Complementary Antibacterial Effects of Bacteriocins and Organic Acids as Revealed by Comparative Analysis of Carnobacterium spp. from Meat

The results of this study demonstrated that both bacteriocins and organic acids are important factors contributing to the antibacterial activities of Carnobacterium from vacuum-packaged (VP) meats. This study demonstrated that formate and acetate are the key organic acids produced by Carnobacterium and demonstrated their association with the inhibitory activity of carnobacteria under VP meat-relevant storage conditions. The role of lactate, on the other hand, may not be as important as previously believed in the antimicrobial activities of Carnobacterium spp. on chilled VP meats. These findings advance our understanding of the physiology of Carnobacterium spp. to better explore their biopreservative properties for chilled VP meats.

ABSTRACT

Carnobacterium maltaromaticum and Carnobacterium divergens are often predominant in the microbiota of vacuum-packaged (VP) meats after prolonged storage at chiller temperatures, and more so in recent studies. We investigated the antibacterial activities of C. maltaromaticum and C. divergens (n = 31) from VP meats by phenotypic characterization and genomic analysis. Five strains showed antibacterial activities against Gram-positive bacteria in a spot-lawn assay, with C. maltaromaticum strains having an intergeneric and C. divergens strains an intrageneric inhibition spectrum. This inhibitory activity is correlated with the production of predicted bacteriocins, including carnobacteriocin B2 and carnolysin for C. maltaromaticum and divergicin A for C. divergens. The supernatants of both species cultured in meat juice medium under anaerobic conditions retarded the growth of most Gram-positive and Gram-negative bacteria in broth assay in a strain-dependent manner. C. maltaromaticum and C. divergens produced formate and acetate but not lactate under VP meat-relevant conditions. The relative inhibitory activity by Carnobacterium strains was significantly correlated (P < 0.05) to the production of both acids. Genomic analysis revealed the presence of genes required for respiration in both species. In addition, two clusters of C. divergens have an average nucleotide identity below the cutoff value for species delineation and thus should be considered to be two subspecies. In conclusion, both bacteriocins and organic acids are factors contributing significantly to the antibacterial activity of C. maltaromaticum and C. divergens under VP meat-relevant conditions. A few Carnobacterium strains can be explored as protective cultures to extend the shelf life and improve the safety of VP meats.

IMPORTANCE The results of this study demonstrated that both bacteriocins and organic acids are important factors contributing to the antibacterial activities of Carnobacterium from vacuum-packaged (VP) meats. This study demonstrated that formate and acetate are the key organic acids produced by Carnobacterium and demonstrated their association with the inhibitory activity of carnobacteria under VP meat-relevant storage conditions. The role of lactate, on the other hand, may not be as important as previously believed in the antimicrobial activities of Carnobacterium spp. on chilled VP meats. These findings advance our understanding of the physiology of Carnobacterium spp. to better explore their biopreservative properties for chilled VP meats.

INTRODUCTION

Carnobacterium divergens and Carnobacterium maltaromaticum are psychrotrophic lactic acid bacteria (LAB). Both species are major components of the microbiota of refrigerated foods, including seafood, meats, and dairy products, especially at later stages of storage (1, 2). Comparison of the spoilage-related activities of 45 C. maltaromaticum strains in air-stored and vacuum-packaged (VP) beef demonstrated that all these strains produce volatile organic compounds in a strain-independent manner; however, the volatile molecules produced by these strains are of low sensory impact, and their contribution to the spoilage of meat under either condition is negligible (3). The predominance of carnobacteria on food may be achieved by inhibiting the food microbiota through production of antimicrobial compounds. Not surprisingly, carnobacteria have been explored as protective cultures to inhibit pathogens and spoilage organisms for meat, seafood, and dairy products (1, 2). C. maltaromaticum strain CB1 has been approved for use in several countries, including the United States, Canada, Mexico, Colombia, and Costa Rica (4, 5).

The effect of carnobacteria on their companion meat microbiota relates to the strain-dependent production of antibacterial compounds, including bacteriocins and organic acids. Bacteriocins are ribosomally synthesized antimicrobial peptides and are classified on the basis of structural properties into three major groups, classes I, II, and III (6–8). Class I bacteriocins are small (<5 kDa), posttranslationally modified peptides; class I bacteriocins particularly include lantibiotics containing the unusual amino acid lanthionine (Lan). Their mode of action varies with their structure (9). Class II bacteriocins are also <5 kDa but are unmodified peptides that typically permeabilize bacterial cell membranes. They are further divided into four groups, IIa to IId, depending on their structures. Both class I and class II bacteriocins are thermostable. Class III comprises bacteriolysins, which are larger than 30 kDa and thermolabile (6, 7). Class IIa bacteriocins, divercin V41 and divergicin M35, and class IId, divergicin A, have been isolated from various C. divergens strains (10–12). Compared to C. divergens, a larger number of bacteriocins have been reported for C. maltaromaticum, including class I (carnolysin), class IIa (carnobacteriocin B2 and BM1, piscicolin 126 and CS526, and maltaricin CPN), and class IId (carnobacteriocin X) (13–20).

Organic acid metabolism by carnobacteria and their role in the microbial ecology of meat and meat products are not as well understood as bacteriocin production. Carnobacteria are homofermentative lactic acid bacteria that metabolize glucose via the Embden-Meyerhof pathway to pyruvate as a central metabolic intermediate (21–24). Pyruvate is further converted to the alternative end product lactate or acetate, ethanol, formate, and CO₂; respiratory metabolism in the presence of oxygen can additionally influence the type and concentration of organic acids that are produced (21–24). Organic acid and bacteriocin production by carnobacteria is not only strain dependent but is influenced by the condition under which they are cultured (21–23, 25, 26). The quest to isolate and characterize new strains continues to draw attention from the research community, in part at least, due to the strain-dependent nature of the antibacterial activities of carnobacteria.

A few recent reports have demonstrated that Carnobacterium spp., particularly C. maltaromaticum, are associated with extremely long storage life of VP beef (27–29). However, the exact mechanisms by which they contribute to shelf life are not well understood, as information on the production of either bacteriocins or organic acids by and the antibacterial activities of these strains/species under VP meat-relevant anaerobic conditions is largely unavailable. Our previous research reported draft genomes of 31 Carnobacterium strains recovered from VP beef and pork and divided them into 11 (I to XI) phylogenetic groups by core genome alignment (30). The objective of the present study was to characterize these strains by phenotypic analysis under anaerobic conditions and genomic analysis with a focus on their antibacterial properties, as well as their phylogenetic positioning in relation to the 18 C. maltaromaticum and 5 C. divergens strains of which genomes are publicly available.

RESULTS

Phylogenetic analysis.

Strains of C. divergens and C. maltaromaticum of which the genomes were available in GenBank were clustered into two distinct groups (Fig. 1). Within the C. maltaromaticum cluster, strains isolated from diseased sharks formed a clade of mostly clonal strains. C. maltaromaticum strains (A1, A5, A7, and A14) isolated from chilled VP beef in Alberta clustered with strains from poultry, milk, or VP meat. The 32 C. divergens strains in the phylogenetic tree were also clustered into two major groups, without any apparent pattern between clustering and source of isolation.

FIG 1.

Phylogenetic tree of the sequenced genomes of C. maltaromaticum and C. divergens. The tree was constructed based on the alignment of concatenated nucleic acid sequences of eight housekeeping genes (ack, dnaB, dnaK, ftsZ, greA, gyrA, ileS, and polA) of the Carnobacterium strains and E. faecalis V583. The value shown on each node is the local bootstrap support value of 1,000 replicates. The scale bar represents the number of substitutions per site. The visualization of the tree was improved by truncating long branches, with the length of each truncated branch labeled.

The orthologous average nucleotide identity (ANI) between strains of C. divergens and C. maltaromaticum was 74.3 to 75.2% (see Table S1 in the supplemental material). The intraspecies ANI was 96.6 to >99.9% for C. maltaromaticum and 93.3 to >99.9% for C. divergens. An ANI of 95 to 96% has been suggested for species delineation (31); the two clusters of C. divergens share an intercluster ANI of 93.3 to 93.9%, which may suggest that they should be classified as distinct species or subspecies.

Genome characterization and pangenome analysis.

The size of the assembled C. maltaromaticum and C. divergens draft genomes was 3.50 to 3.53 and 2.62 to 2.88 Mbp, respectively (Table S2). The GC content was 34.9 to 35.5% and 34.3 to 34.5%, respectively. The number of RNA-encoding genes ranged from 45 to 61 in C. divergens and from 57 to 71 in C. maltaromaticum.

The pangenomes of the C. maltaromaticum (n = 4) and C. divergens (n = 27) strains had 1,087 (30.3%) and 1,867 (48.9%) accessory genes, respectively (Fig. S1A). The categories each accounting for >3% of the accessory genomes in both species were G (carbohydrate transport and metabolism), K (transcription), L (replication, recombination, and repair), M (cell wall/membrane/envelope biogenesis), V (defense mechanism), W (extracellular structures), and genes with unassigned functions or bacteriophage genes (Fig. S1B). C. maltaromaticum had larger fractions of G, K, L, M, V, and W in the accessory genome than C. divergens. A CRISPR gene cluster was identified in the genomes of all strains of C. divergens groups IV and V (Table S2). Prophage genes, important contributors to genomic plasticity, were found in all strains. However, an intact prophage region was only present in the genomes of C. divergens group IV, V, VI (except B7), IX, and XI strains.

Functional genome analysis.

To assess the potential contribution of bacteriocin production to meat preservation, genes encoding putative bacteriocins were identified in the strains of C. divergens and C. maltaromaticum (Table 1). All four C. maltaromaticum genomes include the carnobacteriocin BM1/B1 structural gene, with 100% identity to the reference peptide sequence. Putative encoding genes for both subunits of carnolysins were found in groups I and II C. maltaromaticum. For C. maltaromaticum A5 and A14, a peptide with 98.5% similarity to carnobacteriocin B2 was also predicted. The difference in sequence was caused by a substitution of Asn (N) with Tyr (Y) before the signature motif YGNGV of carnobacteriocin B2 (Fig. S2). Production of divergicin A was predicted for strains of C. divergens groups VI and IX. In addition, a lanthipeptide gene cluster with genes encoding homologues of carnolysin subunit A2, cerecidin, and cytolysin small subunit was predicted for C. divergens C13 (Fig. S3). The respective identities to reference sequences were 46.8, 36.7, and 34.9% (Table 1; Fig. S4).

TABLE 1.

Predicted bacteriocins in Carnobacterium spp. isolated from vacuum-packaged meat

Bacteriocin	% identity between query and reference peptide sequences
	C. maltaromaticum groups and strains			C. divergens groups and strains
	I	II	III	IV	V	VI	VII	VIII	IX	X	XI
	A7b	A5,b A14	A1b	A10b	B1,b B6, C5, C7, C10,b C14	B7,b B8, C6,b C16	C13b	B3,b C1, C12,b C17	C8b	B5,b C3, C4, C9,b C11, C15	A2,b A4, A8, A12
Class IIa
Carnobacteriocin BM1/B1	100	100	100	–c	–	–	–	–	–	–	–
Carnobacteriocin B2/Carnocin CP52	–	98.5	–	–	–	–	–	–	–	–	–
Class IId	–	–	–	–	–	100	–	–	100	–	–
Divergicin A	–	–	–	–	–	100	–	–	100	–	–
Class I (lantibiotics)
Carnolysina
Subunit A1	100	100	–	–	–	–	–	–	–	–	–
Subunit A2	100	100	–	–	–	–	46.8	–	–	–	–
Cerecidin	–	–	–	–	–	–	36.7	–	–	–	–
Cytolysin
Large unit	–	–	–	–	–	–	–	–	–	–	–
Small unit	–	–	–	–	–	–	34.9	–	–	–	–

^aCarnolysin is a two-peptide lantibiotic.

^bThe strain was selected for evaluation of inhibitory activity using the spot-lawn and broth assays.

^c–, the bacteriocin-encoding gene was not found in the genome(s).

The potential of strains to convert carbohydrates to organic acids was assessed by analysis of key genes involved in carbohydrate metabolism and respiration. Genes encoding key enzymes for the homofermentive pathway for both hexose (fructose-bisphosphate aldolase, EC 4.1.2.13; phosphofructokinase, EC 2.7.1.11) and pentose (transaldolase, EC 2.2.1.2; transketolase, EC 2.2.1.1) were present in all genomes (Fig. 2). The gene encoding 2-dehydro-3-deoxyphosphogluconate aldolase (EC 4.1.2.14), the key enzyme of the Entner-Doudorof pathway, was also identified. Phosphoketolase (EC 4.1.2.9), the key enzyme in the phosphoketolase pathway, was absent in all carnobacteria (Fig. 2). Pyruvate formate-lyase (EC 2.3.1.54) was found in both C. maltaromaticum and C. divergens. Genes for four enzymes involved in pyruvate metabolism to lactate, acetate, and CO₂, including lactate dehydrogenase (EC 1.1.1.27/28), pyruvate dehydrogenase (EC 1.2.4.1), phosphotransacetylase (EC 2.3.1.8), and acetate kinase (EC 2.7.2.1) (24), were all found in all strains. C. divergens and C. maltaromaticum differed with respect to the number of genes coding for l-lactate or d-lactate dehydrogenases (Fig. 2). The genes encoding two key enzymes for lactate-to-propionate and propanol, lactaldehyde dehydrogenase (EC 1.2.1.22), and glycerol/diol dehydratase (EC 4.2.1.28) (32, 33), were not found in any of the 31 strains, indicating the inability of C. maltaromaticum and C. divergens to produce propionate or propanol. The genes encoding key enzymes involved in respiration were found in all 31 strains (Fig. 2). IsdA to IsdI mediate heme uptake by Staphylococcus aureus (34). Homologous genes of IsdA, C, E, F, and G/I were present in all 31 Carnobacterium genomes. However, isdB, D, and H were not found in any of the genomes.

FIG 2.

The presence/absence of genes encoding proteins involved in different biological pathways in Carnobacterium genomes. Pathways are differentiated by color, with the intensity of shading increased with the increasing number of genes. The numbers indicate copies of genes. The abbreviations for carbohydrate metabolism are FBA, fructose-bisphosphate aldolase (EC 4.1.2.13); PFK, phosphofructokinase (EC 2.7.1.11); PK, phosphoketolase (EC4.1.2.9); TAK, transketolase (EC 2.2.1.1); TAL, transaldolase (EC 2.2.1.2); EDA, 2-dehydro-3-deoxyphosphogluconate aldolase (EC 4.1.2.14); PFL, pyruvate formate-lyase (EC 2.3.1.54); LDH, lactate dehydrogenase (EC 1.1.1.27/28); PDH, pyruvate dehydrogenase alpha- and beta-unit (EC 1.2.4.1); PTA, phosphotransacetylase (EC 2.3.1.8); ACK, acetate kinase (EC 2.7.2.1); LactAldDH, lactaldehyde dehydrogenase (EC 1.2.1.22); and G/D-DHA, glycerol/diol dehydratase (EC 4.2.1.28/30). The abbreviations for the respiration system are NADH_DH, NADH dehydrogenase (EC 1.6.99.3) and CydABCD, cytochrome d ubiquinol oxidase (EC 1.10.3). menF (EC 5.4.4.2), menD (EC 2.2.1.9), menH (EC 4.2.99.20), menC (EC 4.2.1.113), menE (EC 6.2.1.26), menB (EC 4.1.3.36), menA (EC 2.5.1.74), and ubiE (EC 2.1.1.-) are enzymes for biosynthesis of menaquinone (electron transporter). IsdA to IsdI are proteins involved in the iron surface determinant (Isd) heme uptake pathway.

Putative genes encoding components of importance to acid tolerance in LAB, including ATP synthase (F₁F₀), arginine deiminase (EC 3.5.3.6), glutamate decarboxylase (EC 4.1.1.15), and Na⁺/H⁺ antiporter, were found in all 31 genomes. Other potential acid resistance components, such as ornithine decarboxylase (EC 4.1.1.17), glutaminase (EC 3.5.1.2), agmatine deiminase, and urease, were not found in any of the genomes. Similarly, the enzyme responsible for production of biogenic amines, histidine decarboxylase (EC 4.1.1.22), was not found in any of the genomes, but tyrosine decarboxylase (EC 4.1.1.25) was found in all 31 Carnobacterium genomes. A putative glycogen biosynthesis cluster was found in the strains of C. divergens group XI but not in any of the other strains. Putative tetracycline resistance genes were found in most Carnobacterium strains except for A7, A5, A14, B7, and C13, but genes encoding aminoglycoside, β-lactam, colistin, or fluoroquinolone resistance were absent in the 31 genomes. In addition, among strains with putative tetracycline resistance, the genetic determinant tet(M) was found in group III, IV, and V strains, while tet(S) was found in other groups; mobile elements were found in the flanking region of tet(M)/tet(S) of all these strains.

Antibacterial activity determined by spot-lawn assay and broth assay.

To correlate the potential of carnobacteria for production of organic acids and bacteriocins with their inhibitory activity, inhibition of meat spoilage-associated bacteria and pathogens by 15 Carnobacterium strains was assessed in two assays, a spot-lawn assay and a broth assay. In the spot-lawn assay, C. maltaromaticum A7 (group I) and A5 (group II) showed the largest inhibition spectrum, with inhibition observed for most Gram-positive indicator bacteria, including Listeria monocytogenes, Lactobacillus sakei, Leuconostoc gelidum, Enterococcus spp., C. maltaromaticum, and C. divergens (Table S3). A5 resulted in larger inhibition zones (1.3 to 4.2 mm) than A7 (0.5 to 2.9 mm). C. divergens groups VI (B7 and C6) and IX (C8) had identical inhibition spectra and were inhibitory against all C. maltaromaticum strains and C. divergens A10, B3, C12, and A2. The inhibition of these C. divergens strains was stronger on C. divergens (3.1 to 5.1 mm) than on C. maltaromaticum (0.7 to 1.8 mm). Mutual inhibition was found between C. maltaromaticum A7 and A5 and C. divergens B7, C6, and C8 (Table S3 and Fig. S5). However, the inhibitory spectrum of strains B7, C6, and C8 was limited to Carnobacterium spp. None of the 15 Carnobacterium strains produced an inhibition zone against any of the Gram-negative bacteria tested. Group III, IV, V, VII, VIII, X, and XI Carnobacterium spp. did not show any inhibition by spot-lawn assay against any of the 45 indicator bacteria (Table S3). In short, the inhibitory activity of carnobacteria as determined by the spot-lawn assay (Table S3) matched their potential to produce bacteriocins except for strains A1 and C13 (Table 1).

In broth assay, the cell-free supernatants of all 15 strains of Carnobacterium spp. retarded the growth of indicator bacteria (Table S4 and Fig. S6). Growth inhibition was observed for all combinations except for C. divergens B7 against Hafnia alvei S1, C6 against Rahnella sp. S8, B5 against Serratia liquefaciens, and C9 against H. alvei S1. The mean inhibition strength of combinations of effector and indicator bacteria showing inhibition zones in the spot-lawn assay was significantly larger than that of combinations which did not show inhibition zones in the spot-lawn assay (P < 0.05; Fig. 3).

FIG 3.

Inhibition strength of cell-free supernatants against indicator bacteria determined by broth assay of Carnobacterium strains that did not (NO; 620 combinations) or did (YES; 55 combinations) show inhibition in the spot-lawn assay. Boxes indicate data within the first and third quartile (interquartile range [IQR]), central horizontal lines indicate medians, × marks indicate means, and whiskers indicate the lowest data point within 1.5 IQR of the first quartile and the highest data point within 1.5 IQR of the third quartile.

To investigate the contribution to inhibition by factors other than bacteriocins, the combinations not showing inhibition in the spot-lawn assay were further analyzed. Significant differences were found among the Carnobacterium strains (P < 0.05), with strains C13 (VII) and B3 (VIII) showing the largest and smallest mean inhibition strengths, respectively (Fig. 4). The overall response of indicator bacteria to inhibitory compounds in the cell-free supernatants of Carnobacterium was also strain dependent (Fig. 5).

FIG 4.

Inhibition strength of cell-free supernatants of Carnobacterium strains against indicator bacteria determined using broth assay. The combinations of Carnobacterium spp. and indicator bacteria that showed a visible inhibition zone in the spot-lawn assay were not included in the analysis. The difference between means is significant (P < 0.05) if two strains do not share a letter. Boxes indicate data within the first and third quartile (interquartile range [IQR]), central horizontal lines indicate medians, × marks indicate means, and whiskers indicate the lowest data point within 1.5 IQR of the first quartile and the highest data point within 1.5 IQR of the third quartile.

FIG 5.

Mean inhibition strength of Carnobacterium on the growth of 45 indicator bacteria. The combinations of Carnobacterium spp. and target bacteria which showed inhibition zones in the spot-lawn assay were not included in the statistics. Error bars represent standard errors of mean. Means with different letters are significantly different (P < 0.05).

Production of organic acids by Carnobacterium spp.

The metabolism of strains of Carnobacterium spp. reduced the pH of meat juice medium (MJM) (pH 5.93 ± 0.02) by 0.55 to 0.73 (Fig. 6A), indicating acid production. The concentration of glucose in the supernatants of these cultures was reduced by 1.75 to 2.66 mM compared to blank MJM which contained 5.83 ± 0.24 mM glucose (Fig. 6B). Concentrations of formate and acetate in the supernatants were increased by 3.32 to 4.64 mM and 1.61 to 2.21 mM, respectively, compared to blank MJM (Fig. 6C and D). The concentration of lactic acid, propionic acid, isobutyric acid, butyric acid, isovaleric acid, valeric acid, or caproic acid in the supernatants was not significantly different from their respective concentrations in blank MJM (data not shown). These data correspond well with the metabolite concentrations that are expected for homofermentative metabolism of glucose via glycolysis and pyruvate formate lyase.

FIG 6.

Analysis of the supernatants of Carnobacterium spp. cultured in meat juice medium at 2°C for 10 days under anaerobic conditions. (A) pH. (B) Decrease in glucose concentration compared to blank medium. (C) Increase in acetic acid concentration compared to blank medium. (D) Increase in formic acid concentration compared to blank medium. Error bars represent standard errors of mean of two independent replicates, each with four technical replicates. The difference between means is significant (P < 0.05) if two strains do not share a letter.

For the combinations of effector and inhibitor bacteria that did not show inhibition in the spot-lawn assay, the relative inhibition strength determined in the broth assay was significantly correlated (P < 0.05) with the protonated formic and acetic acids in the supernatant of Carnobacterium MJM cultures (Table 2). The pH was correlated with the production of both formate and acetate (P < 0.05).

TABLE 2.

Correlation between IS, production of formate or acetate by Carnobacterium, and the pH of the spent medium when cultured in MJM^a

Dependent variable and Pearson correlation test parameter	pH	Total formic acid	Total acetic acid	Undissociated formic acid	Undissociated acetic acid
IS
P value	0.085	0.087	0.136	0.042	0.047
Correlation coefficient	–0.459	0.457	0.403	0.531	0.521
pH
P value	\	0.003	0.028	\	\
Correlation coefficient	\	–0.716	–0.565	\	\

^aIS refers to the inhibition strength of Carnobacterium spp. determined by broth assay for effector-indicator combinations which did not produce any visible inhibition zone in the spot-lawn assay. Text in bold indicates analysis where a statistical difference was observed. \, the correlation was not tested.

DISCUSSION

This study investigated the antibacterial potential of Carnobacterium spp. recovered from VP meats via both comparative genomic analysis and phenotypic characterization. The use of carnobacteria as biopreservatives and their role in meat spoilage are controversial. One contributor to this controversy is that carnobacteria were often found to be a sizable component of the microbiota of spoiled products, and all major components were regarded as spoilage bacteria, in particular, in earlier storage life studies. Investigation of the spoilage-related activities of 45 C. maltaromaticum strains on meat revealed that all these strains produced organic volatile compounds; however, their contribution to spoilage was negligible (3). Another factor contributing to the controversial role of carnobacteria is that the by-products generated by carnobacteria when growing on meat are not only species dependent, but also strain dependent (1, 2, 23). The inter- and intraspecies differences found in this study in bacteriocin and organic acid production provide insight into mechanisms by which certain strains of C. maltaromaticum extend the storage life of VP meat (28, 30).

The functional genome analysis of this study was based on the draft genome sequences of 31 Carnobacterium strains with a sequencing depth that is sufficient for phylogenetic analysis and for identification of genes coding for bacteriocin production and carbohydrate metabolism and other metabolic properties that relate to their use as biopreservatives (35, 36).

Carnobacterium spp. tolerate acid conditions, with C. divergens being more tolerant than C. maltaromaticum, in particular, group XI C. divergens (30). Genomic analyses did not reveal any difference with respect to genes related to acid resistance. However, group XI C. divergens differs from other strains with respect to the presence of a glycogen biosynthesis cluster. A comparable glycogen biosynthesis cluster not only mediated growth of Lactobacillus acidophilus in the absence of glucose, but also related to the bile resistance of this organism (37).

Transferable antibiotic resistance genes should be absent in cultures that are intentionally used in food production; in addition, decarboxylation of histamine and tyramine to biogenic amines is undesirable (38, 39). The 31 Carnobacterium genomes encoded tyrosine decarboxylase but not histamine decarboxylase, matching phenotypic data (2). Food and storage condition-related tyramine production should thus be examined prior to using these strains as bioprotective cultures in food. Information on antibiotic resistance genes in carnobacteria is scarce. None of the 31 Carnobacterium strains harbor genes for aminoglycoside, β-lactam, colistin, or fluoroquinolone resistance. Tetracycline is the most commonly used antibiotic in beef production in Canada, and its usage has been linked to increased prevalence of resistant Escherichia coli of bovine origin (40). The prevalence of tetracycline resistance genes in most Carnobacterium strains investigated in this study may relate to their bovine origin (41).

Many LAB respire in the presence of heme and oxygen (42), with menaquinone being required for some species (42–44). Heme-dependent oxygen consumption has been noted for some strains of C. maltaromaticum (25, 45) and contributed to meat spoilage by L. gelidum subsp. gasicomitatum (46). Genes encoding all components of the respiration chain were found in all 31 strains examined in this study. The hemoglobin and heme uptake system in the strains, however, was incomplete, with homologs for uptake of hemoglobin and haptoglobin missing (34, 47). This suggests that carnobacteria are equipped with heme uptake systems, while hemoglobin or haptoglobin may not be used. The cytochrome bd oxidase works at low oxygen concentrations and improves stationary-phase survival of Lactococcus lactis by reducing oxidative and acid stress (48). A low level of oxygen is present on the surface of VP meat, and the heme-based respiration of Carnobacterium spp. may then contribute to their competitiveness through a growth advantage and reducing stress (46).

Genome analysis predicted production of several bacteriocins, particularly by those strains of Carnobacterium that were most inhibitory in the spot-lawn assay. The predicted carnobacteriocin B2 in C. maltaromaticum group II strains (A5 and A14) has an N2Y mutation next to the conserved YGNGV motif at the N terminus. Substitutions in the conserved motif or the C terminus of carnobacteriocin B2 render the bacteriocin inactive (49), but an active N2Y variant of carnobacteriocin B2 has been reported (15). C. maltaromaticum A1 harbors genes for structural (cbnBM1) and immunity (cbiBM1) proteins of carnobacteriocin BM1 (CbnBM1), an antilisterial class IIa bacteriocin, but this strain did not show any antibacterial activity by the spot-lawn assay. In C. maltaromaticum LV17B, carnobacteriocin BM1 production is dependent on the plasmid-borne gene cluster cbnKRTD (47, 50), which is absent in C. maltaromaticum A1. The lack of antibacterial activity by C. divergens C13 in the spot-lawn assay is likely due to the missing structural genes for the carnolysin subunit A1 or cytolysin large subunits, which are required for an active bacteriocin (51–54). The inhibitory spectrum of the strains examined in this study is in agreement with previous reports that the inhibitory spectrum of carnolysin and carnobacteriocin B2 is relatively broad, while divergicin A inhibits only strains of the same genus. Taken together, the antimicrobial activities observed in the spot-lawn assay are well explained by the genome analysis of putative genes involved in producing bacteriocins.

In contrast to bacteriocins, organic acids at low concentrations are often bacteriostatic rather than bactericidal, and their effect is not readily detected in the spot-lawn assay (55). Therefore, a broth assay in MJM was performed. The growth of most Gram-positive and Gram-negative bacteria was retarded by the cell-free supernatants of C. maltaromaticum and C. divergens, including those strains that did not have potential for bacteriocin production. This suggests that bacteriocin-less strains also have potential for shaping the microbial community structure on VP meats. In particular, the inhibition of Gram-negative spoilage organisms was not related to bacteriocin formation but was entirely dependent on the formation of organic acids. Interestingly, all 15 effector strains were self-inhibitory in the broth assay, and often, the strongest inhibition was observed for intraspecific or intrageneric combinations, which could be a strategy for total population control in a closed system (56).

Lactate is the primary metabolite of Carnobacterium from glucose when strains are grown at high substrate concentrations and without the addition of lactate (21, 57, 58). In contrast, we observed formation of acetate and formate as the major or only metabolites in MJM that contained low concentrations of glucose but 40 mM lactate. Similarly, C. maltaromaticum and C. divergens produced acetate but not lactate during storage of modified-atmosphere-packaged shrimp at 5°C (23). The genomes of carnobacteria encoded a comparable set of genes related to carbohydrate metabolism; accordingly, quantitative but not qualitative differences were observed in the metabolite pattern of different strains. Under carbohydrate-limiting conditions, homofermentative LAB metabolize pyruvate via pyruvate formate-lyase or pyruvate dehydrogenase to formate/CO₂, acetate, and ethanol as an alternative to lactate formation (24), with pyruvate formate-lyase being preferentially used under anaerobic conditions at near-neutral pH (59, 60). In addition, downregulation of l-lactate dehydrogenase and upregulation of pyruvate formate-lyase in response to low pH and high lactate concentrations were observed in Lactococcus lactis (61). The lack of lactate production by Carnobacterium spp. observed in this study could then be caused by feedback inhibition of lactate dehydrogenase by the relatively high concentration of lactate in MJM. Our data on the relative amounts of formate, acetate, and lactate produced under anaerobic conditions in MJM, along with the previous observation of a lack of lactate production by Carnobacterium in modified-atmosphere-packaged shrimp, suggest that lactate is not the primary organic acid produced by Carnobacterium when growing on muscle food, where the concentration of naturally occurring lactate is high. Instead, formate and acetate are produced, and they both contribute to the antibacterial activity of Carnobacterium spp.

In conclusion, this study shows that both C. maltaromaticum and C. divergens, especially the former species, can be explored as protective cultures to improve meat quality and safety due to their potential of producing bacteriocins and organic acids, in particular, formate and acetate. To date, investigations of the use of lactic acid bacteria as biopreservatives focused on bacteriocin-producing strains. However, owing to the complementary inhibitory spectra and modes of action of bacteriocins and organic acids, both types of metabolites should be exploited for meat preservation.

MATERIALS AND METHODS

Bacterial strains.

Four C. maltaromaticum and 27 C. divergens strains, covering 11 phylogenetic groups from the culture collection of the Lacombe Research and Development Centre, were included in this study (30) (Table S2). These strains had been recovered from VP beef and pork primal cuts from three different abattoirs after storage at –1.5 and 2°C for up to 180 days. Isolates recovered from the same meat sample and with matching genomes were regarded as clonal isolates of the same strain, and isolates from different samples or with different genome sequences were regarded as different strains (30). All 31 strains were included for genomic analysis, and 15 strains (effector strains) were selected for further phenotypic characterization in relation to their biopreservative activities against 45 indicator bacterial strains. These 45 indicator strains included 17 foodborne pathogens, including Listeria monocytogenes, various serotypes of Shiga toxin-producing E. coli, and Salmonella enterica serovar Typhimurium, 13 spoilage bacteria recovered from meat, including Hafnia, Rahnella, Serratia, Lactobacillus, Leuconostoc, and Enterococcus strains, and the 15 effector strains of Carnobacterium spp. (Tables S3 and S4).

Phylogenetic analysis.

Genomes of the 31 VP meat Carnobacterium strains from our culture collection were sequenced using Illumina MiSeq PE 250-bp technology with a read depth of >50 for each genome (30). SPAdes was used for assembly, and contigs of <1,000 bp were removed. The assembled draft genomes contained 9 to 68 contigs (Table S2). The draft genomes were annotated using RAST version 2.0 (62). The overall similarity between any two genome sequences was calculated using an orthologous average nucleotide identity tool (OAT) (https://www.ezbiocloud.net/tools/orthoani) (63). The phylogenetic positions of these 31 VP meat Carnobacterium strains in relation to other strains of these two species were determined. Briefly, all additional genomes of C. maltaromaticum (n = 18) and C. divergens (n = 5) currently available in GenBank were downloaded (Table S5). The sequences of eight housekeeping genes (41) encoding acetate kinase (ack), replicative DNA helicase (dnaB), molecular chaperone (dnaK), cell division protein (ftsZ), transcription elongation factor (greA), DNA gyrase subunit A (gyrA), isoleucine-tRNA ligase (ileS), and DNA polymerase (polA) were extracted from all the above Carnobacterium genomes (n = 54) and aligned using MUSCLE (Codons) in MEGA X (64). The alignments of each gene were concatenated using Geneious version 11.0.5 (65). A neighbor-joining tree was constructed using the Kimura 2-parameter model in MEGA X, including transitions and transversions. The variation rate among sites was modeled with a Gamma distribution. The tree was rooted against Enterococcus faecalis V583, and the branch support was assessed with 1,000 bootstrap replicates.

Pangenome analysis.

Protein sequences (.faa file) produced from RAST for the 31 VP meat C. maltaromaticum and C. divergens genomes were compiled on the species level. The sequences were then clustered using USEARCH version 10.0.240 at an identity threshold of 0.6 (66, 67). The seed sequences were parsed using a Python script to produce a pangenome file showing the presence/absence of genes for C. maltaromaticum and C. divergens (67). Protein sequences shorter than 50 amino acids were removed prior to downstream analysis. The function of the seed protein sequences of all clusters was predicted using the Web CD-Search tool (https://www.ncbi.nlm.nih.gov/Structure/bwrpsb/bwrpsb.cgi) with the following settings: database, COG (clusters of orthologous groups of proteins); expected value threshold, 0.01; composition-corrected scoring turned on; maximum number of hits, 500.

Functional genome analysis.

Clustered regularly interspaced short palindromic repeats (CRISPRs) and prophage genes were identified for each genome using CRISPR Recognition Tool (CRT) version 1.1 (68) and PHASTER (http://phaster.ca/), respectively. For each genome, putative bacteriocin open reading frames (ORFs) were identified using the online server Bagel 4 (http://bagel4.molgenrug.nl/index.php). Antibiotic resistance genes were searched for each genome using the ResFinder 3.0 server hosted by the Center for Genomic Epidemiology (https://cge.cbs.dtu.dk/services/ResFinder/) (69). Enzyme/protein-encoding genes involved in carbohydrate metabolism, respiration system, and acid resistance were searched in the pangenome files created above. To investigate genes encoding the proteins required for the uptake of heme and glycogen biosynthesis, a local database containing protein sequences of 31 Carnobacterium genomes was searched using BLAST+ (ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/2.7.1/), with reference genes listed in Table S6. An E value of <1e⁻⁵ was used as the cutoff threshold for BLAST (67, 70). Matched protein sequences were subjected to InterPro (https://www.ebi.ac.uk/interpro/search/sequence-search) to examine them for the presence of functional domains.

Preparation of inoculum.

Anaerobic meat juice medium (MJM) was prepared and made anaerobic following the protocol described by Yang et al. (71), with external glycogen being omitted. The 15 Carnobacterium effector strains (Table S2) were grown in half-strength brain heart infusion broth (Oxoid, Mississauga, ON, Canada) at 25°C for 24 h, and 0.1 ml of the overnight cultures was inoculated in 7 ml of anaerobic MJM and incubated at 2 ± 0.5°C until stationary phase (10 days). Growth was assessed by optical density at 600 nm (OD₆₀₀) using a cell density meter (model 40; Thermo Fisher Scientific, Ottawa, ON, Canada). Indicator microorganisms were grown in half-strength brain heart infusion broth at 25°C for 24 h. Cultures of the effector and indicator microorganisms thus prepared were used as inoculum throughout the study unless otherwise stated.

Antibacterial activity of Carnobacterium spp.

Antibacterial activities of the 15 Carnobacterium effector strains were assessed by spot-lawn agar assay and broth assay (55). For the spot-lawn assay, 20 μl of each indicator brain heart infusion culture was plated on tryptone soya agar (Oxoid) as a lawn by a spiral plater (Eddy Jet 2; Neutec, Farmingdale, NY, USA). Two aliquots (technical replicates) of 10 μl of each Carnobacterium effector strain grown in anaerobic MJM were then spotted onto the lawn of each plate, and two independent plates were used for each combination of effector and indicator strains. After air drying for 10 min, the agar plates were placed in an anaerobic jar. Anaerobic conditions were created using an atmosphere generation sachet (AnaeroGen; Thermo Fisher Scientific, Ottawa, ON, Canada). After incubation at 15°C for 5 days, the plates were photographed, and the diameter of the inhibition zone was measured using the software Image J (https://imagej.nih.gov/ij/).

For the broth assay, cell-free supernatant (CFS) of each Carnobacterium effector strain was prepared by filtrating the anaerobic MJM culture through a 0.2-μm pore-sized syringe filter (Nalgene; Thermo Fisher Scientific, Ottawa, ON, Canada). The brain heart infusion culture of each indicator strain was adjusted to OD₆₀₀ of 0.2 to 0.25 (10⁸ CFU/ml). The adjusted bacterial suspension was further diluted using MJM to approximately 10⁵ CFU/ml, and 100 μl of the dilution was added to wells of a 96-well microtiter plate, each containing 100 μl of CFS (treatment), phosphate-buffered saline (buffer, pH 5.98 ± 0.07, control; Mediatech, Inc., Manassas, VA, USA) or MJM (control). The relatively low initial indicator level was intended for exploring the potential of these strains as biocontrol measures for fresh meat, where the initial level of contamination is, in general, less than 4 log CFU/cm² (72). Each treatment or control had duplicate wells. The microtiter plates were incubated at 15°C under anaerobic conditions as before. The OD₆₀₀ was measured every 24 h for up to 5 days using a plater reader (POLARstar; Omega, BMG Labtech, Germany). The OD data were calibrated by subtracting the background values of the blank medium. The time to reach OD₆₀₀ 0.05 (T_0.05), approximately 7 log CFU/ml, was then calculated for each well. For the indicator bacteria in wells containing CFS which did not reach OD₆₀₀ 0.05, the T_0.05 was recorded as 120 h, the duration of the incubation. The effect of CFS on the growth of an indicator strain was regarded as inhibitory if the T_0.05 of treatment (T_CFS) was larger than that of both MJM (T_MJM) and buffer (T_Buffer) controls. The inhibition strength was then calculated as (T_CFS – T_MJM)/T_MJM (55, 73).

Analysis of supernatants of effector Carnobacterium strains.

The CFS of each Carnobacterium effector strain grown in anaerobic MJM for 10 days was analyzed for pH, glucose, and organic acids, with uninoculated MJM included as controls. Two independent replicates were carried out for each measurement. The pH was measured using a portable pH meter (Accumet; Thermo Fisher Scientific, Ottawa, ON, Canada). Concentrations of glucose and lactate were determined using a glucose (HK) assay kit (Sigma, Oakville, ON, Canada) and a lactate colorimetric assay kit II (BioVision, Milpitas, CA, USA), respectively, following the manufacturers’ instructions. Formate was determined using an enzymatic method described previously with slight modification (74). Briefly, 300 μl of each CFS was mixed with 1.5 ml of 5.4% trichloroacetic acid (Thermo Fisher Scientific) and incubated at 0°C for 5 min. The acidified samples were centrifuged at 8,000 × g for 10 min at 4°C to remove proteins. Then, 600 μl of the supernatant was neutralized with 60 μl of 3M potassium carbonate (Sigma) and kept at 4°C for 16 h. For the enzymatic reaction, 240 μl of the treated sample or formate standard was mixed with 1.2 ml of phosphate buffer (pH 7.5; Hardy Diagnostics, Santa Maria, CA, USA), 30 μl of 0.05 M β-NAD (β-NAD sodium salt from Saccharomyces cerevisiae; Sigma), and 30 μl of formate dehydrogenase (20 units/ml; Sigma). The mixture was added into wells of a 96-well microtiter plate with 200 μl for each well and four wells for each sample or standard. The plate was incubated at 37°C for 20 min, and absorbance at 340 nm (A₃₄₀) was measured using the microtiter plate reader. The concentration of formate in each sample was calculated according to A₃₄₀, and a standard curve was generated using known concentrations of formate. The detection limit for glucose, lactate, and formate was 0.1 mM, 0.06 mM, and 0.3 mM, respectively. Concentrations of acetate, propionate, isobutyrate, butyrate, isovalerate, valerate, and caproate were determined using gas chromatography at the Department of Agricultural, Food, and Nutritional Science, University of Alberta, Edmonton, Canada, using conditions described previously (71).

Statistical analysis.

Analysis of variance (ANOVA) was performed using the general linear model (GLM) procedure in SAS version 9.4 (SAS Institute, Cary, NC) to compare the inhibition strength (broth assay) between the combinations of Carnobacterium spp. and indicator bacteria which produced an inhibition zone in the spot-lawn assay and those that did not. For the combinations that did not show a visible inhibition zone, ANOVA was performed to compare the inhibition strength of 15 Carnobacterium strains against 45 indicator bacteria. The difference in pH of spent medium, glucose consumption, and acetate and formate production among Carnobacterium strains was also assessed using ANOVA. Undissociated formic acid (UndisFA) and acetic acid (UndisAA) were each calculated as UndisFA = FA/[1 + 10^{(pH - 3.75)}] and UndisAA = AA/[1 + 10^{(pH - 4.76)}], respectively. The Pearson method was used to examine the correlation between the inhibition strength or pH of CFS of 15 Carnobacterium strains and production of formate, acetate, UndisFA, or UndisAA by them. A significance level of 0.05 was used for all statistical analyses.