Twenty-five vaginal LAB isolates with clear acidification halos on MRS agar supplemented with 0.5% CaCO₃ were recovered. All were Gram-positive rods, catalase negative, and γ-hemolytic. They were subsequently identified at the species level by 16S rRNA gene sequencing and sequence comparison. Seven isolates were identified as belonging to Lactobacillus crispatus, six to Lactobacillus salivarius, four to Lactobacillus jensenii, four to Lactobacillus paracasei, two to Lactobacillus reuteri, and two to Bifidobacterium bifidum (Table 1). A fingerprinting analysis by combining the results of RAPD-PCR and rep-PCR techniques detected 21 different strains among the 25 isolates, distributed as follows: L. crispatus (six strains), L. salivarius (five), L. jensenii (three), L. paracasei (three), L. reuteri (two), and B. bifidum (two strains) (Supplementary Figure S1).

Table 1 shows the MIC values of the 16 tested antibiotics for the 25 vaginal LAB isolates. All isolates were phenotypically susceptible to tetracycline, erythromycin, clindamycin, penicillin, quinupristin-dalfopristin, linezolid, and rifampicin. The distribution of neomycin MICs covered more than nine 2-fold dilutions, ranging from ≤0.5 to 256 µg mL⁻¹. Similarly, a wide distribution of MICs was observed for streptomycin (≤0.5 to >256 µg mL⁻¹), ciprofloxacin (≤0.25 to 64 µg mL⁻¹), and chloramphenicol (1 to 8 µg mL⁻¹). The most common resistance phenotypes observed were those to trimethoprim (MIC 16 to ≥64 µg mL⁻¹) and vancomycin (MIC ≥ 128 µg mL⁻¹). Nine isolates were resistant to kanamycin (MIC values 32 to >1024 µg mL⁻¹). The two B. bifidum strains were susceptible to all tested antibiotics except for streptomycin (MIC >256 µg mL⁻¹). Moderate resistance to chloramphenicol was seen in three isolates and for ampicillin in one isolate. Interestingly, L. salivarius VA40-10 was highly resistant to all four aminoglycosides tested (gentamicin, kanamycin, streptomycin, and neomycin).

The presence of genes coding for the commonest AR genes spread among LAB was investigated by PCR. No genes involved in resistance to chloramphenicol (cat), β-lactams (bla), aminoglycosides [aac(6′)-aph(2″) and aad(E)], macrolides [erm(A), erm(B), erm(C), erm(F), mef(A)], tetracycline [tet(M), tet(O), tet(S), tet(K), tet(L)], clindamycin (lsaA) or vancomycin (vanA) were ever detected in any of the isolates (data now shown). In contrast, PCR analyses for genes encoding ribosomal protection proteins (RPP) causing tetracycline resistance using the degenerate primer pairs DI-DII and Tet1-Tet2 (Supplementary Table S1) produced a positive amplification when DNA from two tetracycline-susceptible B. bifidum isolates was used as a template (Figure 1A). Amplification with gene-specific primers gave a positive result only for the tet(W) gene (Figure 1B). Amplicon sequencing and sequence comparison further proved the presence in these isolates of a tet(W) gene highly homologous to those present in many Gram-positive and Gram-negative bacteria.

Based on the phenotype and genotype results (Table 1; Supplementary Figure S1), six strains were subjected to genome sequencing: L. crispatus VA50-4AN (resistant to kanamycin, ampicillin, and trimethoprim), L. jensenii VA04-2AN (resistant to trimethoprim), L. salivarius VA40-10 (resistant to gentamicin, kanamycin, streptomycin, neomycin, and vancomycin), L. paracasei VA02-1AN (resistant to chloramphenicol and vancomycin), L. reuteri VA24-5 (resistant to vancomycin and trimethoprim), and B. bifidum VA07-1AN (resistant to streptomycin). Supplementary Table S2 shows the general features of their genomes. Their size was, in all cases, around 2.2 Mbp but the number of contigs obtained after assembly ranged from 17 to 300. Supplementary Table S3 summarizes some of the key genetic features of the genomes of the sequenced strains. Genes coding for penicillin binding proteins (PBP) were found in all the genomes, although with different numbers and types for the distinct species. Mutations in PBPs encoding-genes known to confer AR were not identified. One gene coding for a D-alanine-D-alanine ligase (Ddl) was detected in each of the strains. In several LAB species, the presence of phenylalanine at the enzyme active site in Ddl is correlated with intrinsic resistance to vancomycin ^[2]. In addition, in each of the strains, 9-32 genes were classified by the RAST server as belonging to the category “Virulence, Disease, and Defence”, subcategory “Resistance to Antibiotic and Toxic Compounds”. The majority of these genes encoded components dedicated to homeostasis or resistance to heavy metals, such as copper, mercury, and the cobalt-zinc-cadmium triad. Genes encoding elongation factors, efflux pumps, DNA gyrases, and topoisomerases were also included by RAST in this subcategory.

By comparing the genome sequences against the databases CARD, ResFinder, and ARG-ANNOT, no genes known to be involved in AR in L. jensenii VA04-2AN (resistant to trimethoprim), L. paracasei VA02-1AN (resistant to chloramphenicol and vancomycin), and L. reuteri VA24-5 (resistant to trimethoprim and vancomycin) were detected. The only positive correlation between phenotype and genotype was the presence of a conserved phenylalanine (F) residue in the active site of the Ddl ligase, corresponding to amino acid 261 of the Leuconostoc mesenteroides enzyme ^[2], in the deduced sequence of all vancomycin-resistant (Vm^r) strains, while the susceptible (Vm^s) strains were characterized by the presence of a tyrosine (Y) residue at this position (Figure 2).

Genome analysis of L. crispatus VA50-4AN, L. salivarius VA40-10, and B. bifidum VA07-1AN identified no genes known to be involved in aminoglycoside resistance. Therefore, mutations in key genes, such as those coding for the ribosomal S12 protein and others acting on the 16S rRNA molecule, were therefore sought by comparing the DNA and deduced protein sequences from our strains with those in databases. No amino acid differences were observed in the sequences of the ribosomal protein S12 for L. crispatus VA50-4AN and L. salivarius VA40-10 from those belonging to susceptible strains of the same species. Further, alignment of the deduced amino acid sequences for the 16S rRNA guanine(527)-N(7)-methyltranferase (RsmG) proteins of the sequenced strains, showed heterogeneity at several positions between themselves and with respect to sequences in databases. In particular, the RsmG sequence of L. crispatus VA50-4AN showed one amino acid change at position 38 (N→H), while that of B. bifidum VA07-1AN showed three amino acid changes at positions 105 (E→A), 150 (G→D), and 206 (R→G), and that of L. salivarius VA40-10 showed six exclusive amino acid changes at positions 12 (G→E), 67 (D→N), 186 (N→D), 199 (I→V), 208 (Q→K), and 209 (V→I). However, by comparing RsmG sequences from resistant and susceptible strains, none of the changes considered could be associated with streptomycin resistance.

As expected, the genome analysis confirmed the presence of tet(W) in B. bifidum VA07-1AN; this gene was also unequivocally identified by searches in the three AR databases used. The tet(W) gene in B. bifidum VA07-1AN was located in a contig of 76,748 bp. Figure 3 shows the genetic organization of the 40-kbp left extreme of the contig that included the tet(W) gene. The tet(W) sequence of VA07-1AN (1922 bp) was almost identical to that described for Bifidobacterium longum LTBL16 (CP034089.1). Similar tet(W) sequences have also been found in the chromosome of strains belonging to other species such as B. bifidum L22 (NG_048301.1), Lachnospiraceae bacterium KGMB03038 (CP041667.1), and Ruminococcus sp. JE7A12 (CP039381.1); and in plasmids, such as pTZC1 from Cutibacterium acnes TP-CU389 (LC473083.1). Compared to tet(W) in B. longum LTBL16, the tet(W) in B. bifidum VA07-1AN contained an insertion of two extra guanine residues (GG) after nucleotide 731 in the ORF resulting in a frameshift, which produced only a short peptide—289 amino acids long compared to 639 residues for the functional Tet(W). This likely explains the susceptibility of VA07-1AN to tetracycline. The tet(W) gene was flanked by ORFs coding for proteins showing the greatest homology to others from B. longum in the upstream region, and proteins typical of B. bifidum in the downstream region (Supplementary Table S4).

The CARD database further identified in the genome of B. bifidum VA07-1AN a single nucleotide polymorphism (SNP) point mutation in the rpsL gene (encoding the ribosomal S12 protein), a variation causing an amino acid substitution (K→R) at position 43 of the protein (Figure 4). This amino acid change has been associated with streptomycin resistance in many species ^[3].

When the susceptibility of B. bifidum VA07-1AN to tetracycline was assayed using the MICE test, colonies growing within the inhibition halo were noted. Identification and typing showed them to be tetracycline-resistant variants of VA07-1AN. After plating on antibiotic-containing and antibiotic-free plates, about 0.6% of the colonies from an overnight culture were found to be tetracycline-resistant. Amplification and sequencing of tet(W) genes from 13 tetracycline-resistant variants showed the addition of one guanine nucleotide in most revertants to the guanine stretch where the two Gs disrupting the ORF had been inserted (as in R-1; Figure 5). Other mutations consisting of both nucleotide insertions and deletions in the vicinity of the stretch of Gs were occasionally seen (as in R-11; Figure 5). In either case, there was a net gain of one nucleotide, which, together with the two Gs that disrupted the tet(W) ORF, created a new codon that opened the reading frame of Tet(W) producing a functional protein that provided tetracycline resistance. The MIC of tetracycline in the tetracycline-resistant variants ranged from 48 to 96 μg mL⁻¹. In contrast, growing the antibiotic-resistant variants in the absence of tetracycline for about 80-100 generations showed no tetracycline-susceptible revertants, demonstrating high stability of the mutations that restored the resistant phenotype.