Background

Leuconostoc gelidum is a psychrotrophic lactic acid bacterium (LAB) detected by Shaw and Harding [1] in cold-stored, vacuum-packaged beef. In 2014, Leuconostoc gasicomitatum [2] was reclassified as Leuconostoc gelidum subsp. gasicomitatum [3] and another subspecies, Leuconostoc gelidum subsp. aenigmaticum was described. Reclassification of L. gasicomitatum was mainly based on the results from the phylogenetic analyses of atpA, pheS, and rpoA housekeeping genes. The sequence analyses of concatenated atpA, pheS, and rpoA genes had demonstrated that the novel strains, designated as subsp. aenigmaticum, as well as the type and reference strains of L. gelidum and L. gasicomitatum, were phylogenetically closely related. Until recently, L. gelidum was thus considered to comprise three subspecies i.e., subsp. gelidum, gasicomitatum and aenigmaticum, but based on the ANI and dDDH values of L. gelidum type strains, Wu and Gu [4] proposed to reject the proposal of Rahkila et al. [3], and to restore the taxonomic position of L. gelidum subsp. gasicomitatum back to the species level as L. gasicomitatum.

L. gelidum and L. gasicomitatum have commonly been associated with the spoilage of perishable food items, such as meat and poultry or minimally processed vegetables [5,6,7]. Metagenomic analyses have detected these species mainly in cold-stored meat and vegetables [8], but rarely in samples from animal microbiomes or environmental sources. Apart from food spoilage, L. gelidum and L. gasicomitatum have often been associated with the fermentation of kimchi [9, 10], a Korean traditional side dish made most commonly of Chinese cabbage and Korean radishes with a variety of seasonings. Kimchi is often fermented at low temperature (2–6 °C) to restrict the growth of spoilage bacteria, but psychrotrophic LAB, L. gelidum and L. gasicomitatum can grow at temperatures below 6 °C [1, 3, 6]. Therefore, they may become abundant in kimchi microbiomes [9,10,11]. Thus, the growth of L. gelidum in refrigerated foods may either be unwanted and lead to spoilage or be endorsed to ensure proper ripening and preservation according to the expectations of kimchi fermentation.

The significance of L. gelidum and L. gasicomitatum in food, prompted our interest to further studying these species at the genome level. Despite the importance of these species in the food system, relatively little is known about them at the genome level, and at the time of this study, only few genomes were available in the public databases. Until now, most studies have applied analysis of molecular fingerprints [6, 7] or multi-locus sequence typing [12] to examine these species. In addition, the recent taxonomic proposal warranted further genotaxonomic studies performed with a larger set of strains than just the type strains that were used by Wu and Gu [4]. Since we have worked with these species during the course of two decades, our strain collection allowed us to select a diverse set of well-characterised strains in terms of sources, previously conducted genotyping results and geographical distribution (Supplementary Table S1) for the present study.

Results

Genome features reveal little intraspecies variation

The 37 L. gelidum/L. gasicomitatum strains (our L. gelium subsp. gasicomitatum strains were renominated as L. gasicomitatum in this paper according to the proposal of Wu and Gu [4]) had an average genome size of 1.95 Mbp (range 1.82–2.12 Mbp) with GC content ranging from 36.4 to 36.8% (Table 1, Supplementary Table S1). The number of contigs ranged from 6 to 49 (av. 24 ± 12; Table 1) with 28 of 37 (75%) of the genomes that were assembled to a maximum of 32 contigs. The N50 values ranged from 102 kbp to 1.45 Mbp (Table 1), with an average of 354 kbp (Supplementary Table S2). On average, the genomes encoded 1,842 predicted proteins (range 1,704–2,034). The number of pseudogenes was estimated to range from 33 to 67 (43 ± 7.6). Furthermore, the number of transfer RNA (tRNA) genes varied from 40 to 47 with an average of 44 tRNA genes per assembled genome. (Supplementary Table S2).

Table 1 Leuconostoc strains (n = 37) sequenced and genomic features of the draft genomes
Full size table

Genomic relatedness indices suggest species status for L. gelidum subsp. gasicomitatum

Pairwise average nucleotide identity (ANI) values were used to assess inter- and intra-subspecies relatedness. The values for L. gelidum subsp. aenigmaticum assemblies against L. gelidum subsp. gelidum assemblies (Table 2;Supplementary Table S4) were above 95%, the cut-off value for species delineation [13], which supports their classification to the same species. In contrast, L. gasicomitatum and L. gelidum subsp. gelidum DSM 5578 T type assembly or L. gelidum subsp. aenigmaticum DSM 19375 T type assembly presented ANI ≤ 94.8% (Table 2; Supplementary Table S4), and were, thus, lower than the proposed species cut-off values for ANI. Similarly, the average values for L. gasicomitatum against to the assemblies of subsp. gelidum and subsp. aenigmaticum were below the cut-off of 95% (Table 2); although 5% (21 of 416) of the pairwise comparisons were between 95.0–95.2% (Supplementary Table S4). The ANI values obtained for all the intra-subspecies comparisons between L. gelidum assemblies were high, i.e. ≥ 98% (Table 2, Supplementary Table S4).

Table 2 Pairwise average nucleotide identity (ANI) and dDDH digital DNA-DNA hybridisation (dDDH) values determined
Full size table

In addition to ANI, we assessed the genome relatedness by calculating the pairwise digital DNA-DNA hybridisation (dDDH) scores for the 47 genomes (Table 2, Supplementary Table S4). For dDDH, similarity values ≥ 70% are considered an argument to classify the strains to the same species [14]. Similarly, with ANI metrics, L. gasicomitatum assemblies showed relatively low dDDH values against L. gelidum subsp. gelidum (57.7 ± 1.7%) and L. gelidum subsp. aenigmaticum (59.2 ± 0.5%) assemblies. The inter-subspecies values obtained for L. gelidum subsp. aenigmaticum against L. gelidum subsp. gelidum genomes varied more. They ranged from 67.1 to 87.4 (71.1 ± 4.6%) with the values obtained for the L. gelidum subsp. gelidum DSM 5578T type assembly (69.0–72.4%) thus being also around the species cut-off of 70% (Table S4). However, the dDDH similarity was 70.6% between the type assemblies of these subspecies (DSM 5578T × DSM 19275T) (Supplementary Table S4).

Phylogenetic analyses classified subsp. gasicomitatum separately from subs. gelidum

To assess the evolutionary relationships between the strains belonging to the former three L. gelidum subspecies, we constructed a maximum likelihood phylogenetic tree (Fig. 1) from concatenated nucleotide sequences of three housekeeping genes atpA, pheS and rpoA used in the previous taxonomy studies to distinguish Leuconostoc species [3, 15]. In the tree topology (Fig. 1), the L. gasicomitatum strains were separated with high bootstrap support from subsp. gelidum and subsp. aenigmaticum strains. For this finding, we obtained even stronger support with a tree constructed using a core gene set of 1,141 single-copy genes (Fig. 2). In both phylogenetic treeing approaches, the three subsp. aenigmaticum strains clustered together and were placed in the same group as the subsp. gelidum strains (Figs. 1 and 2). In addition, we did not notice any clustering based on the geographical locations of the L. gasicomitatum strains. Strains from Belgium (strains coded with R-, HS- and a- or numbers only) and Latvia (Jla4-8) clustered among those isolated in Finland.

Fig. 1
figure 1

Maximum-Likelihood phylogenetic tree for concatenated housekeeping genes atpA, pheS and rpoA nucleotide sequences. Bootstrap support (%) for 1,000 replicates are given at branch points

Full size image
Fig. 2
figure 2

Maximum-Likelihood phylogenetic tree for 1,141 single-copy orthologous genes shared by all genomes. Bootstrap support (%) for 1,000 replicates are given at branch points

Full size image

Pangenome analyses

We estimated the pangenome and core genome sizes separately for L. gasicomitatum and for L. gelidum subsp. gelidum and subsp. aenigmaticum using the model of Tettelin et al. [16] for curve fitting. Gene accumulation curves (Figs. 3 A and B) showed that the number of new genes added by each assembly continued to increase, indicating open pangenomes for both analyses. However, the expansion of the L. gasicomitatum pangenome was decreasing (Fig. 3B), and after 22 genomes fewer than 15 genes were added to the pool per new genome, suggesting that the pangenome was closing slowly. The core genome analyses (Figs. 3C and D) fitted the exponential decay model. The fitted curve for L. gasicomitatum plateaued at around 1300 genes (Fig. 3D), suggesting relatively stable core genomes.

Fig. 3
figure 3

Estimates after 10 random genomes for pangenome (A) and core genome (C) for 15 Leuconostoc gelidum subsp. gelidum and L. gelidum subsp. aenigmaticum genomes, and for pangenome (B) and core genomes (D) for 32 L. gasicomitatum genomes. Curve fitted using the Tettelin model [16]. Residual standard errors reported to show the goodness of fit

Full size image

The pangenome of L. gasicomitatum was estimated to include 3,046 genes of which 1,313 (43.1%) were considered the core genes present in 99% of the assemblies. On average, the core genome consisted of 71% (64–75%) of the coding potential of a gasicomitatum assembly. The open pangenome for L. gelidum subsp. gelidum and aenigmaticum consisted of 2703, gene, and of these 1,341 were included in the core genome. The unique genes detected in L. gelidum/L. gasicomitatum (Supplementary tables S5-S7) did not encode any properties associated with specific lifestyles related to food spoilage. Genes unique to L. gelidum subsp. gelidum and L. gasicomitatum strains were not detected.

Properties associated with lifestyle in food

Regarding to properties typical for leuconostocs, all strains contained genes encoding at least one Glucansucrase and half of them carried also genes encoding a fructansucrase. No genes associated with the formation of biogenic amines through amino acid decarboxylation or deamination were found.

L. gelidum and L. gasicomitatum are known to be able to cause spoilage of food products by producing yellow carotenoids when growing in food that contains fat, e.g. ham or sausage [17, 18]. But the genes responsible for carotenoid production in LAB (including other Leuconostoc species), e.g. crtN and crtM [19], were not detected in L. gelidum/L. gasicomitatum. It is, therefore, not clear to us how L. gelidum and L. gasicomitatum produce carotenoids or what type of carotenoids they produce.

Genetic determinants for fosfomycin and fusidic acid resistance

Next, we analysed the acquired antimicrobial resistance (AMR) determinants by comparing the L. gelidum/ L. gasicomitatum assemblies with the Comprehensive Antibiotic Resistance Database (CARD). The searches did not return any hits with good support (using “Perfect, Strict, complete genes only” as criteria) suggesting that these food-associated L. gelidum/L. gasicomitatum strains do not carry acquired or transferable antimicrobial resistance genes or allelic variants that cause antimicrobial resistance.

Regarding intrinsic resistance, the genus Leuconostoc is known to be resistant to vancomycin and the resistance mechanism is well-characterised [20]. Consistently, all 37 newly sequenced L. gelidum genomes harboured a resistant genotype of ddl gene. In addition to vancomycin, resistance to fosfomycin is common among leuconostocs [21, 22], and so we analysed the genomes for genes associated with fosfomycin resistance [23]. Among the alignments, the only known resistance determinant we identified was an amino acid substation in the fosfomycin target gene murA. A multiple protein sequence alignment of the MurA demonstrated (Supplementary Fig. 1) that all 47 L. gelidum strains had Asp substitution at position 115 (E. coli numbering) at the site encoding the active site of the MurA, a substitution known to confer fosfomycin resistance in E. coli [24].

Furthermore, we assessed the known genetic determinants behind fusidic acid resistance, another resistance characteristics of leuconostocs [25]. Sequence analysis of fusA (Supplementary Fig. 2), a chromosomal gene encoding elongation factor G (EF-G), the target protein of fusidic acid, revealed alterations that cause resistance-associated amino acid substitution in EF-G [26]. Comparison with Staphylococcus aureus showed (Supplementary Fig. 2) that L. gelidum/L. gasicomitatum genomes harboured an amino acid substitution V90I, His457Q, Leu461M and S416T (S. aureus numbering) of which at least V90I and His457Q result in high fusidic acid resistance in S. aureus [26].

Discussion

The genomic features of the 37 L. gelidum/L. gasicomitatum strains sequenced in the present study showed only little variation (Table 1; Supplementary Table S1), even though we selected as heterogenous a set of strains as possible from our database. We used MLST, fingerprinting and source attribution data accumulated over the past 25 years since 1995 (Table 1; Supplementary Table S1) to select the strains for the present study. The genome sizes (1.95 ± 0.06 Mbp), GC contents (36.6 ± 0.1%), numbers of CDS (1842 ± 72) and tRNA genes (44 ± 2) were quite similar (Table 1; Supplementary Table S2) and alike with the features present in the genomes of the 10 L. gelidum/L. gasicomitatum strains (Table 1; Supplementary Table S3) available in the public NCBI database.

The genetic relatedness indices (Table 2; Supplementary Table S4) as well as the phylogenetic analyses (Figs. 1 and 2) revealed that the strains formerly designated as L. gelidum subsp. gasicomitatum formed a genetically coherent group that was evolutionary different from the group containing the strains of the two other L. gelidum subspecies. According to the ANI and dDDH results (Table 2 and Supplementary Table 4), the taxonomic position of subsp. gasicomitatum within L. gelidum is not supported. Based on the ANI and dDDH results between the type strains of L. gelidum subspecies, Gu and Wu [4] have already suggested rejecting the proposal of Rahkila et al. [3] to reclassifiy L. gasicomitatum as L. gelidum subsp. gasicomitatum. Our study conducted with these 37 strains supports their findings and thus, the strains previously considered as subsp. gasicomitatum should be designated as L. gasicomitatum.

Rahkila et al. [3] justified the splitting of L. gelidum between three subspecies mainly because of the results obtained in the phylogenetic and optical DDH analyses. At that time, the phylogenic analyses were targeted at three housekeeping genes atpA, pheS and rpoA that had been reported to distinguish LAB species well [15]. Rahkila et al. [3] used 13 strains consisting of 6, 3 and 4 representatives of L. gelidum subsp. aenigmaticum, L. gelidum subsp. gelidum and L. gasicomitatum, respectively. Since the analysis of the concatenated atpA, pheS and rpoA sequences divided these 13 strains between three separate clusters, they subsequently conducted the optical DDH experiments using representative strains from each three clusters. In these DDH experiments, Rahkila et al. [3], used, among others, the type strains LMG18297T and LMG18811T of L. gelidum and L. gasicomitatum, respectively, and POUF4d, which is the current type strain (DSM 19375T) of subsp. aenigmaticum. The DDH result of 75% between L. gelidum and L. gasicomitatum type strains that they reported contravened the values previously published in two other studies. Björkroth et al. [2] and Kim et al. [27] had reported DDH values below 70%, i.e., 22 and 6%, respectively. In addition, Björkroth et al. (2000) had reported another value below 70%, i.e., 34%, between L. gasicomitatum LMG 18812 and L. gelidum subsp. gelidum LMG 9850T type stain. Thus, the results of the present study are in accordance with the conclusions of Björkroth et al. (2000), Kim et al. (2000) and Gu and Wu (2021), supporting that the taxonomic position of L. gelidum subsp. gasicomitatum should be restored as L. gasicomitatum.

Our analyses also show that subsp. aenigmaticum has been correctly classified as a subspecies of L. gelidum. During the past 25 years, we have isolated and constructed a comprehensive culture collection containing hundreds of L. gelidum and L. gasicomitatum food isolates. However, we have not detected more subsp. aenigmaticum strains than those used in the study of Rahkila et al. [3]. To our knowledge, only one additional scientific study of this subspecies has been published thus far. Mun et al. (2021) found L. gelidum subsp. aenigmaticum LS4 strain to improve the organoleptic qualities of kimchi juice and suggested that LS4 could be used as a functional starter culture for food (vegetable or fruit) fermentation at low temperatures. The genome and functional properties of LS4 are very similar to those detected in this study and by Rahkila et al. [3]. The reason for the low detection rate of subsp. aenigmaticum is not known to us, but the genetic differences between subsp. aenigmaticum and gelidum, and even between L. gasicomitatum, do not explain any specific traits to suggest reasons for the rare detection (Supplementary Table S5).

Figures 2 and 3 present phylogenetic trees generated with the concatenated atpA, pheS and rpoA and WG sequences, respectively. Both approaches resulted in trees containing three separate clusters with either L. gelidum subsp. gelidum, L. gelidum subsp. aenigmaticum or L. gasicomitatum strains. In both trees L. gelidum subsp. gelidum and L. gelidum subsp. aenigmaticum strains clustered adjacently, whereas L. gasicomitatum strains were located more distantly from the other two clusters. This shows that L. gelidum and L. gasicomitatum differ clearly from each other also phylogenically. No clustering to suggest any attribution to specific isolation sources was found as in the previous L. gasicomitatum MLST study by Rahkila et al. [12].

Antimicrobial resistance in Leuconostoc is of interest to the researcher community since leuconoctocs are present in various foods [6, 7, 11, 12, 22] and antimicrobial resistance, particularly, transferable or acquired resistance-related determinants, is considered a food safety concern [28]. Antimicrobial resistance properties can also be used to develop tools for a counterselection marker for recombination [29] or to design differentiating or selecting growth media for LAB [25]. Leuconostocs are Generally Regarded As Safe (GRAS, US Food and Drug Administration); or some have the Qualified Presumption of Safety by the European Food Safety Authority. However, mainly because of their intrinsic antimicrobial vancomycin resistance, they have also been associated with infections in patients receiving vancomycin. According to our analyses of the 47 L. gelidum/L. gasicomitatum genomes, these strains did not carry any acquired antimicrobial resistance genes. In addition to intrinsic vancomycin resistance, leuconostocs are intrinsically resistant to teicoplanin, fosfomycin and fusidic acid [16, 17, 26]. The vancomycin/teicoplanin resistance mechanism is known and it results from the use of D-lactate instead of D-ala in the synthesis of peptidoglycan [20]. Based on our findings, we consider that the intrinsic fosfomycin resistance in Leuconostoc is due to the Asp in MurA position 115 (E. coli numbering) known to confer resistance in E. coli [24]. In addition to leuconostocs, Fructobacillus, Weissella, Oenococcus, Pediococcus and lactobacilli appear to carry the resistant genotype, whereas other LAB genera have the sensitive genotype. Our findings related to fusidic acid resistance suggests that it results from point mutations within the fusA gene [26]. Leuconostoc, Fructobacillus and Convivina have the resistant genotype associated with a Q at the position 457 (Staphylococcus numbering), whereas Weissella and Oenococcus have the sensitive genotype related to a H at position 457. Recently [30], IsaA gene similar to a clindamycin resistant phenotype was detected in Leuconostoc fallax ATCC 700006T and Leuconostoc pseudomesenteroides NCDO 768T. However, we did not detect this gene in the 47 L. gelidum/ L. gasicomitatum genomes analysed.

A form of spoilage manifested by the formation of slime has been associated with L. gelidum/L. gasicomitatum [5]. Since all strains contained genes encoding at least one glucansucrase and half of them carried also genes encoding a fructansucrase, it can be concluded that these species have a general potential to form dextran if sucrose is available as the precursor and favourable growth condition exists. On the other hand, we did not find clear evidence of ability of forming biogenic amines (BAs) that result from decarboxylation or iminase reactions related to amino acids. BAs are unwanted metabolites in foods since they are associated with health hazards and a malodour [31]. Decarboxylation of histidine as histamine leads to a food poisoning resembling a severe allergic reaction after consumption of spoiled histidine-rich food such as tuna and some fish belonging to Scromboidaceae. Tyramine produced from tyrosine is known to trigger migraine, whereas the polyamines putrescine, cadaverine and agmatine are considered as indicators of food spoilage. We searched the genomes but did not detect genes related to agmatine demininase or histidine or tyrosine decarboxylases. We detected a gene that is annotated as an orn/lys/arg-decaboxylase by some annotations (not NCBI) in strains AMKR21, C220d, HS9, JPBL22, Kg1-2 & Vvan9). However, this gene is only half of the length of the characterised orn/lys/arg-decaboxylases. Thus, it is more likely an unknown aminotransferase. Based on the genomic data, these leuconostocs are unlikely to produce any BAs. The ability of leuconostocs to form BAs has been reported time to time but it is probably related to the former classification of “leuconostocs” that are currently considered as Oenococcus, Weissella and Fructobacillus. Due to the many changes in the taxonomy of leuconostocs, one should be cautious while reading older literature and not to mix properties of these genera in these with the current leuconostocs.

Conclusions

Genome features of 47 strains belonging to L. gasicomitatum and L. gelidum subspecies aenigmaticum and gelidum showed only limited variation even though the strains selected for this study originated from different sources, represented variable MLST and genotyping groups, and were isolated in four different countries over a time span of more than 25 years. The genetic relatedness indices as well as the phylogenetic analyses revealed that the strains formerly designated as L. gelidum subsp. gasicomitatum are not a subspecies of L. gelidum. Thus, the proposal of Rahkila et al. [3] to reclassify L. gasicomitatum as L. gelidum subsp. gasicomitatum should be rejected and the species status for L. gasicomitatum restored as also recently suggested by Wu and Gu [4]. Analyses of the genomes did not reveal any source attribution or other clustering associated with the lifestyle of the strains. The pangenome of L. gasicomitatum was estimated to include 3,046 genes of which were 1,313 (43.1%) were considered the core genes present in 99% of the assemblies. After 22 genomes, less than 15 genes were added to the pool per new genome, suggesting that this pangenome was closing slowly, whereas L. gelidum pangenomes (2,703 genes) remained open. According to our analyses on the 47 L. gelidum/ L. gasicomitatum genomes, these strains did not carry any acquired antimicrobial resistance genes or genes associated with production of harmful biogenic amines.

Methods

Selection and sequencing of the strains

Thirty-seven L. gelidum strains isolated during previous research activities from various foods and food-related samples (Table 1) were selected for genome sequencing. The source and county of origin are listed in Table 1.

For sequencing, DNA was extracted from cells collected from MRS broth cultures (25 °C) as described earlier [3]. DNA purity was confirmed using a NanoDrop 2000™ spectrophotometer and quantified using a Qubit® 3.0 fluorometer (Thermo Scientific, USA). Library preparation and sequencing were carried out at the Institute for Molecular Medicine Finland (FIMM), HiLIFE, University of Helsinki, Finland. First, a paired-end DNA library for Illumina sequencing was prepared and normalised with ~ 300 bp inserts using a Nextera XT DNA library Preparation Kit (Illumina, CA, USA). The prepared library was sequenced using an Illumina HiSeq 2500 platform for 100 bp paired-end reads.

In addition to 37 sequenced strains, three complete and seven draft genomes of L. gelidum/L. gasicomitatum, including the type assemblies of each subspecies, were retrieved (Supplementary Table S2). Among them are four L. gasicomitatum strains sequenced by Andreevskaya et al. [32] that at the time of writing were incorrectly classified as Leuconostoc inhae by the European Nucleotide Archive (ENA). Furthermore, the genome sequence for Leuconostoc kimchii IMSNU 11154 T (BioProject PRJNA40837) to was used as a root in the phylogenetic analyses. All software were run with default parameters, unless otherwise stated.

Genome assembly

Following sequencing, the quality control of the raw sequence reads was conducted with FastQC v.0.11.9 available at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc/, the adaptor removal and trimming were performed with Trimmomatic v. 0.40 [33]. The draft genomes were assembled with Velvet v.1.2.08 [34] together with VelvetOptimiser v.2.2.6 (https://github.com/tseemann/VelvetOptimiser) and their quality for completeness and contamination was evaluated using CheckM v.1.0.10 [35]. The assembled genomes were annotated using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) [36].

The accession numbers and details of the genomes used in this study are provided in Supplementary Tables S1 and S2.

To obtain genome-to-genome relatedness indices, Pairwise average nucleotide identity (ANI) values based on BLAST were determined using pyani.py v.0.2.7 (https://github.com/widdowquinn/pyani) and digital DNA–DNA hybridisation (dDDH) with the genome-to-genome distance calculator 2.1 (GGDC) server (available at: http://ggdc.dsmz.de/ggdc.php) using formula 2 recommended for draft genomes. The species boundary for ANI and dDDH values were set for/at? 95 ~ 96 and 70%, respectively.

Pangenome and core genome

The pangenome and core genome sizes were estimated using BPGA v.1.3.0 [37].

Phylogenetic analysis

The nucleotide sequences of all 1,141 single-copy orthologous genes shared by all 47 of the genomes were retrieved by homologous searches using GET_HOMOLOGUES v.3.2.4 [38] and aligned using Prank v.140603 [39] with the “-codon” option for codon aware alignment. Gblocks v.0.91b [40] was used to select conserved regions for the construction of concatenated nucleotide sequences. An approximative Maximum-Likelihood phylogenetic tree was built using FastTree v.2.1.11 [41], with the generalised time-reversible (GTR) model. Maximum-Likelihood phylogenetic trees were constructed for concatenated housekeeping genes atpA, pheS and rpoA using IQ-TREE v.1.6.12 [42] with the TPM2u + F + I and the HKY + F + G4 nucleotide substitution models, respectively.

Prediction of antibiotic resistance

The Resistance Gene Identifier (RGI) tool of the Comprehensive Antibiotic Resistance Database “CARD” version 3.1.3 [43] was used to predict resistomes by using contigs file with the parameters “Perfect, Strict, complete genes only” as search criteria. Literature information on intrinsic antibiotic resistance in Leuconostoc species was also collected. The protein sequences were aligned using multiple sequence alignment MAFFT v.7.408 [44]. Aligned sequences were visualized with GeneDoc v.2.7 [45].