Multilocus Sequence Typing of Leuconostoc gelidum subsp. gasicomitatum, a Psychrotrophic Lactic Acid Bacterium Causing Spoilage of Packaged Perishable Foods

Abstract

Leuconostoc gelidum subsp. gasicomitatum is a psychrotrophic lactic acid bacterium (LAB) that causes spoilage of a variety of modified-atmosphere-packaged (MAP) cold-stored food products. During the past 10 years, this spoilage organism has been increasingly reported in MAP meat and vegetable products in northern Europe. In the present study, the population structure within 252 L. gelidum subsp. gasicomitatum strains was determined based on a novel multilocus sequence-typing (MLST) scheme employing seven housekeeping genes. These strains had been isolated from meat and vegetable sources over a time span of 15 years, and all 68 previously detected pulsed-field gel electrophoresis (PFGE) genotypes were represented. A total of 46 sequence types (STs) were identified, with a majority of the strains (>60%) belonging to three major STs, which were grouped into three clonal complexes (CCs) and 17 singletons by Global Optimal eBURST (goeBURST). The results by Bayesian analysis of population structure (BAPS) mostly correlated with the grouping by goeBURST. Admixture analysis by BAPS indicated a very low level of exchange of genetic material between the subpopulations. Niche specificity was observed within the subpopulations: CC1 and BAPS cluster 1 consisted mostly of strains from a variety of MAP meats, whereas vegetable strains grouped together with strains from MAP poultry within CC2 and BAPS cluster 2. The MLST scheme presented in this study provides a shareable and continuously growing sequence database enabling global comparison of strains associated with spoilage cases. This will further advance our understanding of the microbial ecology of this industrially important LAB.

INTRODUCTION

Leuconostoc gelidum is a psychrotrophic lactic acid bacterium (LAB) currently comprising three genetically distinct subspecies, Leuconostoc gelidum subsp. gelidum, Leuconostoc gelidum subsp. gasicomitatum, and Leuconostoc gelidum subsp. aenigmaticum, associated with modified-atmosphere-packaged (MAP) nutrient-rich foods (1). L. gelidum subsp. gasicomitatum was first encountered in spoiled MAP tomato-marinated raw broiler meat strips in 1997 (2). The spoiled packages were bulging due to rapid accumulation of CO₂ within 4 to 5 days of manufacture. L. gelidum subsp. gasicomitatum has also been prevalent in Finnish high-oxygen MAP beef and pork products (3, 4).

In MAP meat, L. gelidum subsp. gasicomitatum spoilage is typically characterized by gas formation, buttery or sour off odors, and/or green discoloration of the meat (2, 3). In addition, L. gelidum subsp. gasicomitatum has been shown to form slime and gas in acetic-acid-preserved herring (5) and in cooked vegetable “sausages” packaged under vacuum (6). Recently, the LAB has been associated with similar spoilage changes in cooked, brined eggs (slime formation) and various packaged meat and vegetable products (off odors) in Belgium (7–9).

Our previous study employing pulsed-field gel electrophoresis (PFGE) typing for characterization of 384 L. gelidum subsp. gasicomitatum isolates from various meat and vegetable sources differentiated them into 68 genotypes, allowing identification of the genotypes associated with spoilage of meat and vegetable-based foods (10). L. gelidum subsp. gasicomitatum was found to persist in many types of cold-stored MAP foods from several manufacturers in Finland. The meat-derived PFGE genotypes were generally not specific to the type of meat, product type, or producer, but meat genotypes were generally not associated with the vegetable-based products. This was hypothesized to be attributable to the fact that vegetable-derived cell lineages are not disseminated into the meat-processing chain or, alternatively, that these genotypes lack the ability to develop to high numbers in MAP meats.

We have previously sequenced the genome (accession no. FN822744) of L. gelidum subsp. gasicomitatum LMG 18811^T and annotated the spoilage-associated pathways (11). The aim of the present study was to utilize the genomic sequence to establish a multilocus sequence-typing (MLST) approach for the subspecies. MLST would provide insights into the phylogeny of L. gelidum subsp. gasicomitatum, especially whether cell lineages frequently associated with meat have evolved from a restricted group of ancestral genotypes. In addition, establishment of a sequence-based MLST library would enable global surveillance of this spoilage LAB.

MATERIALS AND METHODS

Selection of isolates and growth conditions.

For this study, 252 L. gelidum subsp. gasicomitatum strains from vegetables and MAP meat sources from 22 producers were selected from our culture collection based on PFGE types and sources to obtain maximal genetic and ecological diversity (see Table S1 in the supplemental material). The strains represented 62 different PFGE types divided into 21 PFGE clusters showing pattern variations from 12 to 20.6%. The strains had been isolated as described by Vihavainen and Björkroth (10) and were grown using Man-Rogosa-Sharpe (MRS) medium (Oxoid, Basingstoke, United Kingdom) and incubated at 25°C under an anaerobic atmosphere (Anaerogen; Oxoid).

DNA isolation, PCR, and sequencing.

DNA from all the strains was extracted as described by Pitcher et al. (12) with modifications by Björkroth and Korkeala (13). Each gene was amplified using the primers listed in Table 1. The PCR cycle for the 16S rRNA gene comprised denaturation at 98°C for 2 min, followed by 30 amplification cycles (94°C for 45 s, the annealing temperature [T_a] [Table 1] for 45 s, and 72°C for 1 min), and final elongation at 72°C for 5 min. Each PCR mixture consisted of 200 μM deoxynucleoside triphosphates (dNTPs), 0.02 IU DyNaZyme II polymerase ml21 (Thermo Fisher Scientific Inc.), and 0.5 mM each primer. The PCR products were purified with the Millipore MultiScreen PCR 96 system. Bidirectional sequencing of the PCR products was done using the same primers that were used for amplification, using an Applied Biosystems BigDye Terminator v3.1 Cycle Sequencing kit and an ABI 3130XL genetic analyzer.

TABLE 1.

Genes and primers used in this study

Gene	Locus taga	Primer	Sequence	Length (bp)	PCR temp (Ta) (°C)
dnaA	LEGAS_0001	dnaA-F-001	GTGTTGGCCTTGGTAAAAC	346	44
		dnaA-R-001	GGCAAATCTGGTTTTTGAATATC
gyrB	LEGAS_0005	gyrB-F-001	GAGAAGATGTGCGTGAAGG	379	51
		gyrB-R-001	GTTCAAAATTTTACCACGAATTGG
ddl	LEGAS_0344	ddl-F-001	CACGATGTATCAAAACGTTC	373	45
		ddl-R-001	ACTACATATTTTGTATTACGAAT
pgm	LEGAS_0668	pgm-F-001	TGTTTGGATTTGAAGAAAG	400	42
		pgm-R-001	TAGAACTTTAATTTTGGTTC
dnaK	LEGAS_0807	dnaK-F-001	CTTGGTGGTGATGACTTTGA	322	50
		dnaK-R-001	GGAATACGAGTTGATCCACC
lepA	LEGAS_0891	lepA-F-001	TCACAAGCTTTAGGATTTGG	382	48
		lepA-R-001	TCATAATCTAATGATGCATACCC
pheS	LEGAS_1446	pheS-F-010	CGCGTGATATGCAAGATAC	364	49
		pheS-R-010	GCTCCTAGTACTTCAATCCA
rpoC	LEGAS_1677	rpoC-F-001	CAAGACGTTATTGTTCGTGA	400	48
		rpoC-R-001	AGGCAAACCTTGTGTAATATC
rpoA	LEGAS_1705	rpoA-F-010	CCAAATATTCATAACATCGAAGA	362	45
		rpoA-R-010	AATGCCGTGATTGTCATAT
atpA	LEGAS_1766	atpA-F-010	CAAAAAGACCAAGACATGAT	401	46
		atpA-R-010	CGTTGGAATATATGCTGAAA

^aLocation of each gene in the genome of L. gelidum subsp. gasicomitatum LMG 18811 (accession no. FN822744) (11).

DNA sequence analyses.

Multiple-sequence alignments were performed using MAFFT (14). The dN/dS ratio (the ratio of mean nonsynonymous to synonymous substitutions), the π value (nucleotide diversity), Tajima's D values, and the minimum number of recombination events (R_m) were calculated with DnaSp v5.1 (15).

Population structure.

The Global Optimal eBURST (goeBURST) algorithm (16), as implemented in PHYLOViZ (17), was used to cluster the 46 sequence types (STs) into clonal complexes (CCs) based on their allelic profiles (with respect to their numbers of single-locus variants [SLVs], double-locus variants [DLVs], or triple-locus variants [TLVs]) and to further infer a hypothetical pattern of evolutionary descent by constructing a minimum spanning tree (MST) from the eBURST data. Isolate-specific metadata, including the isolation source, producer, PFGE cluster, and Bayesian analysis of population structure (BAPS) cluster, was then overlaid on top of the MST. Allele numbers and sequence types were assigned manually.

To analyze the population structure, BAPS (18) linkage clustering and the corresponding admixture model were used. The estimation algorithm was used with 10 replicate runs. The maximum number of clusters was set to values between 2 and 20, and the STs were assigned to the clusters with the highest posterior probability. Admixture inference was calculated using 100 Monte Carlo runs and 100 Monte Carlo reference samples to estimate the P values. An admixture inference threshold for possible recombination was set at ≤0.05.

A phylogenetic tree was constructed for the 46 STs from concatenated sequences of the seven genes (2,620 bp in total). First, the appropriate substitution model (K81uf+I) was selected by Modeltest (19) under the Akaike information criterion (AIC). Then, maximum-likelihood (ML) analysis was performed using the PhyML algorithm (20) implemented via PALM (21).

ClonalFrame (22) was used to estimate the recombination ratio for the population. For each data set, two runs of the ClonalFrame Markov chain Monte Carlo (MCMC) were performed, each consisting of 400,000 iterations. The first 25% of the chains were discarded, and the rest were sampled every hundred iterations. The Gelman-Rubin statistic (23) was then computed for r/m, the ratio of recombination to mutation events, in each data set to assess the convergence and mixing properties of the MCMC to ensure that the Gelman-Rubin statistic was below a threshold of 1.1.

Nucleotide sequence accession numbers.

The GenBank/EMBL accession numbers for the ddl, dnaK, gyrB, lepA, pgm, pheS, and rpoC gene sequences of the novel strains are KP146146 to KP146397, KP146398 to KP146649, KP146650 to KP146901, KP146902 to KP147153, KP147154 to KP147405, KP147406 to KP147657, and KP147658 to KP147909, respectively.

RESULTS

Nucleotide sequence data.

Initially, 10 housekeeping genes were selected for MLST from the complete genome of L. gelidum subsp. gasicomitatum LMG 18811^T (11), based on their locations in the genome and their putative roles as housekeeping genes under neutral selection. Three of the 10 genes (atpA, dnaA, and rpoA) were rejected because they either contributed with too little variation or were located too close to another gene on the chromosome. The lengths of the PCR products analyzed for the remaining seven genes varied from 322 to 400 bp (Table 1). The genes were unlinked and evenly distributed along the genome of LMG 18811^T (see Fig. S1 in the supplemental material). The genes encode functions associated with cell wall synthesis (ddl), DNA biosynthesis (dnaK), DNA replication (gyrB), translation (lepA and pheS), carbohydrate metabolism (pgm), and transcription (rpoC). The average pairwise nucleotide diversity per site ranged from 0.0005 to 0.00952 (Table 2). The number of alleles for each locus ranged from 2 to 15, and the number of polymorphic sites (S) ranged from 5 to 23. dnaK and pgm were the genes with the highest numbers of alleles (10 and 5, respectively) and polymorphic sites (19 and 23, respectively), whereas ddl, gyrB, lepA, pheS, and rpoC contributed less to the variation. The ratio of mean nonsynonymous substitutions and mean synonymous substitutions (dN/dS) was low, <0.1 (0 to 0.071), for all loci. The values from Tajima's D test (D), which measures deviation from the standard neutral model of evolution, ranged from −1.87 to 1.18. These values imply balancing (neutral) selection for all the selected genes, which is typical for housekeeping genes (24).

TABLE 2.

Average allelic and nucleotide diversity of the seven MLST loci

Gene	Size (bp)	D	dN/dS	π	% G+C	S	No. of alleles
ddl	373	1.18141	0	0.00546	0.385	8	7
dnaK	322	−1.87354	0.010	0.00289	0.367	19	10
gyrB	397	−0.81872	0	0.00152	0.436	6	2
lepA	382	−0.20005	0.008	0.00152	0.366	4	6
pgm	400	−0.19706	0.030	0.00952	0.375	23	15
pheS	364	−1.44745	0.013	0.00050	0.417	5	7
rpoC	400	−1.25345	0.071	0.00067	0.426	5	7

Population structure.

The seven housekeeping alleles of each of the 252 strains yielded a total of 46 STs. A majority of the strains (>60%) represented the three most prevalent STs, ST1 (n = 86), ST2 (n = 40), and ST7 (n = 35). ST1 and ST2 consisted of strains from MAP beef, pork, poultry, and lamb, whereas most of the strains representing ST7 (31 out of 35 strains) were from MAP poultry, including the type strain, LMG 18811^T. All strains representing ST14 (n = 5), ST22 (n = 4), ST29 (n = 6), ST34 (n = 3), and ST41 (n = 5) were from vegetable sources. Twenty-four STs accounted for only one strain.

The 46 STs were grouped into three CCs and 17 singletons at the SLV level by goeBURST (Fig. 1). CC1 consisted of 14 STs, including ST1 as the putative primary founder and ST2 as the secondary founder. The primary founder, ST1, was surrounded by eight SLVs and two DLVs. One of the SLVs, ST2, had become successful and had diversified further into two SLVs and a DLV. CC1 accounted for 155 isolates. The founder ST in CC2 was predicted to be ST7. ST7 had six SLVs, of which ST29, ST3, and ST12 had diversified further. Eighty-eight isolates represented STs in CC2. Three STs, ST11, ST16, and ST39, accounted for 17 isolates and formed CC3. Fifteen isolates represented ST11, whereas only one isolate represented both ST16 and ST39. Seventeen STs had no SLVs in the data set and were thus singletons.

FIG 1.

Minimum spanning tree reflecting clonal relationships of 46 L. gelidum subsp. gasicomitatum STs at the SLV level constructed using goeBURST. Each ST is represented by a circle, and the size of the circle is logarithmically proportional to the number of strains represented by the ST. The color of each circle represents the source of the strains belonging to the ST.

The STs in CC1 consisted mostly of strains from MAP poultry, pork, beef, lamb, and minced pork and beef (Fig. 1). CC1 contained 86% of the strains from MAP meat and 44% of the strains from MAP poultry. All STs containing strains from vegetable sources were predicted to belong to CC2 (60% of the strains) or were singletons. CC2 contained 43% of the poultry strains. CC3 contained 15 strains from MAP meat and 2 strains from MAP poultry. STs predicted to be singletons contained one to four strains from one source only (vegetable, MAP meat, or MAP poultry). The four strains from a fish product containing vegetables were grouped in CC1 (two strains representing ST17) and CC2 (two strains representing ST29).

The Bayesian approach was used to obtain an alternative view of the genetic population structure of the 252 L. gelidum subsp. gasicomitatum strains. Maximum posterior probability was found for a solution with three clusters, indicating that the population was divided into three ancestral subpopulations (Fig. 2). The population structure predicted by BAPS was mostly congruent with the grouping predicted by goeBURST, except that BAPS clustered CC2 and CC3 together. BAPS cluster 1 was mostly formed of STs from CC1, also including the founder STs, ST1 and ST2. BAPS cluster 1 also included four STs that were predicted to be singletons by goeBURST. Cluster 2 included almost all of the STs from CC2, all 3 STs from CC3, and 13 STs that were predicted to be singletons by goeBURST. Cluster 3 included ST37 and ST38 from CC1 and ST41 from CC2. The BAPS approach identifies admixture within the population, i.e., genotypes with sequences characteristic of more than one subpopulation. Based on the admixture analysis, the level of recombination within the L. gelidum subsp. gasicomitatum population studied was found to be very low, with only ST5 containing genetic elements from both BAPS clusters 1 and 2.

FIG 2.

Admixture analysis of 46 L. gelidum subsp. gasicomitatum STs by BAPS. Each row represents a single ST and is colored according to the percentage of genetic material from each ancestral subpopulation (K = 3). CC refers to clonal complexes predicted by goeBURST. Each vertical bar represents a single ST and is colored according to the proportion of genetic variation assigned to each BAPS group.

In addition to sources, we tested if strains from the same producer or PFGE cluster represented the same ST or grouped in the same clonal complex by goeBURST (see Fig. S2 and S3 in the supplemental material). No clonality was observed within strains from the same producer; the predominant STs, ST1, ST2, ST7, and ST11, were represented by strains from 6 to 15 producers, and each clonal complex contained strains from 9 to 22 producers. Strains from the same PFGE cluster, however, often represented the same ST or clonal complex. Strains in the predominant clusters represented one to four different PFGE clusters.

A phylogenetic tree (see Fig. S4 in the supplemental material) was constructed for the 46 STs by the maximum-likelihood method implemented in PALM (21) from concatenated sequences of the seven genes (2,638 bp). The STs grouped together by goeBURST were mostly located close together in the phylogenetic tree. However, only the three STs in BAPS cluster 3 (ST37, ST38, and ST41) were clearly differentiated from the other STs with a bootstrap value of 96%, thus supporting the grouping detected by BAPS. The bootstrap values for the tree were in general very low (<70%), indicating that the structure of the tree was statistically unsupported.

ClonalFrame (22) was used to measure the impact of homologous recombination on the genetic diversification of the population studied. The r/m value for the population was 1.9, with 95% confidence bounds from 1.84 to 1.88.

DISCUSSION

The novel MLST scheme presented in this study allowed the detection of 46 STs within a population of 252 food-related L. gelidum subsp. gasicomitatum strains. The strains represented different PFGE profiles and ribotypes and were from both vegetable and meat sources from 22 producers in Finland (a majority of the strains), the Baltic states, Spain, and New Zealand (see Table S1 in the supplemental material). L. gelidum subsp. gasicomitatum has been constantly associated with spoilage of, e.g., beef, pork, poultry, and vegetable products in Finland and other countries (2, 3, 5–9). This MLST scheme provides a tool for characterizing isolates from spoilage case studies globally.

All analyses indicated that the seven housekeeping genes sequenced for MLST were under stabilizing selection, and we consider them to provide a sufficient level of sequence diversity for studying the population structure of the subspecies. We used methods based on both allelic profiles (goeBURST) and sequence data (BAPS) to gain a comprehensive view of the L. gelidum subsp. gasicomitatum population studied. The analyses indicated that the population was divided into three subpopulations, with evidence of niche specificity in two major subpopulations. Furthermore, the population was predicted to be highly clonal, with the majority of the strains belonging to the three most prevalent STs and almost no exchange of genetic material between these subpopulations.

The population structure predicted by the Bayesian approach (BAPS) correlated well with the grouping detected by goeBURST. goeBURST aims to identify closely related strains based on allelic profiles and is less affected by recombination than many sequence-based methods (25). BAPS, on the other hand, is a Bayesian method that predicts whether the differences in sequences are due to recombination or mutation and takes this into account when grouping presumably related strains (18). goeBURST divided the L. gelidum subsp. gasicomitatum population into three CCs and 17 singletons, whereas BAPS grouped isolates within CC3 with strains from CC2, creating BAPS cluster 3. This may indicate that STs intermediate between these CCs are unsampled or extinct in our material. ST5, which according to BAPS contained genetic elements from both major BAPS clusters, and its SLV, ST20, were the only genotypes that were located in CC2 but clustered together with STs from CC1 in BAPS. The goeBURST singletons were grouped by BAPS with strains from the same ecological niche (poultry/vegetable or meat), which supports ecological separation of strains.

One of the two major subpopulations detected by goeBURST (CC1) was formed by strains mostly from MAP poultry and vegetable sources and the other (CC2) by strains from MAP meat (MAP beef, pork, poultry, and minced pork and beef). This may indicate that there is a subpopulation that grows especially well in a variety of MAP meats and thus causes spoilage of these products. Members of this group may also possess characteristics that aid survival in a meat-processing environment and that are nonexistent in other L. gelidum subsp. gasicomitatum strains. Johansson et al. (11) showed that the type strain, LMG 18811^T, isolated in spoiled MAP broiler meat and belonging to this group, has genes associated with adhesion, e.g., an ortholog for the putative collagen adhesion protein enabling biofilm formation in staphylococci. However, L. gelidum subsp. gasicomitatum has never been detected as prevailing in a meat-processing environment (26–29). In our previous contamination study, L. gelidum subsp. gasicomitatum was not associated with pork or poultry at the beginning of the slaughter line (27, 28). Thus, it is not likely that slaughter animals are introducing this subpopulation into meat-processing plants.

Interestingly, all analyses indicated that the vegetable strains grouped together with poultry strains. goeBURST predicted that CC2 contained 44% of the poultry strains and 60% of the vegetable strains, whereas BAPS cluster 2 contained 45% of the poultry strains and 97% of the vegetable strains. However, the 15 strains from MAP pork and beef from CC3 were also included in BAPS cluster 2. L. gelidum subsp. gasicomitatum has been associated with spoilage of MAP poultry products in Finland (2). The vegetable strains in this study were isolated from spoiled carrots, lettuce, and cooked vegetable sausages during our spoilage investigations for the food industry (6, 11). L. gelidum subsp. gasicomitatum and L. gelidum subsp. gelidum have also recently been identified as a major part of the spoilage microbiota of packed food, including vegetable salad, in Belgium (7–9). Pothakos et al. (8) studied the contamination of ready-to-eat vegetable salads at a processing plant and reported more pronounced contamination with L. gelidum subsp. gasicomitatum than in our previous studies associated with poultry meat production (28); every vegetable handled in the production environment was found to be cross-contaminated. Understanding why poultry and vegetable strains belong to the same subpopulation requires further studies with more isolates from different niches.

BAPS and phylogenetic analysis indicated that a third subpopulation of L. gelidum subsp. gasicomitatum exists that was represented by only three STs and nine strains within the population studied. This group was distinct from the other strains in the phylogenetic tree and was represented by strains from both vegetable and meat sources and from four producers. goeBURST placed these STs at the peripheries of CC1 (ST37 and ST38) and CC2 (ST41), probably due to the small number of strains belonging to the subpopulation. Strains in this group may have been underrepresented within the population studied.

Previous studies on the population genetics of food-associated bacterial species have shown that division into subpopulations reflects adaptations to different niches rather than geographical isolation (30, 31). The strains in this study were from food products manufactured at 22 production plants located across Finland. There was no evidence of clonal in-house populations of L. gelidum subsp. gasicomitatum, since strains from all producers represented several STs. This correlates with previous studies suggesting that leuconostocs are constantly introduced into production plants, even though some strains are able to persist longer than others in this harsh environment (10, 28). The L. gelidum subsp. gasicomitatum population apparently comprises strains adapting to and surviving the stressors present in the food-processing environment and flourishing in the niche created by MAP.

Diversification into subpopulations can occur through recombining or clonal expansion. In the L. gelidum subsp. gasicomitatum population studied, admixture analysis by BAPS indicated that only one ST, ST5, with three isolates from MAP broiler meat, possessed a sequence characteristic of two subpopulations. This indicates that recombination between subpopulations occurs rarely. The clonal complexes created by goeBURST appeared as star-like structures that are typical for populations with a low level of recombination within subpopulations (16). ClonalFrame predicted an intermediate recombination rate for the population (an r/m value of 1.9), which was somewhat higher than expected considering the r/m values obtained by Vos and Didelot (32) for other Firmicutes. However, the level of variance in r/m values was remarkably high for almost all species in that study. Both strategies of evolution seem to occur within food-associated LAB. In a study by Dan et al. (33), the MLST of 50 Leuconostoc lactis isolates revealed a highly clonal population structure with an indication of genetic exchange only within the two separately evolved subpopulations. Leuconostocs are known to possess restriction modification (RM) systems in the genome that destroy foreign DNA, which may partly explain their predilection for clonal evolution (34). A similar population structure was found for the wine-associated species Oenococcus oeni (31). Conversely, recombination both within and between subpopulations was common for Lactobacillus sakei (30).

In conclusion, the MLST scheme developed in this study is a robust tool for investigating the population structure of L. gelidum subsp. gasicomitatum. The 46 STs from vegetable and meat products provide a basis for a growing MLST database when more strains from new spoilage cases are added. It will open the information in our database to the international research community and promote global studies of this spoilage organism.