Lactic acid bacteria and bifidobacteria deliberately introduced into the food chain may act as a reservoir of antimicrobial resistance genes (ARGs), which is considered a safety concern. In the present study, resistance to antimicrobials of commercial probiotic strains, probiotic candidate strains, and starter cultures (n = 20) was characterised based on integration of phenotypic and in silico data. Minimum inhibitory concentrations (MICs) of 16 antimicrobials were determined for lactobacilli and bifidobacteria that were isolated from pharmaceutical products or obtained from the manufacturers or culture collections. Using different databases and bioinformatic tools, we predicted ARGs, mutations, genomic islands, and mobile genetic elements (MGEs) in their whole genome sequences. In addition, a comprehensive in silico analysis of the prevalence of the tetW gene and its genetic environment across lactobacilli and bifidobacteria (n = 1423) was conducted. Several strains exhibited phenotypic resistance to kanamycin, tetracycline, chloramphenicol, quinupristin-dalfopristin, ciprofloxacin, or neomycin. These resistances, however, did not always correspond to the presence of ARGs and vice versa. We detected an acquired tetW gene in four commercial strains of Bifidobacterium animalis subsp. lactis, whereas homologs of antimicrobial resistance (AR) proteins were predicted in all 20 proteomes. The prevalence of the tetW gene, which was often flanked by MGEs, was higher in analysed bifidobacteria (31.9%) than lactobacilli (6.3%). In addition, sequences flanking tetW were associated with putative genomic islands and were conserved in several strains, including potential pathogens. Our findings provide an insight into AR of probiotics, probiotic candidates, and starter cultures with an emphasis on tetracycline and into the safety of these strains in the context of AR.

Two novel strains AV382 and AV436 were isolated from a submerged industrial bioreactor for production of apple cider vinegar in Kopivnik (Slovenia). Both strains showed very high (≥98.2%) 16S rRNA gene sequence similarities with Komagataeibacter species, but lower 16S–23S rRNA gene internal transcribed spacer (ITS). The highest similarity of the 16S–23S rRNA gene ITS of AV382 was to Komagataeibacter kakiaceti LMG 26206^T (91.6%), of AV436 to Komagataeibacter xylinus LMG 1515^T (93.9%). The analysis of genome sequences confirmed that AV382 is the most closely related to K. kakiaceti (ANIb 88.2%) and AV436 to K. xylinus (ANIb 91.6%). Genome to genome distance calculations exhibit for both strains ≤47.3% similarity to all type strains of the genus Komagataeibacter. The strain AV382 can be differentiated from its closest relatives K. kakiaceti and Komagataeibacter saccharivorans by its ability to form 2-keto and 5-keto-D-gluconic acids from glucose, incapability to grow in the presence of 30% glucose, formation of C_19:0 cyclo ω8c fatty acid and tolerance of up to 5% acetic acid in the presence of ethanol. The strain AV436 can be differentiated from its closest relatives K. xylinus, Komagataeibacter sucrofermentans, and Komagataeibacter nataicola by its ability to form 5-keto-D-gluconic acid, growth on 1-propanol, efficient synthesis of cellulose, and tolerance to up to 5% acetic acid in the presence ethanol. The major fatty acid of both strains is C_18:1ω7c. Based on a combination of phenotypic, chemotaxonomic and phylogenetic features, the strains AV382^T and AV436^T represent novel species of the genus Komagataeibacter, for which the names Komagataeibactermelaceti sp. nov. and Komagataeibacter melomenusus are proposed, respectively. The type strain of Komagataeibacter melaceti is AV382^T (= ZIM B1054^T = LMG 31303^T = CCM 8958^T) and of Komagataeibacter melomenusus AV436^T (= ZIM B1056^T = LMG 31304^T = CCM 8959^T).

The Shine-Dalgarno (SD) sequence is a key element directing the translation to initiate at the authentic start codons and also enabling translation initiation to proceed in 5′ untranslated mRNA regions (5′-UTRs) containing moderately strong secondary structures. Bioinformatic analysis of almost forty genomes from the major bacterial phylum Bacteroidetes revealed, however, a general absence of SD sequence, drop in GC content and consequently reduced tendency to form secondary structures in 5′-UTRs. The experiments using the Prevotella bryantii TC1-1 expression system were in agreement with these findings: neither addition nor omission of SD sequence in the unstructured 5′-UTR affected the level of the reporter protein, non-specific nuclease NucB. Further, NucB level in P. bryantii TC1-1, contrary to hMGFP level in Escherichia coli, was five times lower when SD sequence formed part of the secondary structure with a folding energy -5,2 kcal/mol. Also, the extended SD sequences did not affect protein levels as in E. coli. It seems therefore that a functional SD interaction does not take place during the translation initiation in P. bryanttii TC1-1 and possibly other members of phylum Bacteroidetes although the anti SD sequence is present in 16S rRNA genes of their genomes. We thus propose that in the absence of the SD sequence interaction, the selection of genuine start codons in Bacteroidetes is accomplished by binding of ribosomal protein S1 to unstructured 5′-UTR as opposed to coding region which is inaccessible due to mRNA secondary structure. Additionally, we found that sequence logos of region preceding the start codons may be used as taxonomical markers. Depending on whether complete sequence logo or only part of it, such as information content and base proportion at specific positions, is used, bacterial genera or families and in some cases even bacterial phyla can be distinguished.