Eleven M. anserisalpingitidis clinical isolates were selected for the examinations based on differences in their phenotypic and genetic characteristics which has been previously described (Grózner et al., 2021; Table 1). In brief, the isolates were cultured in Oxoid Mycoplasma broth medium (Thermo Fisher Scientific, Inc./Oxoid, Inc., Waltham, MA) supplemented with Mycoplasma Supplement G (Thermo Fisher Scientific, Inc./Oxoid, Inc.), 0.5% (wt/vol) sodium pyruvate, 0.5% (wt/vol) glucose, 0.15% L-arginine hydrochloride and 0.05% (wt/vol) phenol red. MIC values of enrofloxacin (group of fluoroquinolones), lincomycin (from lincosamides), tiamulin (from pleuromutilins) and tilmicosin, tylosin, tylvalosin from the group of macrolides were determined with the broth micro-dilution method (Hannan, 2000). Based on the results of the MIC testing, three strains with high and three isolates with low MICs were chosen for each antimicrobial group for further analyses. Overall, six strains exhibited reduced antibiotic susceptibility to multiple widely used antibiotics: MycAv 47, 70 and 91 for enrofloxacin; MycAv 47, 67, and 68 for lincomycin and tiamulin; MycAv 67, 68, and 177 for macrolides. Additionally, five strains showed sensitivity to several antimicrobial agents: MycAv 63, 65, and 668 for enrofloxacin; MycAv 55, 63, and 668 for lincomycin; MycAv 50, 55, and 668 for tiamulin and macrolides (Table 1). The low and high MIC values were determined based on previous studies, as there are no official breakpoints available for veterinary mycoplasmas: low MIC value ≤0.5 μg/mL for enrofloxacin (Landman et al., 2008), tylvalosin (Behbahan et al., 2008; Ghaleh Gol et al., 2008); ≤2 μg/mL for lincomycin (Kempf et al., 1988); ≤1 μg/mL for tylosin (Landman et al., 2008); ≤8 μg/mL for tilmicosin (Landman et al., 2008); <0.625 μg/mL for tiamulin (Ghaleh Gol et al., 2008); high MIC value: ≥2 μg/mL for enrofloxacin (Landman et al., 2008); tylvalosin (Behbahan et al., 2008; Ghaleh Gol et al., 2008); ≥8 μg/mL for lincomycin (Kempf et al., 1988); ≥4 μg/mL for tylosin (Landman et al., 2008); ≥32 μg/mL for tilmicosin (Landman et al., 2008); ≥0.625 μg/mL for tiamulin (Ghaleh Gol et al., 2008). Whole genome sequences of the isolates were determined on Illumina platform (Illumina, Inc., San Diego, CA, United States) and multilocus sequence typing (MLST) was performed for the genetic characterization of the isolates previously (Grózner et al., 2021). The selected isolates belong to different clades based on MLST analysis (Table 1; Grózner et al., 2021).

A total of 18 in vitro cultivated mutants were selected by serial passages from field isolates inhibited by lower MIC values for the examined antibiotics (Sulyok et al., 2017). The parent strains, the applied antibiotics and the 18 in vitro cultivated mutants can be seen in Supplementary Dataset S1. Briefly, two-fold dilution series were made from each antibiotic in sub-inhibitory concentrations for every selected parent strain with low MICs (listed in Table 1 and Supplementary Dataset S1). Enrofloxacin, lincomycin, tiamulin, tilmicosin, tylosin were produced by Vetranal^® Sigma-Aldrich, USA and tylvalosin was originated from ECO^® Animal Health Ltd., United Kingdom. Stock solutions were diluted in the concentration of 1 mg/mL and aliquots were stored at −70°C. Two-fold dilutions of each antibiotic were freshly prepared in Oxoid Mycoplasma broth medium (Thermo Fisher Scientific) in the range of 0.039–0.156 μg/mL for enrofloxacin and tiamulin, 0.125–0.5 μg/mL for lincomycin and macrolides in the first dilution panel. The selected isolates were incubated in Oxoid Mycoplasma broth medium (Thermo Fisher Scientific) at 37°C for 5 days, and the M. anserisalpingitidis cultures changing color (the red media turned yellow) at the highest antibiotic concentrations were inoculated into a fresh, two-fold dilution panel of the antibiotics, with increased concentrations. The serial passages were continued until the strains’ MICs reached above the sensitive MIC value determined by previous studies: ≥0.5 μg/mL for tylvalosin (Ghaleh Gol et al., 2008) and enrofloxacin (Landman et al., 2008), ≥1 μg/mL for tylosin (Landman et al., 2008), ≥8 μg/mL for tilmicosin (Landman et al., 2008), ≥2 μg/mL for lincomycin (Kempf et al., 1988), and ≥0.625 μg/mL for tiamulin (Ghaleh Gol et al., 2008). However, two (MycAv 63 and 65) of the three selected parent strains’ (MycAv 63, 65, 668) MIC values for enrofloxacin were already above 0.5 μg/mL, hence it was decided to set a higher threshold of 2 μg/mL for this antibiotic (Landman et al., 2008). In order to reach these MIC values, the sensitive M. anserisalpingitidis parent strains (MycAv 50, 55, 63, 65, 668) were submitted to four, increasing dilution series of enrofloxacin, lincomycin, tiamulin, tilmicosin, tylosin, and five, increasing dilution series of tylvalosin supplied liquid medium. In vitro developed mutants were passaged once in antibiotic-free medium and the MIC values were determined again (Supplementary Datasets S1,S2).

Antimicrobial susceptibility was measured by using broth micro-dilution method as described by Hannan (2000). For the MIC testing, twofold dilutions of each antibiotic were freshly prepared in the range of 0.039–10 μg/mL for enrofloxacin and tiamulin and 0.25–64/128 μg/mL for lincomycin, tilmicosin, tylosin and tylvalosin.

M. anserisalpingitidis type strain (ATCC BAA-2147) and five field isolates (the parent strains of the in vitro cultivated mutant strains, MycAv 50, 55, 63, 65, 668; Table 1) were used as control and the cultures were adjusted in Oxoid Mycoplasma broth medium (Thermo Fisher Scientific) to 10⁵ color changing unit (CCU)/mL. The experiment was carried out on 96-well plate and every experiment was performed in duplicates. The plates were incubated at 37°C and were checked every day for a week or until they did not show further color changes (red to yellow shift) for 2 days. Initial MICs were measured when the growth control changed color.

Broth micro-dilution assay (Hannan, 2000) was used in order to determine the sensitivity of M. anserisalpingitidis to CCCP, OV, and RSP. The efflux pump inhibitors were produced by Vetranal^® Sigma-Aldrich, United States. Two-fold dilutions of each efflux pump inhibitor were freshly prepared in the range of 0.191–97.92 μg/mL for CCCP, 0.07–35.84 mM for OV and 0.005–1.28 mg/mL for RSP. M. anserisalpingitidis cultures were adjusted in Oxoid Mycoplasma broth medium (Thermo Fisher Scientific) to 10⁵ CCU/mL. The experiment was carried out on 96-well plate and every experiment was performed in duplicates. The plates were incubated at 37°C and were checked every day until they did not show further color changes (red to yellow shift). The lowest efflux pump inhibitor concentration at which no color change was detected was considered the lethal dose of the efflux pump. The adequate concentrations of the efflux pump inhibitors to be used in the MIC testing were determined to be at least four-fold below of the lethal concentrations of each efflux pump inhibitors (Raherison et al., 2002; Li et al., 2017).

A modified broth micro-dilution method was used to examine the effect of efflux pump inhibitors (Hannan, 2000; Antunes et al., 2015; Li et al., 2017). The MIC values against the field isolates (MycAv 47, 70, 91 for enrofloxacin; MycAv 47, 67, 68 for lincomycin and tiamulin, MycAv 67, 68, 177 for macrolides) and in vitro cultivated mutant strains with decreased antibiotic susceptibility (MycAv 63-M1, 65-M1, 668-M1 for enrofloxacin, MycAv 55-M2, 63-M2, 668-M2 for lincomycin, MycAv 50-M6, 55-M6, 668-M6 for tiamulin, MycAv 50-M3, 55-M3, 668-M3 for tilmicosin, MycAv 50-M5, 55-M5, MycAv 668-M5 for tylosin, MycAv 50-M4, 55-M4, 668-M4 for tylvalosin) were measured for each antibiotic both in the presence and in the absence of efflux pump inhibitors. A modified dilution series was applied where 10% decrease steps were used instead of two-fold dilutions in order to make a more sensitive system, except for the tests with CCCP (Supplementary Dataset S2). In this case, two-fold dilutions were applicable (applied antibiotic ranges were the same as the ones used for MIC testing of the control strains) due to the strong effect on MIC values which had been experienced during the pilot studies (data not shown). The efflux pump inhibitors were used in the following concentrations: 0.765 μg/mL CCCP, 0.56 mM OV and 0.02 mg/mL RSP. The efflux pump inhibitors were freshly prepared before every experiment and every test was carried out on 96-well plate. The M. anserisalpingitidis type strain (ATCC BAA-2147, GenBank accession number CP042295) was used as a quality control. M. anserisalpingitidis cultures were diluted in Oxoid Mycoplasma broth medium in the concentration of 10⁵ CCU/mL. The plates were incubated at 37°C and checked after 16, 20, 24, 40, 44, 48, 64, 68, and 72 h of incubation, then on a daily basis for 1 week, or until they did not show further color changes for 2 days. Each experiment was performed in duplicates. Initial MICs were measured when the growth control showed color change.

Two-sample paired (Wilcoxon) signed rank test was used to ascertain whether there is significant difference between MIC values in the presence and absence of each tested efflux pump inhibitor. The (Wilcoxon) signed rank test was performed by R 4.2.1. software (R Core Team, 2022).

Genomic DNA of the 18 in vitro cultivated mutants (listed in Supplementary Dataset S1) were extracted from 2 mL of logarithmic-phase cultures with the help of a commercial kit (QIAamp DNA Mini Kit, Qiagen Inc., United States) according to the manufacturer’s instructions. The whole-genome sequencing was carried out by using NextSeq 500 Illumina equipment (Illumina, Inc., San Diego, CA, United States). The short reads were assembled using SPAdes software version 3.15.4 (Bankevich et al., 2012). The annotation of the genomes was done using Prokaryotic Genome Annotation Pipeline (PGAP) software version 6.3 (Li et al., 2021). The newly assembled genome sequences (Accessibility of the SRA reads: BioProject ID: PRJNA912395) and the field isolates with high MICs were compared to the parent strains of the in vitro developed M. anserisalpingitidis mutants, and to the type strain (ATCC BAA-2147) with progressive MAUVE algorithm (Darling et al., 2004). SNPs, deletions and insertions were searched by Geneious Prime 2019.2.1 software (Kearse et al., 2012). The putative SNPs were manually reviewed and sorted based on their capability to cause synonymous or non-synonymous SNP (nsSNP). The functions of those genes which contain nsSNPs were examined with the help of CARD database (Alcock et al., 2019), NCBI nucleotide database (Sayers et al., 2022) and current literature (Google Scholar, using the following key words: “antibiotic resistance” and the gene name or functional group, for example: “gyrA” “antibiotic resistance”). The common nsSNPs were checked in all M. anserisalpingitidis strains with publicly available whole genome sequences and MIC data (Grózner et al., 2022).

Putative regulatory sequences were identified using the Promotech software version 1.0 (Chevez-Guardado and Peña-Castillo, 2021) and the genome of the M. anserisalpingitidis type strain. In order to minimize the number of false positive promoters, the score threshold was suggested to be above 0.6 (Weber et al., 2012; Siqueira et al., 2014; Lloréns-Rico et al., 2015). As the regulatory regions of mycoplasmas are not fully characterized, we established a more stringent threshold for our investigation. Sequences located upstream of an efflux pump coding gene with a score of 0.8 or higher were considered potential regulatory regions involved in efflux pump regulation. The putative regulatory sequences’ promoter regions and ribosome-binding sites were determined based on the literature (Lloréns-Rico et al., 2015; Fisunov et al., 2016; Montero-Blay et al., 2019; Yus et al., 2019) and MEME suite software (Bailey et al., 2015). The Promotech-detected potential efflux pump regulatory regions were used as the input for the MEME motif-finding software to identify the putative Pribnow boxes in the sequences.

In this study initial MIC values of 18 in vitro cultivated mutants (MycAv 63-M1, 65-M1, 668-M1 for enrofloxacin, MycAv 55-M2, 63-M2, 668-M2 for lincomycin, MycAv 50-M6, 55-M6, 668-M6 for tiamulin, MycAv 50-M3, 55-M3, 668-M3 for tilmicosin, MycAv 50-M5, 55-M5, MycAv 668-M5 for tylosin, MycAv 50-M4, 55-M4, 668-M4 for tylvalosin) and six isolates with decreased antibiotic susceptibility (MycAv 47, 70, 91 for enrofloxacin; MycAv 47, 67, 68 for lincomycin, tiamulin, MycAv 67, 68, 177 for macrolides) were examined in the presence and in the absence of efflux pump inhibitors (Supplementary Dataset S2 and Figures 1, 2). In case of CCCP notable increases were detected in the efficiency of the antibiotics using twofold dilution series.

The MICs for enrofloxacin were halved in all cases out of the three field isolates tested (MycAv 47, 70, 91; Figure 1A); however, the in vitro cultivated mutants (MycAv 63-M1, 65-M1, 668-M1) did not show any detectable changes (Figure 2A). The MICs against the in vitro mutants with decreased lincomycin susceptibility (MycAv 55-M2, 63-M2, 668-M2) were reduced at least two-fold below (Figure 2B and Supplementary Dataset S2), while there could not be measured any changes in the MICs of clinical isolates of low susceptibility to this agent (MycAv 47, 67, 68 with >128 μg/mL MICs; Figure 1B). The MICs of in vitro mutants (MycAv 50-M6, 55-M6, 668-M6) and field isolates (MycAv 47, 67, 68) with decreased susceptibility to tiamulin dwindled at least two-fold below the original values, except in case of MycAv 68 clinical isolate with 0.625 μg/mL original MIC value (Figure 1C and Figure 2C). Regarding field isolates with increased MIC values for macrolides (MycAv 67, 68, 177), two-fold MIC decreases could be seen using CCCP in three cases: MIC value of MycAv 177 against tylosin and MycAv 177, 68 against tylvalosin. However, clinical isolates with extremely high MICs (MycAv 177, 67 with >128 μg/mL MIC for tilmicosin and MycAv 68 with >128 μg/mL MICs for tilmicosin and tylosin) did not show any changes (Figures 1D–F). Additionally, there was a field isolate (MycAv 67 in cases of tylosin and tylvalosin) where we could see differences only in modified dilution series (Supplementary Dataset S2). In cases of in vitro cultivated mutants with decreased susceptibility to macrolides, there were two- (MycAv 50-M3, MycAv 55-M3, MycAv 50-M4, MycAv 55-M4, MycAv 668-M4, MycAv 50-M5, MycAv 55-M5, MycAv 668-M5) or four-fold (MycAv 668-M3) reductions in the MIC values (Figures 2D–F and Supplementary Dataset S2). The difference was statistically significant between the MICs in the presence and absence of CCCP (in twofold dilution series) when all the mutants and field isolates were analyzed together (p < 0.001).

In the presence of OV or RSP MIC value differences could be seen only if modified dilution series were used (10% decrease steps were used instead of two-fold dilutions). There were detectable changes in 24 occasions out of the 36 examined combinations with OV (Supplementary Dataset S2), but these differences were not statistically significant. As for RSP, there were nine detectable, statistically not significant differences out of the 36 combinations (Supplementary Dataset S2).

The 18 in vitro cultured mutants (listed in Supplementary Datasets S1, S2, BioProject ID: PRJNA912395) were compared to their parent strains (listed in Table 1 and Supplementary Dataset S1) to identify novel mutations associated with lower antibiotic susceptibility (Supplementary Datasets S3, S4). Furthermore, MycAv 47, 67, 68, 70, 91, and 177 (M. anserisalpingitidis strains with high MICs for several antibiotics, listed in Table 1) were compared to the M. anserisalpingitidis type strain (ATCC BAA-2147) and the sensitive parent strains (MycAv 50, 55, 63, 65, 668; listed in Table 1 and Supplementary Datasets S3, S4). ABC family genes that exhibited shared mutations in clinical isolates (or nsSNPs observed in both in clinical and in vitro mutants) were analyzed across all available M. anserisalpingitidis strains with publicly accessible whole genome sequences and MIC data (Supplementary Dataset S5).

Gene function predictions, which were based on M. anserisalpingitidis type strain (ATCC BAA-2147) gene annotation (GenBank accession number CP042295), showed that most SNPs were located in hypothetical proteins and intergenic regions. Numerous mutations were identified within genes that were previously described as antibiotic resistance-related genes. These genes encode ABC family members (Supplementary Dataset S3), 23S ribosomal RNA and a topoisomerase coding gene (Supplementary Dataset S4).

In the present study, numerous nsSNPs were identified in several members of the ABC family, which encompass diverse functions. Clinical isolates with high MIC values (MycAv 47, 67, 68, 70, 91, 177) exhibited nsSNPs within 14 genes that belong to the ABC family: WP_146368471.1, WP_146368786.1, WP_226364443.1, WP_146368361.1, WP_146368222.1, QDY87225.1, WP_146368766.1, WP_146368557.1, WP_146368360.1, WP_146368785.1, WP_146368574.1, WP_146368765.1, WP_146368337.1, WP_146368347.1 (Supplementary Dataset S3). The functional classification of the proteins was determined through in silico methods, utilizing Hidden Markov Model (HMM) or Subfamily Protein Architecture Labeling Engine (SPARCLE; Lu et al., 2020) employed by PGAP. Among these genes, WP_146368786.1, WP_146368785.1, and WP_146368557.1 are ABC transporter ATP-binding proteins that potentially serve as ATPase and permease components of an ABC-type multidrug transport system. WP_146368766.1 is an ABC transporter ATP-binding protein, likely functioning as the catalytic subunit of an ATP transporter complex involved in macrolide export. QDY87225.1 and WP_146368347.1 represent ATP-binding cassette domain-containing proteins with potential ATPase activity. Additionally, WP_146368361.1, WP_146368222.1, WP_146368360.1 are ABC transporter permease subunits, while WP_146368765.1 and WP_146368337.1 are ABC transporter permeases, indicating their probable involvement as permease components. Furthermore, nsSNPs were also found in two presumed ATP-binding genes (WP_146368471.1 and WP_226364443.1 ATP-binding cassette domain-containing proteins) and a potential energy-coupling factor transporter ATPase (EcfA, WP_146368574.1). Among these, five probable ABC transporter coding genes (NCBI reference: WP_146368786.1, WP_226364443.1, WP_146368361.1, WP_146368222.1, QDY87225.1) displayed one common mutation among isolates with decreased susceptibility to enrofloxacin, macrolides, lincomycin and tiamulin (Supplementary Datasets S3, S5).

Regarding in vitro cultivated mutants, nsSNPs were identified in three ABC protein coding genes: WP_146368461.1, which encodes an ABC transporter ATP-binding protein and is a potential ATPase component of an ABC-type multidrug transport system; WP_146367439.1, which encodes the energy-coupling factor transporter transmembrane protein EcfT (presumed T component of ECF-type transporters); and WP_146368471.1, an assumed ATP-binding gene encoding an ATP-binding cassette domain-containing protein. Specifically, MycAv 668-M1/2/3/4 exhibited nsSNPs within WP_146368461.1, and MycAv 65-M1 displayed nsSNPs within WP_146367439.1 (Supplementary Dataset S3). Furthermore, in the case of WP_146368471.1 ATP-binding cassette domain-containing protein nsSNPs were identified both in clinical isolates with decreased antibiotic susceptibility (MycAv 47, 67, 68, 70, 91, 177) and in vitro cultivated mutants (listed in Supplementary Dataset S3). In the field isolates (MycAv 47, 67, 68, 70, 91, 177) either one or both nsSNPs were found at the nucleotide positions 250 and 1,339 (according to the numbering of the concerned gene of the reference strain). On the other hand, in the in vitro cultivated mutants, strains with decreased susceptibility to lincosamide, tiamulin and macrolides (MycAv55-M2/3/4/5/6) exhibited a mutation at the nucleotide positions 786 and 789 in the same gene (Supplementary Dataset S3).

Those ABC transporter protein coding genes that displayed common mutations in clinical isolates (MycAv 47, 67, 68, 70, 91, 177) or both in clinical and in vitro developed mutants were examined in all M. anserisalpingitidis strains to see any possible patterns among field isolates with low and high MIC values (Table 2 and Supplementary Dataset S5). The examined nsSNPs could be detected in higher percentages in clinical isolates with decreased antibiotic susceptibility. Most prominent differences were observed in the cases of the ABC transporter ATP-binding protein (WP_146368786.1; a potential ATPase and permease component of an ABC-type multidrug transport system), the ABC transporter permease subunit (WP_146368222.1) and the ATP-binding cassette domain-containing protein (QDY87225.1) (Table 2 and Supplementary Dataset S5).

Besides nsSNPs in ABC efflux pump coding genes, different mutations were identified in other genes previously associated with elevated MICs (Supplementary Dataset S4; Grózner et al., 2022). In terms of decreased susceptibility to fluoroquinolones, one common nsSNP was found in DNA gyrase subunit A (gyrA) gene [in 3/3 field isolates (MycAv 47, 70, 91) and in 2/3 (MycAv 63-M1, 668-M1) in vitro mutants] at the nucleotide position 446. Examining this mutation on all M. anserisalpingitidis strains with publicly available whole genome sequences and MIC data, there were three types of nsSNPs: strains with extremely high MICs (initial MICs were > 10 μg/mL) and eight strains with elevated MICs (six strains with initial MICs of 5 μg/mL, two strains of 2.5 μg/mL initial MIC values) had an nsSNP at the nucleotide position 446 which altered the original amino acid from Threonine to Isoleucine. Fourteen strains with increased MIC (initial MICs were 5 μg/mL in four cases and 10 strains had 2.5 μg/mL initial MIC value) and three strains with lower MIC (initial MICs were 1.25 μg/mL) displayed Guanine at the nucleotide position 445 which caused a changed from Threonine to Alanine. Strains in the third group were inhibited at lower MIC values (initial MICs were < 1.25 μg/mL) and contained Adenine and Cytosine at the nucleotide positions 445 and 446, coding Threonine (Supplementary Dataset S5).

Evaluating decreased tiamulin susceptibility, in vitro cultivated mutants (MyAv 50-M6, 55-M6, 668-M6) displayed mutations in the 23S rRNA gene at different positions (179, 434, 1,502, 1,568, 2,457) according to the numbering of the concerned gene of the type strain, while field isolates (MycAv 47, 67, 68, 70, 177) showed mutations at the nucleotide positions 793, 1,502 and 2067. In terms of the in vitro cultivated mutants with higher MIC values for lincomycin (MycAv 55-M2, 63-M2, 668-M2), mutations were determined in the 23S rRNA gene at different positions (179, 434, 1,568, 2067, 2,621). Regarding field isolates with decreased macrolide susceptibility (MycAv 67, 68, 177), three different nsSNPs were identified in the 23S rRNA gene at the nucleotide positions 793, 1,502 and 2067. The same nsSNP at the nucleotide position 793 was also present in three in vitro mutants (MycAv668-M3/5, MycAv55-M4), beside others (nucleotide positions 434, 544, 1,502, 1,568). For all strains of this bacterium with publicly available whole genome sequences and MIC data, the nsSNPs in the 23S rRNA gene were analyzed and reported in a previous work (Grózner et al., 2022).

Using the Promotech software, five putative efflux pump regulatory regions were identified related to the following genes: BMP family ABC transporter substrate-binding protein (NCBI reference: QDY86627.1), energy-coupling factor transporter ATPase (QDY86993.1), ATP-binding cassette domain-containing protein (WP_226364443.1), ABC transporter ATP-binding protein (WP_146368785.1), and multidrug transporter MATE (WP_201798414.1). The identified putative regulatory sequences with potential Pribnow box and RBSs (Figure 3A) showed statistically insignificant E-values according to the MEME analyses (E-value of the potential RBSs sequences is 0.33 and the E-value of the assumed Pribnow box is 2.4; Figures 3B,C). The putative efflux pump regulatory regions of the M. anserisalpingitidis type strain were compared to those of all M. anserisalpingitidis strains with publicly available whole genome sequences. However, no significant nsSNPs were found in these regions that could potentially contribute to lower antibiotic sensitivity.