ characterized the protein as cytoplasmic (CYT) [47]. Using SurfG+, we predicted 94 secreted proteins in L. lactis NCDO 2118.
From this data, the secreted proteins of L. lactis NCDO 2118 were compared to the proteins identified in L. lactis IL1403 using OrthoMCL [48]. In this comparison, 26 of the secreted proteins were exclusive from L. lactis NCDO 2118. Because the probiotic effect was searched using secreted proteins previously expressed in vitro, we searched for proteins that were expressed in  Table 7). The complete lists of genes identified in proteomic analyses, in the prediction of subcellular location and the exclusive proteins of L. lactis NCDO 2118 are described in S4 Table. Discussion Genomic characterization of L. lactis NCDO 2118 and comparison with other species The genomic lengths of the Lactococcus species analyzed here are highly variable (from~1.95 Mb to~2.60 Mb). However, the finding that L. garvieae strains have the smallest genomes compared to L. lactis strains is in agreement with the lifestyle of L. garvieae, isolated from diseased fish. Because pathogenic bacteria may scavenge compounds from the host for their own metabolism, they tend to lose genes involved in biosynthetic pathways, thus, presenting smaller genomes [49]. The high similarity at the subspecies level may be related with some specific characteristics already described in literature, such as the propensity of L. lactis subsp lactis to form longer chains. Besides, L. lactis subsp. lactis are able to produce GABA, ammonia from arginine, carbon dioxide and diacetyl formation from citrate as opposing to L. lactis subsp. cremoris subspecies. Additionally, analyses using southern hybridization, PFGE, 16 rRNA and housekeeping genes (atpA, rpoA, pheS, pepN, bcaT, pepX) showed two separate clusters formed by L. lactis subsp. lactis and L. lactis subsp. cremoris with a low degree of similarity between them [50][51][52].
From the genome synteny analyses, we have found a high degree of synteny between L. lactis KF147 and L. lactis IL1403, which was already reported in a previous work [25]. However, there was no other genome sequence of any Lactococcus species correlated with plants available at the time the work was performed. Here, we found that the most conserved genome compared to L. lactis NCDO 2118 was L. lactis KF147. The material of fermented plant covers a highly variable niche according to some characteristics as: chemical composition and physical conditions. Thus, plant-related strains posses a great metabolic diversity that certainly extrapolates that from dairy strains [53].
Finally, although L. lactis NCDO 2118 shares several pathways in common with L. lactis KF147 and L. lactis IL1403, it presents several exclusive metabolic features that may be explored for future utilization in industry.
Evaluation of safety aspects in the use of L. lactis NCDO 2118 by genome plasticity and antibiotic resistance approaches Plasmid-linked antibiotic resistance is not very common among LAB, but it does occur, and safety implications should be taken into consideration. Strains harboring resistance plasmids should not be used as human or animal probiotics. Checking the ability of a proposed probiotic strain to act as a donor for conjugative antibiotic resistance genes may be a sensible precaution in some instances [54].
To provide a better understanding of the putative plasticity of L. lactis NCDO 2118, we have predicted putative phage and genomic islands of this species. The presence of phage regions may contribute to the acquirement of antibiotic resistance, the ability to survive in a new environment, the improvement of adhesion ability, or even to turning the bacteria pathogenic [55]. Here, we found 5 phages; the 3 intact phages harbored important genes such as rusA, arsC1, arsC3, amtB, rpmE2, carA, pyrB, pyrP and pepT. The rusA gene is associated with the prophage sequences of several genera of bacteria, including Bacillus, Streptococcus, Staphylococcus, and Enterococcus, and it is also present in Lactococcus lactis phage r1t [56]. The arsC1 gene is related to arsenate resistance in Corynebacterium glutamicum [57]. arsC3 codes for a thioredoxin-dependent arsenate reductase of the Mycobacterium sp. A33 [58]. amtB is a gene of the ammonia transporter family, which is found in eubacteria, archaea, fungus, plants and animals, whereas in prokaryotes, its homologue is co-transcribed with a PII paralogue, GlnK, in response to nitrogen limitation [59]. The rpmE2 gene codes for a L31 ribosomal protein.
The genes carA, pyrB and pyrP are organized as an operon in L. cremoris MG1363, where pyrP encodes a membrane-bound protein with high affinity to uracil permease and pyrimidines, and pyrB and carA encode pyrimidine biosynthetic enzymes [60]. Finally, the gene pepT encodes for a tripeptidase.
Additionally, we predicted 9 GEIs, 5 MIs, 4 SIs and 3 MSIs in the genome sequence of L. lactis NCDO 2118. Interestingly, all MIs present deletions in the pathogenic species L. garvieae, which is a common feature of pathogenic bacteria that adapted to scavenge nutrients from the host [61]. Additionally, MI3 is only present in the L. lactis NCDO 2118 and L. lactis KF147 and may be important for the adaptation of those strains to plants.
We have also assayed L. lactis NCDO 2118, aiming to characterize its antibiotic resistance profile. L. lactis NCDO 2118 is susceptible to most of the antibiotics assayed here. Although L. lactis NCDO 2118 presented resistance to oxacillin and susceptibility to penicillin, it only harbored genes coding for a VanZ family protein, which may be related to the vancomycin resistance, penicillin-binding proteins, and multidrug efflux pump proteins.
The efflux pumps are membrane transporter proteins responsible for the extrusion of relevant antibiotics, which are found in both Gram-positive and Gram-negative bacteria [62; 63]. Penicillin-binding proteins are transpeptidases or caboxypeptidases that harbor specific motifs that limit the active site serine penicillin-recognizing enzyme family, including class A and C β-lactamases [64]. Vancomycin is a glycopeptide antibiotic used in severe infections. Some species used in the food industry or found naturally in raw food material present an intrinsic resistance to vancomycin, including L. rhamnosus, L. casei, Lactobacillus plantarum, and Leuconostoc lactis [65].
Finally, although L. lactis NCDO 2118 does present genes putatively coding for antibiotic resistance-related proteins, none of those genes present anomalous G+C or codon usage deviation, nor are they harbored by the putative horizontally acquired regions predicted by GIPSy or PHAST. More interestingly, no Resistance Island was identified in L. lactis NCDO 2118, corroborating its safety aspects [66].

In vitro and in silico analyses of survival, exclusion mechanisms and probiotic properties of L. lactis NCDO 2118
Susceptibility of L. lactis NCDO 2118 to acid stress and bile salts. Concerning the acid stress, lowering the intracellular pH reduces the transmembrane pH difference and the activity of acid-sensitive enzymes and damages proteins and DNA [67].The first mechanism used by L. lactis species to cope with acid stress is to maintain a low intracellular pH (pHi) by using membrane ATPase FoF1 [68; 69] and the generation of alkaline substances through the catabolism of amino acids (deamination, for example) [70; 71]. Bile salts, on the other hand, are surface-active, amphipathic molecules with a potent antimicrobial activity, and they act as detergents that disrupt biological membranes [67]. The percentage of resistance to bile salts also tends to vary among LAB and even between strains of the same species [72].
Here, we have identified 25 and 16 genes previously shown to be involved in acid stress and bile resistance in other species, respectively. In an in vitro assay, however, only 48% of L. lactis NCDO 2118 was able to grow after pH challenge, and 95% of bacteria was inhibited by bile salts. Other authors have already found that bacteria with an intestinal origin tend to be more resistant to stomach acids [73]. Therefore, this finding corroborates our results because L. lactis NCDO 2118 was isolated from frozen peas. Most of the genes found in L. cremoris MG1363 were also identified in L. lactis NCDO 2118. Additionally, a work using proteomics analyses identified some genes related to acid response and they are present in L. lactis NCDO 2118 genome (clpEP, ahpC, tig, hpr and luxS) [74] showing that other approaches may better elucidate the mechanism of survival to acid stress on this strain.
The high susceptibility of L. lactis NCDO 2118 to bile salts, on the other hand, must be further explored in vitro and in vivo using transcriptomics analyses to determine the expression rates of the described genes.
Competitive exclusion mechanisms of L. lactis NCDO 2118. There are several mechanisms used by bacteria to competitively exclude other species, such as bacteriocin production, space competition through the use of adhesins or receptors that bind to specific surface features, predation and even rapid growth [75].
Adhesins are responsible for the recognition and colonization of host tissues through specific binding. This process may activate the innate host cells or the expression of new genes.
Adhesins may be characterized as hair-like attachments named pili or fimbriae or in other cases, named non-pilus adhesin, related to the microbial cell surface [76].
In L. lactis NCDO 2118, we have identified the gene chiA (NCDO2118_2053) and the genes coding for the Chitin binding protein (CBP-NCDO2118_2054) and the laminin-binding protein (NCDO2118_1446), which are normally related to adhesion in other bacteria. Chitin is degraded by chitinases that belong to members of the glycoside hydrolase of family 18 [77]. One example of bacteria that produces chitinase is Serratia marcescens, one of the most efficient organisms in chitin degradation [78]. When E. coli was cloned with a chitin-binding protein of Serratia marcescens, there was a significant increase in its ability to adhere to human colon cells [77].
Chitin-binding encoding genes are broadly distributed in many microorganisms. The L. lactis IL1403 genome, for example, harbors chitinolytic machinery represented by one family 33 CBP (yucG; referred as LlCBP33A), one family 18 chitinase (chiA, referred as LlChi18A) and one family 20 N-acetylhexosaminidase [3; 79]. Another example of bacteria that present a high adhesion degree is Borrelia burgdorferi, which is able to bind to mammalian laminin, an important extracellular matrix (ECM) component [80]. A laminin-binding protein has also been identified in L. lactis NCDO 2118.
Additionally, we have found using MATS experiments that L. lactis NCDO 2118 presents a 52% of association to xylene, which supports the presence of genes coding for adhesion-related proteins in this strain. The hydrophobicity is directly related to the capacity of strains to adhere to surfaces. This capacity is determined by hydrophobic components present in the outer membrane of microorganisms, and it is known that hydrophobic interactions have an important role in the adhesion of bacteria to the epithelium. The application of MATS experiments facilitates a qualitative assessment of the polarity or non-polarity of the bacterial surface, which is important because it indicates the potential for probiotic adhesion to apolar surfaces in the intestinal and vaginal epithelia. However, this test is only a primary indicator of the adherence of microorganisms [81; 82].
The other bacterial competitive exclusion mechanism assayed here was the production of exclusion antimicrobial peptides, named bacteriocins. Bacteriocins produced by a bacterium may be activated against others, even ones from the same species, while the producer is immune to its own peptides [43]. This exclusion mechanism is very important for probioses, as it renders probiotic organisms able to compete with and kill pathogenic ones, promoting a health benefit to the host [2]. We have predicted one bacteriocin for each of the three classes in L. lactis NCDO 2118 (class I-III), which may be important for exclusion mechanisms of this bacteria. However, the lack of nisT and the pseudogenization of nisP on the class I gene cluster, the lack of ABCtransporters in the class II cluster and, also, the lack of information regarding the product of the putative bacteriocin in the class III cluster have to be further studied using in vitro analyses to elucidate whether those bacteriocins are produced and present antimicrobial activity or not.
We have also performed a deferred agar spot assay for the initial determination of antagonistic activity produced by L. lactis NCDO 2118. This test indicates the activity against various Gram-positive and -negative bacteria. This inhibitory effect may be due to H 2 O 2 , lactic acid, bacteriocins, antibiotic-like substances, or a combination of these compounds [83]. However, L. lactis NCDO 2118 showed no effect on the growth of the pathogenic strains assayed here.
Secreted proteins and immunomodulatory effects. According to Luerce et al., (2014), the secreted proteins of L. lactis NCDO 2118 are possibly responsible for the immunomodulatory effects of this transient bacterium inside the host. In a comparison of the anti-inflammatory effects between L. lactis NCDO 2118, L. lactis IL1403 and L. cremoris MG1363 strains, only the L. lactis NCDO 2118 supernatant was able to decrease the IL-8 production (45%), showing its immunomodulatory ability against inflammation [18].
Here, we predicted 5 proteins that are present in the 26 secreted proteins exclusive from L. lactis NCDO 2118 and in the 867 expressed proteins from proteomic analyses and may thus be related to the probiotic effect of this strain (Table 7). From those 5 exclusive, secreted and expressed genes of L. lactis NCDO 2118, epsK and epsL are part of the operon epsABCDEF-GHIJKLX, whereas there is an epsR gene located in another genomic region.
The EPSs are a type of biopolymer able to facilitate intense interactions of biofilm cells through adhesion, aggregation of bacterial cells, cohesion of biofilms, protective barriers, and cell component export [84]. Through microarray and electron microscopy analyses, Denou et al., 2008 found an eps cluster of genes exclusive from a probiotic Lactobacillus strain compared to a type strain and they have shown that deletion of this cluster from the probiotic strain results in lack of the fuzzy layer on the outside of the cell wall [85].
Altogether, the lack of further knowledge of the eps cluster of genes and the presence of three other genes coding hypothetical exclusive/secreted/expressed proteins highlight the need for additional studies to better elucidate the underlying mechanisms involved in the antiinflammatory and immunomodulatory activities of this strain.

Genome sequences
The genome sequences of L. lactis NCDO 2118 [24] and 15 other strains of Lactococcus were retrieved from the GENBANK dataset of NCBI (Table 1). Briefly, the dataset is composed of 8 strains of Lactococcus lactis subsp. lactis, 2 of which were isolated from legumes (L. lactis NCDO 2118 and Lactococcus lactis subsp. lactis KF147), 6 Lactococcus lactis subsp. cremoris isolated from dairy or other fermented foods, and 2 Lactococcus garvieae isolated from diseased fish. L. garvieae was added to the analyses because it is a closely related pathogenic species. S. thermophilus LMD-9 was used as an outgroup to root the phylogenetic tree. Only complete genomes were used to avoid bias.

In silico analyses
Heatmap of genome similarities and 16S phylogenetic tree. The heatmap analyses of the 17 strains were performed with Gegenees [86]. The input files consisted of complete genomes in.fna format. Streptococcus thermophilus LMD-9, a closely related species, was used as an outgroup to root the tree. The analyses were performed with default parameters for comparative analyses using the alignment method BLASTn. Gegenees performs an all-versus-all alignment process of the fragments generated from the 17 genomes. The result was exported from Gegenees as a heatplot image. Additionally, a phylogenetic tree was made using the 16S sequences from all genomes as identified by RNAmmer [87]. After that, they were aligned in MUSCLE [88], and the phylogenetic tree was inferred using the Neighbor-Joining method with 1000 bootstrap replicates.
Genome synteny. The genome synteny analyses were performed using Mauve, with the "progressiveMauve" option and all genome sequences in the.fna format. Mauve predicts gene synteny by merging locally collinear blocks of conserved genome orthologous regions and ordering them according to a reference genome [89].
Genome plasticity. The genome plasticity analyses were performed by searching for horizontally acquired regions such as genomic islands and phage sequences. The genomic islands were searched using the software GIPSy: Genomic Island Prediction Software [90], which updates the methodology of the software PIPS: Pathogenicity Island Prediction Software. Briefly, GIPSy performs the prediction of four different classes of genomic islands: Pathogenicity Islands, Resistance Islands, Metabolic Islands and Symbiotic Islands. In this work, we searched for metabolic and symbiotic islands in the genome of L. lactis NCDO 2118 using Lactococcus lactis subsp. cremoris MG1363 and Lactococcus garviae Lg2 genomes as subjects. After, we consolidated and manually curated the results. The choice of metabolic and symbiotic islands was made based on the lifestyle of L. lactis NCDO 2118, a strain isolated from vegetables, and its metabolic importance.
All the analyses were performed using GENBANK files and default parameters. The results were exported in tabulated format and used in BRIG (Blast Ring Image Generator) to generate circular genome comparative views [91]. Finally, the prophage prediction was performed using the GENBANK file and the software Phast [92], and the results were exported in table format and used as input in BRIG.
Bacteriocin prediction. The bacteriocin prediction was performed in BAGEL software using the.fna file from L. lactis NCDO 2118. Briefly, the software works with a curated dataset of bacteriocins, in which the input data are evaluated based on a Hidden Markov Model. The genetic information is analyzed based on combinations of PFAM domains [42]. For the putative bacteriocin predicted on L. lactis NCDO 2118 (NCDO2118_1768), we used the Transporter Classification Database (TCDB) [93] with an e-value of e-07.
Circular comparison map of genomic sequences. To create circular genome comparisons, we used the software BRIG and all genome sequences in the.fna format; we created the figure with L. lactis NCDO 2118 as reference strain. Additionally, we added the coordinates of the genomic islands and phage regions to the figure to visualize genome plasticity events. Finally, all genomes underwent BLAST analyses against the reference strain to create the circular comparison map.
Metabolic pathway prediction. A genome sequence in.fasta and a genome annotation in the.gbk format were used for reconstructing the Lactococcus species metabolic pathways. Posteriorly, the Pathway/Genome Databases (PGDB) for each of the 16 strains were computationally predicted using Pathway Tools software version 16.5 [94], developed by SRI International. The MetaCyc, a highly curated and non-redundant reference database of small-molecule metabolism, was used as a reference database for the PathoLogic component of the Pathway Tools software [95]. The metabolic pathways of L. lactis NCDO 2118 were used as a reference for the comparative analysis using the following comparisons: i) L. lactis NCDO 2118, L. lactis KF147 and Lactococcus lactis subsp. lactis IL1403, ii) non-pathogenic strains of L. lactis (L. lactis subsp. lactis and cremoris), and iii) all strains in this study.
Identification of the secretome. The prediction of the putative subcellular localizations of L. lactis NCDO 2118 proteins was performed in silico using SurfG+. This software contains such tools as SignalP, LipoP and TMHMM for the identification of motifs [46]. Interestingly, SurfG+ uses the size of the membrane wall to better differentiate the membrane (MEM) and potentially surface exposed (PSE) proteins. Here, the measurements of the membrane wall were performed with electron microscopy with EM10A equipment (Zeiss), as previously described [96].
L. lactis NCDO 2118 was grown at 30˚C for 18 h in M17 medium (Difco) containing 0.5% glucose [18] and then centrifuged. The resulting precipitate (~500 mL) was placed in an Eppendorf tube, fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.2) for 6 h at 8˚C and washed three times with 0.1 M sodium cacodylate buffer (pH 7.2). After washing, the sample was post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (pH 7.2) + 1.5% potassium ferrocyanide for 90 minutes, washed with 0.1 M with sodium cacodylate buffer (pH 7.2), dehydrated in a graduated ethanol series (50% EtOH, 70% EtOH, 95% EtOH, and 100% EtOH), and incorporated in Eponate-Araldite resin. Ultrathin sections were obtained using uranyl acetate and lead citrate and then verified by Zeiss-EM-10A [97]. The micrograph was obtained by one CCD Mega View camera. The thickness of the L. lactis NCDO 2118 wall was determined from the image analysis micrograph in ImageJ software (available at imagej.nih.gov/ij/).
To measure the wall, we used at least five micrographs of L. lactis NCDO 2118 with magnifications of 50,000 and 100,000 times. We calculated the mean size of the cell walls, and the average number of amino acids for the obtained wall thickness was~55 amino acids. This value was added to the SurfG+ software together with the.fasta sequence of amino acids (.faa) exported from the strain of interest.
After this process, we used OrthoMCL tool to predict the orthologous and paralogous genes between L. lactis NCDO 2118 and L. lactis IL1403.
temperature. The tube was agitated vigorously, and after 30 minutes, the aqueous phase was collected for optical density OD 600nm measurement. The reduction percentage of optical density was calculated. The results were based on the average of three independent assays.
Antagonistic activity. Bacterial isolates were cultured in MRS broth for 24 h at 37˚C within an anaerobic chamber. A 5 μL aliquot of the culture was then spotted onto MRS agar. After incubation at 37˚C for 48 h under anaerobic conditions, the cells were killed by exposure to chloroform for 20 min. Residual chloroform was allowed to evaporate, and Petri dishes were overlaid with 3.5 mL of a soft agar containing brain heart infusion (Acumedia, Neogen Co., Lansing, MI, USA), tryptone soy broth (Difco) supplemented with 0.5% yeast extract Bacterial strain, growth conditions and preparation of proteins from culture filtrates for proteomic analysis. L. lactis NCDO 2118 and L. lactis IL1403 were pre-inoculated in M17 medium (Difco, New Jersey, USA) and incubated at 30˚C for 16 h. The precultures were then inoculated (1:100) in fresh M17 medium supplemented with 0.5% (w/v) glucose (M17Glc) at 30˚C until reaching an OD 600 = 0.8 (three independent experiments). The cultures were then centrifuged for 20 min at 2,700 x g. The supernatants were filtered using 0.22-μm filters, 30% (w/v) ammonium sulfate was added to the samples, and the pH of the mixtures was adjusted to 4.0. Next, 20 mL of N-butanol was added to each sample. The samples were centrifuged for 10 min at 1,350 x g and 4˚C. The interfacial precipitate was collected and resuspended in 1 mL of 20 mM Tris-HCl pH 7. 2 [102]. To perform label-free proteomic analysis, the protein extract was concentrated using a spin column with a 10 kDa threshold (Millipore, Billerica, MA, USA). The protein was denatured (0.1% RapiGEST SF at 60˚C for 15 min) (Waters, Milford, CA, USA), reduced (10 mM DTT), alkylated (10 mM iodoacetamide) and enzymatically digested with trypsin (Promega, Sequencing Grade Modified Trypsin, Madison, WI, USA).
Following the identification of proteins, the quantitative data were packaged using dedicated algorithms [104; 105] and searching against a database with default parameters to account for ions [106]. The databases used were reversed "on-the fly" during the database queries and appended to the original database to assess the false positive rate during identification. For proper spectra processing and database searching conditions, the ProteinLynxGlobalServer v.2.5.2 (PLGS) with Identity E and Expression E informatics v.2.5.2 (Waters) was used. Uni-ProtKB with manually reviewed annotations was used, and the search conditions were based on taxonomy (L. lactis). The maximum allowed missed cleavages by trypsin were up to one, variable modifications by carbamidomethyl (C), acetyl N-terminal, phosphoryl (STY) and oxidation (M) were allowed, and a peptide mass tolerance value of 10 ppm was used [107]. The collected proteins were organized by the PLGS Expression E tool algorithm into a statistically significant list that corresponded to higher or lower regulation ratios among the different groups. For protein quantification, the PLGS v2.5.2 software was used with the Identity E algorithm using the Hi3 methodology. The search threshold to accept each spectrum was the default value in the program with a false positive value of 4%. The quantitative values were averaged over all samples, and the standard deviations at p < 0.05 were determined using the Expression software [107].

Conclusions
Although L. lactis NCDO 2118 presented a high similarity to the other L. lactis strains, it presents an SI that is commonly shared with L. lactis KF147, along with high genomic synteny conservation with this strain. Additionally, the antibiotic resistance of this strain to vancomycin, amikacin and oxacillin could be an obstacle for its use as a probiotic. However, the absence of resistance-related genes in regions acquired by HGT and the absence of RIs in the genome sequence corroborates its safety aspects and supports its use as a probiotic strain. Moreover, the high susceptibility of L. lactis NCDO 2118 to acid and bile salts stresses have to be further evaluated in a complete digestion simulation, using transcriptomics and proteomics analyses, to elucidate whether the identified genes are differentially expressed in those environmental conditions. Interestingly, the adhesion of L. lactis NCDO 2118 to xylene and the putative production of three classes of bacteriocins are important indicators of the exclusion mechanisms used by this strain. However, the in vitro analyses have not shown any sign of an antagonistic effect against the assayed pathogenic bacteria. Future works could also take advantage of combined transcriptomics and proteomics analyses of L. lactis NCDO 2118 in vitro before and after intestinal passage to evaluate the expression of the identified genes. Additionally, the identification of the EPS cluster of genes putatively associated with the probiotic effect of L. lactis NCDO 2118 could be further explored in 16S metagenomics analyses of gut microbiota, after expression, purification and administration of EPS proteins. Finally, through the analyses of the safety, survival and probiotic aspects of L. lactis NCDO 2118, we highlight here the potential use of this strain as a target for the future development of probiotic foods. represented according to the nucleotide conservation and synteny. Low similarity regions are represented as white regions inside the blocks, highlighted by a red ( Ã ). Regions of deletions are represented as blank spaces between the blocks, letter (A). Insertion regions are highlighted with the letter (B), and inversion regions are represented by the letter (C). To perform the genome synteny analysis, we used the software Mauve, which compares the genomes by identifying and clustering homologous genes between the genomes into large collinear blocks of genes [89]. The most conserved genome compared to L. lactis NCDO 2118 was L. lactis KF147. Between these two strains, it is possible to see some regions of: deletion; insertion; inversion and specific areas with low or no similarity with the reference genome. The comparison of those features with other strains shows: a deletion on the genome position 1,200,000 of Lactococcus lactis subsp. lactis IO-1; a big inversion region in Lactococcus lactis subsp. lactis AI06 in the range from 800,000 to 1,600,000; a small insertion near the genome position 200,000 of L. lactis KLDS 40325 (in green); and a block on Lactococcus lactis subsp. lactis S0 (2,000,000 position) with low similarity to the reference genome.