Isolation and characterization of Corynebacterium spp. from bulk tank raw cow's milk of different dairy farms in Germany

We detected Corynebacterium spp. in raw milk samples of three farms by means of a selective, tellurite-containing medium. The isolated strains were identified based on full 16S rRNA gene sequences and partial rpoB gene sequences as C. xerosis, C. variabile, C. lactis, C. callunae, C. confusum, C. glutamicum and C. crudilactis. The identification based on 16S rRNA and rpoB sequences was not reliable for isolates of C. xerosis. Chemotaxonomic markers of the isolates, fatty acids, acyl type of peptidoglycan, presence and length of mycolic acids, quinone patterns, and polar lipids, were in accord with the known characteristics of these species. Biochemical profiles, analyzed with the API Coryne system, were able to differentiate all groups, but were unable to identify the strains due to an inappropriate database for raw-milk associated corynebacteria. Most of the tested isolates showed a single-substance resistance against oxacillin, but three single isolates were classified as multidrug resistant.


Introduction
Species of the genus Corynebacterium were found ubiquitously in the environment, although often their natural habitat-especially the habitat of nonmedical Corynebacterium speciesremains unknown [1]. In various studies, pathogenic corynebacteria were detected in raw milk samples and Corynebacterium spp. are known to cause subclinical mastitis in dairy cows [2]. Corynebacterium bovis is a common agent of bovine subclinical mastitis [3,4] and other species, e.g., C. amycolatum, C. minutissimum, C. ulcerans and C. pseudotuberculosis, were associated with clinical or subclinical bovine mastitis as well [5,6].
Non-pathogenic Corynebacterium species were also frequently isolated from raw milk or raw milk products [7,8]. Among them were also some species with beneficial functions in food processing. For example, the species C. glutamicum and C. variabile are well-known aminoacid producers [9] and the species C. casei, C. mooreparkense, C. ammoniagenes and C. stationis, have been detected on the surface of smear ripened cheese and are supposed to contribute to the flavor of the cheese [10,11].
The knowledge about corynebacterial diversity in raw milk is still fragmentary because of inappropriate routine test systems and high numbers of misidentifications [6,12]. For example, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

16S rRNA and rpoB gene sequencing
Extraction of genomic DNA, amplification and sequencing of 16S rRNA genes was performed as described previously [23,24]. 16S rRNA genes were amplified with the universal bacterial primers 8F (5´-AGAGTTTGATCMTGGC-3´) and 1492R (5´-TACCTTGTTACGACTT-3´) and as sequencing primers we used 787R (5´-GGACTACCAGGGTATCTAAT-3´) and 518F (5´-CCAGCAGCCGCGGTAAT-3´) [24]. Amplification of partial rpoB genes was performed with the Corynebacterium specific primers C2700F (5´-CGWATGAACATYGGBCAGGT-3´) and C3130R (5´-TCCATYTCRCCRAARCGCTG-3´) [15]. These primers were also used as sequencing primers. The obtained sequences were manually edited with Chromas Lite 2.1.1. (Technelysium Pty Ltd, South Brisbane, AU) and assembled with BioEdit 7.2.5. [25] to obtain either almost complete 16S rRNA sequences of 1,400-1,500 base pairs (bp) or partial rpoB gene sequences (300-400 bp). The sequences were compared to the sequences of type strains by using the Basic local alignment search tool (BLAST) [26]. Phylogenetic trees of isolates, closest related type strains and other raw milk associated Corynebacterium species were obtained by maximum-likelihood algorithm with MEGA 6.06. [27] and the alignment of the sequences was performed by ClustalW [28]. Model parameters were estimated using the "find best DNA" option of MEGA and models were chosen according to lowest Bayesian information criterion (BIC) and Akaike information criterion (AIC) values [27]. As best fit substitution model for 16S rRNA gene sequences, the Tamura-3-parameter model was chosen with discrete gamma distribution and presence for invariant sites. The rpoB gene sequences were analyzed based on the Tamura-Nei model with discrete gamma distribution and presence for invariant sites. The nearest-neighbor-interchange search method was used for both trees.

Chemotaxonomic and biochemical properties
Chemotaxonomic characteristics. Fatty acid patterns, presence and length of mycolic acids, polar lipid patterns, quinones and the acyl type of peptidoglycan were determined as described previously [23]. API Coryne test system. Biochemical tests were performed with the API Coryne test system (BioMèrieux, F) for identifying coryneform bacteria according to the manufacturer's specifications. Microorganisms were cultured on TSA for 24 h at 30˚C and the inoculated test strips were incubated for 24-48 h at 30˚C. Reaction profiles were analyzed with the apiweb TM software. The database version was V3.0.
Proteolytic and lipolytic activity. Proteolytic activity was determined on skim milk agar (TSA with 5.0% w/v skim milk powder) and lipolytic activity on tributyrin agar (TSA with 1.0% v/v tributyrin) [29]. Both tests were performed at 30˚C for 48 h and at 10˚C for 7 d. Strains were considered positive for proteolysis or lipolysis by formation of a transparent halo around the colonies.

Evaluation of different selective media
All reference strains showed intense growth on brain-heart infusion agar without supplements. The addition of potassium tellurite inhibited the growth of the Gammaproteobacteria strains Escherichia coli M3_I20, Acinetobacter guillouiae M3_I21 and Pseudomonas gessardii M3_I22 at all concentrations. Colonies of the Corynebacterium, Staphylococcus, Bacillus and Lactococcus strains were visible after two days of incubation at 30˚C. Earlier studies showed that Gramnegative species are especially sensitive against tellurite and its strongly oxidizing potential [31], whereas Gram-positive genera like Staphylococcus, Enterococcus and Corynebacterium are able to grow in the presence of tellurite [19,31]. The content of Gammaproteobacteria increases during the storage of raw milk in the bulk tank at approximately 4˚C [20]. These organisms usually grow fast on standard media and may overgrow Corynebacterium strains, but are inhibited by tellurite.
Strains grown on tellurite formed black and small colonies. Growth of the Corynebacterium, Staphylococcus, Bacillus and Lactococcus reference strains was weaker at 0.36 g/L potassium tellurite than at 0.25 g/L or 0.15 g/L. Therefore, 0.25 g/L tellurite was used as additive for the selective cultivation of Corynebacterium spp. Neither 0.1% nor 1.0% Tween 80 had an enhancing or inhibiting effect on the growth of the non-lipophilic Corynebacterium reference strains. Because the supplementation of Tween 80 is essential for cultivation of lipophilic species, it was added at a concentration of 1.0% to the selective agar.

Bacterial counts and isolation procedure
On BHT-agar, the mean number of bacterial counts in nine raw milk samples was 5.0 x 10 3 cfu/mL (colony-forming unit) with a range from 1.0 x 10 2 to 2.4 x 10 4 cfu/mL. The mean number of bacterial counts on TSA was 8.1 x 10 4 cfu/mL (range from 2.0 x 10 3 to 2.7 x 10 5 cfu/mL). The average ratio of bacterial counts on BHT-agar to total bacterial counts on TSA was 6.1% with a range from 1.2% to 165.0%. Colonies on BHT-agar were screened microscopically for a rod-shaped cell morphology and rods without endospores were subcultivated as presumptive Corynebacterium spp. These isolates (n = 68) were obtained from five raw milk samples taken from three dairy farms.

Identification of raw milk isolates
Raw milk associated Corynebacterium species. The isolates were identified based on their partial or full 16S rRNA gene sequences and partial rpoB gene sequences and out of 68 isolates, 28 were identified as Corynebacterium spp. The other isolates were members of the genera Brevibacterium, Microbacterium, Arthrobacter, Dietzia or Psychrobacillus. The Corynebacterium isolates were assigned to six different species: C. callunae, C. xerosis, C. variabile, C. confusum, C. glutamicum and C. lactis. One isolate, JZ16 T , could not be assigned to any of the known Corynebacterium species. For this strain, the new species C. crudilactis was proposed recently [32]. Corynebacterium isolates were recovered from raw milk of three different dairy farms; none were detected in the raw milk samples of the other four dairy farms. This confirms a report [33], in which Corynebacterium species were only detected in 25% of the analyzed raw milk samples. All of the identified Corynebacterium species in this study were recovered from raw milk before, except for C. callunae, and may be part of the natural raw milk microbiota [8,9,23]. C. xerosis was the only species isolated from raw milk of three different dairy farms and isolates of this species have frequently been recovered from raw cow´s milk [6,8]. As it is considered a commensal of the mammalian and bovine mucous membrane [34,35], it may contaminate raw milk as part of the cow's natural udder microbiota. The isolated Corynebacterium species are considered as non-pathogenic and an impact on human health is unlikely. Except for C. confusum, which is a rare human pathogen and was rarely isolated from clinical material [36,37], but no data is available considering the pathogenicity of C. confusum in animals or the potential of a human infection via zoonotic transmission.
16S rRNA and rpoB gene sequence analyses. Results of 16S and rpoB gene sequencing are given in Table 1 and GenBank Accession numbers of the gene sequences in S1 Table. Phylogenetic relationships of the isolates and type strains based on maximum-likelihood analysis of their 16S rRNA or rpoB gene sequences are shown in Figs 1 and 2. The identification of C. lactis, C. callunae, and C. confusum was reliably based on a pairwise similarity of at least 99.7% to the 16S rRNA gene sequences of the type strains. Additionally, the pairwise similarity to the gene sequences of the next related type strain did not exceed the proposed threshold for species delineation of 97.8% [38]. The isolates identified as C. variabile, C. xerosis, C. glutamicum and C. crudilactis were not reliably differentiated from their next relatives by 16S rRNA gene sequencing and the deviation of their16S rRNA gene sequences ranged from 0.1-2.1%. Partial rpoB gene sequence analyses were needed to reliably identify the isolates of C. variabile, C. glutamicum and C. crudilactis. Here, the pairwise similarity to the next related type strain was below 90.2% and complied with the proposed cutoff for species delineation of 95.0% [39]. However, C. xerosis and the most closely related species C. freneyi were hardly distinguishable by rpoB gene sequencing as well. The deviation of their rpoB gene sequences ranged between 2.5 and 4.5% and the average pairwise similarity was 95.4% (94.9-95.8%). This supports findings [15] that the rpoB gene sequence similarity of C. xerosis and C. freneyi is the highest among all Corynebacterium species, which are closely related. For C. xerosis and C. freneyi, restriction length polymorphism analysis of the 16S-23S spacer region has been proposed to clearly differentiate the two species [40]. Additionally, multilocus sequence analyses of several housekeeping genes, e.g. atpA, dnaA, fusA, odhA, or whole genome sequencing can be applied to improve the resolution of phylogenetic relationships between these closely related Corynebacterium species [41].

Chemotaxonomic and biochemical features of Corynebacterium isolates
Fatty acid pattern. All isolated Corynebacterium strains showed long-chain saturated and unsaturated fatty acids, as described for the genus Corynebacterium [1]. In contrast to other genera of the Corynebacteriales, the species of this genus contain no or low amounts of tuberculostearic acid [1]. Species-specific fatty acid patterns were detected for the seven identified Corynebacterium species ( Table 2). The species showed quantitative and qualitative differences among one another, especially in the presence of minor compounds. For example, C. lactis JZ6 was clearly separated from the other strains because it contained minor amounts of the fatty acids C 17:1 cis 9 (9.9%) and C 17:0 (17.1%) and the strains of C. variabile contained little to moderate amounts (2.3-15.3%) of the diagnostic compound tuberculostearic acid (TBSA; C 18:0 10-methyl). Only few Corynebacterium species contain moderate amounts of TBSA, e.g. C. variabile, C. ammoniagenes and C. bovis [1]. Traces of TBSA were detected for C. confusum as well https://doi.org/10.1371/journal.pone.0194365.g002 [36], but this could not be confirmed in this study. Fatty acid analyses may allow differentiation between C. xerosis and the closely related C. freneyi. According to Funke and Frodl [40], strains of C. freneyi contain little amounts of the unsaturated fatty acid C 17:1 cis 8, which was not detected for C. xerosis in this study (Table 2). Additionally, C. freneyi strains contain lower levels of C 18:1 cis 9 (21%) [40], compared to C. xerosis, where C 18:1 cis 9 was the main fatty acid (66.8-85.3%; Table 2). The strains of C. confusum contained large amounts (18.6-46.5%) of an unidentified component (ECL 16,697) that even had, in one case, a higher percentage than the main fatty acid C 18:1 cis 9. Mass spectra revealed these compounds as saturated and unsaturated aldehydes, which were presumptive pyrolysis products of the corynemycolic acids [42]. The fatty acid composition proved to be a useful feature for a differentiation of Corynebacterium species because each one of the seven different species showed a unique fatty acid profile. Additionally, the presence of diagnostic fatty acids, e.g. TBSA, which are only present in a few Corynebacterium species, enables a quick distinction between species.
Mycolic acids, acyl type of peptidoglycan, quinones, polar lipid pattern. A summary of the chemotaxonomic characteristics of representative isolates is given in Table 3. The isolated strains contained mycolic acids with a chromatographic mobility comparable to the mycolic acids of C. glutamicum DSM 20300 T , as described for members of these species [1], except for strain C. lactis JZ6. C. lactis belongs, together with the species C. amycolatum, C. caspium, C. ciconiae and C. kroppenstedtii, to the small group of corynebacteria without mycolic acids [1,23]. Mycolic acids are long-chained, saturated and unsaturated, β-hydroxy fatty acids with a long α-alkyl branch, characteristically synthesized by members of the order Corynebacteriales [43]. They vary in structure, chain length and in the degree of unsaturation between the different genera of the order Corynebacteriales and allow a differentiation of the genus Corynebacterium that characteristically shows short mycolic acids with 22-38 carbons [1]. All strains showed the acetyl type of peptidoglycan. Dihydrogenated menaquinones with nine isoprene units [MK-9 (H 2 )] were detected as major menaquinones (> 60%) for all of the strains, except for the strains of C. confusum, which showed MK-8 (H 2 ) as major menaquinone. The analysis of bacterial isoprenoid quinones and polar lipids is used for the characterization of corynebacteria and related genera [44,45,46]. The phospholipids diphosphatidylglycerol (DPG) and phosphatidylglycerol (PG) were detected in all of the strains. Phosphatidylinositol (PI) and phosphatidylinositol mannoside (PIM) were detected as well, but whereas PI was present in all strains, PIM was not detected in the strains of C. variabile. Aminolipids were not detected. This confirms data from earlier publications, where DPG, PG, PI and PIM were detected for C. xerosis, C. glutamicum and C. lactis [23,47,48]. Phospholipid patterns were not determined so far for C. variabile, C. confusum and C. callunae. PG and phosphatidylethanolamine (PE) are useful markers to differentiate corynebacteria from the related genera Mycobacterium, Nocardia, Gordonia and Rhodococcus [48]. PG was detected in substantial amounts only in Corynebacterium species and not in the genera Mycobacterium, Nocardia or Gordonia [48]. In contrast, PE is absent only in members of the genus Corynebacterium, but not in the other genera of the order Corynebacteriales [1,48]. While DPG is a common compound in bacteria, PI was only detected in Actinomycetes and Corynebacteriales. Not all of the polar lipids could be identified in this study. Some strains contained molybdenum blue negative lipids with a high mobility in the second dimension. According to literature, this is characteristic for acidic glycolipids [49] and they have been detected for C. xerosis and C. bovis as well [48]. Thin-layer chromatograms of mycolic acids and polar lipids are given in Figs 3 and 4. API Coryne. Identification of Corynebacterium species by the API Coryne test system is a fast and easy method and earlier studies showed that the numbers of misidentifications of clinical isolates are relatively low [50,51,52]. Results of the API Coryne test system for our isolates are shown in Table 4. The numerical code was obtained after an incubation period of 24 h. Test strips of the strains of C. xerosis and C. confusum were incubated 48 h because of their weak growth after 24 h. All of the tested strains were positive for pyrazinamidase and negative for gelatinase, pyrolidonyl arylamidase, N-acetyl-β-glucosaminidase and fermentation of glycogen and xylose. The numerical code was identical for the strains within the species C. variabile, C. callunae and C. confusum (Table 4), the three strains of C. xerosis showed different results in the API Coryne test. Two of the three tested strains of C. xerosis (JZ2 and JZ3) were negative for α-glucosidase activity and two strains (JZ3 and N1) were negative for alkaline phosphatase. This may also differentiate the isolated C. xerosis strains from the closely related C. freneyi. Strains of C. freneyi are described consistently positive for α-glucosidase activity and alkaline phosphatase [40]. None of the isolates were correctly identified by this test system, which is explained by the lack of these raw milk associated species in the present database version. Therefore, correct identification of milk-associated Corynebacterium species is critical with this system. Proteolytic and lipolytic activity. In order to determine the potential as food spoiling organisms, proteolytic and lipolytic activity of the Corynebacterium isolates were tested at 30˚C and 10˚C. None of the Corynebacterium strains showed proteolytic activity. Lipolytic activity was detected for the strains of C. lactis, C. callunae, C. xerosis and C. variabile at 30˚C and the strains of C. variabile were able to perform lipolysis at 10˚C as well. Although growth below 20˚C is rarely detected within the genus Corynebacterium [1], all raw milk isolates, except strains of the species C. confusum, were able to grow at 10˚C on TSA within 48 h.

Antimicrobial susceptibility of Corynebacterium isolates
Nine selected strains were tested for susceptibility against 16 different antimicrobial agents: C. lactis JZ6, C. callunae JZ13, C. glutamicum JZ15, C. crudilactis JZ16 T , C. variabile JZ20, C. confusum JZ27 and C. xerosis JZ2, JZ3 and N1. All of the strains were resistant against oxacillin. Additionally, C. callunae JZ13 was resistant against pirlimycin and C. glutamicum JZ15 against streptomycin. The three isolates C. xerosis N1, C. confusum JZ27 and C. crudilactis JZ16 T [32] were resistant against antimicrobial agents of three different classes, which qualified them as multidrug resistant (MDR) according to the definition of the CLSI [30]. C. xerosis N1 and C. confusum JZ27 were resistant against oxacillin, erythromycin and pirlimycin and C. crudilactis JZ16 T was resistant against oxacillin, ampicillin, thrimethoprim/sulfonamide, kanamycin and streptomycin [32]. Multidrug resistance has been described for clinically relevant Corynebacterium species, e.g. C. resistens, C. striatum or C. amycolatum, but rarely for non-medical corynebacteria [53,54,55]. Susceptibilities of raw milk associated Corynebacterium isolates against antimicrobial agents has not been described so far, expect for C. bovis and C. amycolatum isolates from bovine mammary glands [56]. Here, the Corynebacterium isolates were generally susceptible against the 15 tested antimicrobial agents. This supports findings of this study that antimicrobial resistance of raw milk associated corynebacteria and the potential to distribute antimicrobial resistance is generally low, except for single strains with high levels of antimicrobial resistance.

Conclusion
Results of this study confirm that Corynebacterium species are a minor but regular part of the raw milk microbiome. Additionally, strains of other genera of the phylum Actinobacteria were isolated from raw milk on the selective medium in this study, e.g. Arthrobacter and Dietzia, which are rarely described as raw milk associated bacteria. Abundance and impact on raw milk of these organisms may need further investigation. A delineation of Corynebacterium spp. from other closely related genera of the order Corynebacteriales is possible by chemotaxonomic markers. Some of these markers, especially fatty acid profiles or the presence or absence of mycolic acids could also be used to differentiate between several milk-associated species within this genus. Sequencing of the 16S rRNA gene is appropriate for the identification of most Corynebacterium strains. For some species, a reliable identification needs additional sequence information from a less conserved gene like the rpoB gene, but for some closely related species, like C. xerosis and C. freneyi, additional tests (fatty acid pattern, α-glucosidase activity and alkaline phosphatase) are highly recommended. The high physiological diversity within the genus Corynebacterium, covering amino-acid producers, colonizers of smear-ripened cheese but also animal and human pathogens, gives reasons for further in-depth analyses of raw milk associated Corynebacterium species.
Supporting information S1 Table. Accession numbers of the sequences used in this study. (PDF)