Mycoplasma bovis in Spanish Cattle Herds: Two Groups of Multiresistant Isolates Predominate, with One Remaining Susceptible to Fluoroquinolones

Mycoplasma bovis is an important bovine pathogen causing pneumonia, mastitis, and arthritis and is responsible for major economic losses worldwide. In the absence of an efficient vaccine, control of M. bovis infections mainly relies on antimicrobial treatments, but resistance is reported in an increasing number of countries. To address the situation in Spain, M. bovis was searched in 436 samples collected from beef and dairy cattle (2016–2019) and 28% were positive. Single-locus typing using polC sequences further revealed that two subtypes ST2 and ST3, circulate in Spain both in beef and dairy cattle, regardless of the regions or the clinical signs. Monitoring of ST2 and ST3 isolates minimum inhibitory concentration (MIC) to a panel of antimicrobials revealed one major difference when using fluoroquinolones (FQL): ST2 is more susceptible than ST3. Accordingly, whole-genome sequencing (WGS) further identified mutations in the gyrA and parC regions, encoding quinolone resistance-determining regions (QRDR) only in ST3 isolates. This situation shows the capacity of ST3 to accumulate mutations in QRDR and might reflect the selective pressure imposed by the extensive use of these antimicrobials. MIC values and detection of mutations by WGS also showed that most Spanish isolates are resistant to macrolides, lincosamides, and tetracyclines. Valnemulin was the only one effective, at least in vitro, against both STs.


Introduction
Isolated in the early 60s, Mycoplasma bovis is an important bovine pathogen that has a major economic impact on the global cattle industry [1,2]. M. bovis is usually associated with a variety of clinical manifestations, including pneumonia, mastitis, arthritis, keratoconjunctivitis, otitis media, and genital disorders [2,3]. In the absence of an efficient vaccine, the control of M. bovis infections mainly relies on antimicrobial treatments [4]. However, many countries have reported that the in vitro antimicrobial sensitivity of M. bovis isolates has been dramatically reduced [5][6][7][8][9][10][11][12][13][14].
M. bovis belongs to the class Mollicutes, a large group of wall-less bacteria with reduced genome and limited metabolic capacities, but a remarkable adaptive potential [15,16]. Treatment with ß-lactams, were distributed in Spanish herds when considering a large number of field isolates, and whether their antimicrobial susceptibility profiles were congruent with polC typing. Spain, which allowed unrestricted use of FLQ until very recently, may serve as a clear in vivo model to study the effects of the indiscriminate use of these antimicrobials.
The present study objectives were (i) to assess the circulation of M. bovis in Spanish cattle herds using a large collection of isolates collected from beef and dairy cattle and from different sample sources (nasal, auricular, conjunctival, synovial fluid and tissues swabs, and mastitic milk); (ii) to subtype this collection by single-locus sequencing of polC [41]; (iii) to determine the antimicrobial susceptibility of M. bovis isolates studying differences between STs, with a focus on antimicrobial agents approved to treat bovine respiratory disease and/or mastitis in Spain; and (iv) to assess the occurrence of genetic mutations conferring antimicrobial resistance in a selection of isolates representative of each ST.

M. bovis Circulating in Spanish Beef and Dairy Herds Belongs to STs 2 and 3
In this study, 93 (35.7%) of the 260 analyzed animals were infected with M. bovis. Among the 436 analyzed samples, a total of 165 tested positive for Mycoplasma spp. and M. bovis was the most commonly found species, with 122 PCR-positive samples.
Among beef cattle, M. bovis was detected in 84 (40.9%) of the 205 analyzed animals. Specifically, the pathogen was detected in 81 (44.3%) of the 183 feedlots calves and 3 (13.6%) of the 22 pasture-raised animals. The pathogen was detected in 40 (32%) of the 125 healthy animals and 44 (55%) of the 80 animals with clinical signs of respiratory disease or arthritis. Within the 331 analyzed samples, 102 were tested positive. Most positive samples were obtained from nasal swabs (85/278) and the remaining were identified in auricular swabs (5/27) and tissues swabs from lung (9/16), spleen (1/1), liver (1/2), and mediastinal lymph nodes (1/1). However, the pathogen was not found in conjunctival swabs (n = 3) nor synovial fluid (n = 3). The positive samples were obtained from 26 of the 30 analyzed farms and 5 of the 8 analyzed regions ( Figure S1). Among dairy cattle, M. bovis was detected in 9 (16.36%) of the 55 analyzed animals. Specifically, the pathogen was detected in 9 (23.1%) of the 39 dairy cows with mastitis but was not detected in any of the 5 dairy calves with clinical signs of respiratory disease nor any of the 11 asymptomatic calves. Within the 105 analyzed samples, positive samples were only detected in mastitic milk (20/66), while any positive results were detected in BTM (n = 9), or nasal (n = 27), auricular (n = 1), or conjunctival (n = 2) swabs. The positive samples were obtained from 2 of the 7 farms and the milk analysis laboratory, and 3 of the 5 analyzed regions ( Figure S1).
Globally, M. bovis was successfully isolated from 112 PCR-positive samples. Based on their origin, 95 representative isolates were chosen for further characterization (epidemiological background provided in Table S1 and illustrated in Figure 1). Briefly, the collection included isolates from beef (n = 75) and dairy cattle (n = 20). Beef cattle isolates were obtained from nasal (62/75), auricular (6/75), lung (6/75) and spleen swabs (1/75), asymptomatic (35/75) or with clinical signs of respiratory disease (33/75), arthritis (6/75), or both (1/75). Dairy cattle isolates were obtained from mastitic milk. Single-locus sequence analysis of polC revealed two ST profiles: ST2 (n = 37) and ST3 (n = 58). Both STs were found in beef and dairy cattle, in healthy or diseased animals and in different sample sources. Both STs were found concomitantly in animals from the same farm, or even in different samples from the same animal ( Figure 1, Table S1). For example, isolates J96 and J102 (ST3) and J103 (ST2) were collected from spleen, nasal, and lung swabs of the same animal respectively (Table S1). Sequences corresponding to ST2 and ST3 are provided in Table S2.
Hence, no other STs than ST2 or ST3 were found in Spanish herds. Both STs were present in asymptomatic beef cattle or with clinical signs of respiratory disease or arthritis and in dairy cows with mastitis.

The Antimicrobial Susceptibility Profiles of The Spanish Isolates to FLQ Differ Between PolC ST2 and ST3
The MIC values for the reference strain PG45 are shown in Table 1. Individual MIC values for  each isolate are listed in Table S1. Statistical analyses revealed a significant difference in antimicrobial susceptibility to FLQ between ST2 and ST3 isolates (p < 0.01). No significant changes between STs were observed for macrolides, lincomycin, doxycycline, or valnemulin. The antimicrobial susceptibility profile of these two STs is illustrated in Table 1, Figures 1 and 2.
Therefore, most of the M. bovis Spanish field isolates have a similar antimicrobial susceptibility profile against macrolides, lincomycin, and doxycycline with high MIC values and for valnemulin with low MIC values. On the contrary, antimicrobial susceptibility profiles against FLQ differed between ST2 and ST3, with high MIC values mainly associated with ST3 (Table 1).

Analysis of Point Mutations Conferring Resistance to Antimicrobials: The Main Differences between ST2 and ST3 are Found in The QRDR of GyrA and ParC Genes
A total of 36 M. bovis isolates belonging to ST2 (n = 16) and ST3 (n = 20) were subjected to wholegenome sequencing to compare nucleotide changes at QRDR, and rRNA (16S and 23S) and protein (L3, L4, and L22) genes (Table 2-4). The epidemiological background of these isolates is provided in Table  S1 and illustrated in Figure 1.
Nucleotide changes at QRDR revealed important differences between each ST, mainly located in gyrA and parC. While sequence analysis did not reveal any non-synonymous mutations in gyrA or parC for ST2 isolates, ST3 isolates were all characterized by at least one non-synonymous mutation in one or both genes. ST3 isolates were all characterized by a parC non-synonymous mutation at codon 10 (Gln10Arg). This mutation was associated with a substitution from serine to phenylalanine at gyrA codon 83 (Ser83Phe) and serine to isoleucine at parC codon 80 (Ser80Ile) in isolates with MIC values ≥ 1 µg/mL for FLQ. Among the few exceptions were the isolates J28, J228, and J279 having no mutation at parC codon 80, but a non-synonymous mutation at codon 116 (Ala116Pro in J228 and J279) or codons 81 and 84 (Ser81Pro; Asp84Asn in J28). Interestingly, while most of the ST2 and ST3 isolates showed a gyrB

Analysis of Point Mutations Conferring Resistance to Antimicrobials: The Main Differences between ST2 and ST3 Are Found in The QRDR of GyrA and ParC Genes
A total of 36 M. bovis isolates belonging to ST2 (n = 16) and ST3 (n = 20) were subjected to whole-genome sequencing to compare nucleotide changes at QRDR, and rRNA (16S and 23S) and protein (L3, L4, and L22) genes (Tables 2-4). The epidemiological background of these isolates is provided in Table S1 and illustrated in Figure 1.
Nucleotide changes at QRDR revealed important differences between each ST, mainly located in gyrA and parC. While sequence analysis did not reveal any non-synonymous mutations in gyrA or parC for ST2 isolates, ST3 isolates were all characterized by at least one non-synonymous mutation in one or both genes. ST3 isolates were all characterized by a parC non-synonymous mutation at codon 10 (Gln10Arg). This mutation was associated with a substitution from serine to phenylalanine at gyrA codon 83 (Ser83Phe) and serine to isoleucine at parC codon 80 (Ser80Ile) in isolates with MIC values ≥ 1 µg/mL for FLQ. Among the few exceptions were the isolates J28, J228, and J279 having no mutation at parC codon 80, but a non-synonymous mutation at codon 116 (Ala116Pro in J228 and J279) or codons 81 and 84 (Ser81Pro; Asp84Asn in J28). Interestingly, while most of the ST2 and ST3 isolates showed a gyrB non-synonymous mutation associated with a substitution Asp362Asn, ST3 isolates J479, and J482 (MIC values ≥ 8 µg/mL for FLQ) were characterized by a substitution at gyrB codon 323 (Val323Ala) in combination with mutations Ser83Phe in gyrA, and Gln10Arg, Ser80Ile, and Val156Ile in parC.
Mutations in the 23S rRNA and 16S rRNA genes and the ribosomal proteins L3, L4, and L22 are listed in Table 3; Table 4. Regarding 23S rRNA, positions A534T, G748A were notably altered in both rrl alleles of all the isolates. Mutation A2058G affecting the majority of isolates (34/36) in one or both alleles was only absent in those with low MIC values for lincomycin (1 µg/mL). Mutations G954A in one or both alleles were altered in 31/36 isolates from both STs and the remaining five isolates had many compensatory non-synonymous mutations in L3, L4, and L22 proteins. Mutation T1249C in one allele was altered in 31/36 isolates from both STs. Mutations A1251T (1/36) and G2157A (5/36) in one allele and G2848T (2/36) in one allele were only found in ST3 isolates while G452A was present in one allele of a few number (5/36) of ST3 isolates. Some isolates from both STs (6/36) showed a single non-synonymous mutation in L4 or L22 (Table 3). Regarding 16S rRNA, mutations A965T and A967T were altered in both rrs alleles of all the isolates (MIC ≥ 1 µg/mL for doxycycline). Mutations C1192A in both alleles and T1199C in one or both alleles were altered in 31/36 isolates from both STs. Mutations C335T and C859T were present in one rrs allele of five isolates (from both STs) and one isolate (ST2) respectively (Table 4).
Hence, the main differences between ST2 and ST3 are found in the QRDR of gyrA and parC genes. None of the ST2 isolates have any amino acid substitution in either gyrA or parC while ST3 isolates with MIC values ≥ 1 µg/mL for FLQ have the mutation Ser83Phe in gyrA in combination with at least non-synonymous mutation in parC (positions 80, 81, 84, 116, and156).

Discussion
M. bovis was found to be widely distributed in Spanish cattle herds. More specifically, M. bovis was mainly detected in feedlot calves (81/183) and to a lesser extent in pasture-raised animals (3/22) housed in 26 different farms from 5 Spanish regions. This pathogenic species was not only detected in animals suffering from respiratory infections and/or arthritis (44/80), but also in asymptomatic carriers (40/125). These results consolidate previous studies that reported the isolation of M. bovis from young cattle with respiratory disease in Spain between 2010-2012 and 2015-2016 [37,38]. Although the complete epidemiological background of those isolates was not provided, the authors indicated that each isolate was obtained from a different farm. Altogether, these data indicate that, at least among beef cattle, the infection may have already become endemic, as reported in other European countries [25][26][27]. The presence of asymptomatic carriers and the movement of cattle between beef cattle farms, which frequently involves the mix of animals of diverse origins [39], may explain the current situation in Spain. The isolation of M. bovis from clinical mastitis cases was unusual given the low prevalence of this infection in other European countries. Therefore, further studies are needed to confirm whether this particular situation only reflects a bias of the sampling procedure or indicates that Spain is facing an important increase in the number of mastitis cases associated with M. bovis.
M. bovis isolates circulating in Spain are divided into two polC STs, 2 and 3. These two STs are similar to recent French isolates [41][42][43]. Compared with France, where ST2 has been predominant since 2000 [41][42][43], almost two thirds (58/95) of the characterized Spanish isolates belong to ST3. Both STs are widely distributed among different farms and regions, and can be isolated from beef and dairy cattle, from animals with different clinical conditions, and even from different anatomic locations of the same animal. This argues in favor of an efficient circulation and transmission of both STs, as already suggested with French isolates. Thus, animal movement between farms, a common practice in the Spanish beef cattle industry, is likely contributing to the dissemination of M. bovis [39]. Animal movements between dairy farms is less common, but asymptomatic carriers can be introduced into the herd when the replacement rate of animals born in the same herd is insufficient to maintain milk production. Furthermore, artificial insemination may be another way of entry for M. bovis. This was recently documented in Finland, where semen was reported to be the source of M. bovis mastitis outbreaks in two dairy herds [40].
Antimicrobial susceptibility profiles against FLQ differed between ST2 and ST3 isolates. The analysis of the QRDR revealed that the main differences between these STs were located in gyrA and parC. Remarkably, ST3 isolates were all characterized by an unusual Gln10Arg mutation in parC. This mutation is unrelated to antimicrobial resistance, since it was found in ST3 isolates associated with high and low MIC values (≥ 1 and ≤ 1 µg/mL, respectively), and are likely to reflect phylogenetic evolution. ST3 isolates with MIC values ≥ 1 µg/mL were all characterized by mutation Ser83Phe in gyrA in combination with one or more amino acid substitution (Ser80Ile, Ser81Pro, Asp84Asn, Ala116Pro, or Val156Ile) in parC. Only three of these parC mutations, Ser80Ile, Ser81Pro, andAsp84Asn, have been previously described [42,[45][46][47][48]. A point mutation Ser83Phe in GyrA is sufficient to reach an intermediate level of susceptibility to FLQ but additional substitutions in parC are required for resistance [42,[44][45][46][47][48]. Interestingly, ST2 and a majority of ST3 (18/20) isolates had the mutation Asp362Asn in gyrB. This mutation also appears in recent French isolates and is related to phylogenetic evolution rather than drug resistance [41,42]. Two ST3 isolates harbor a Val323Ala mutation in gyrB, but its contribution to FLQ resistance is unknown.
Our results are consistent with in vitro studies showing that under selective pressure, ST3 isolates are more prone to accumulate QRDR mutations than ST2 isolates. Therefore, the widespread circulation of FLQ-resistant ST3 isolates in Spain might reflect the overuse of these antimicrobials in the field. Remarkably, two ST2 isolates were also found to be resistant to FLQ. They were isolated from a cow with clinical mastitis together with susceptible ST2 isolates. This may be the result of long-term treatment with FLQ, leading to the generation of resistant strains, and re-infection with susceptible strains. Globally, our results contrast with other countries where most M. bovis strains are susceptible to this family of antimicrobials [6,[9][10][11][12][13].
MIC values confirmed the general decrease of M. bovis susceptibility to macrolides and lincomycin (MIC 90 > 128) [5,[9][10][11][12][13]. Analysis of 23S rRNA genes revealed that isolates with MIC values > 128 µg/mL for macrolides and lincomycin acquired mutations G748A (in both rrl alleles) and A2058G (in one or both rrl alleles). A combination of mutations in these hotspots is necessary and sufficient to achieve resistance to other macrolides, such as tylosin and tilmicosin, while mutation A2058G in one or both alleles has been linked to lincomycin resistance in M. synoviae [43,49,50]. Isolates J28 and J137 showed high MIC values (16-128 µg/mL) for macrolides but did not carry the mutation A2058G. Consistently, they are the only isolates with low MIC values for lincomycin (1 µg/mL). However, both isolates have several non-synonymous mutations in L4 and L22 proteins including Gln93His in L22, which is related to macrolide resistance and could explain the observed high MIC values for these antimicrobials [43]. No other point mutations related to antimicrobial resistance have been found in the rrl alleles or in L4 and L22 proteins. Since they appear together with other mutations conferring resistance, it is difficult to determine their importance.
As expected by the in vitro antimicrobial activity of pleuromutilins against a broad range of veterinary mycoplasmas [22], valnemulin was the only antimicrobial that demonstrated activity against both STs. Indeed, no mutation previously associated with pleuromutilin resistance [47] has been observed in any isolate. This is consistent with the fact that pleuromutilins are only registered for treatment in swine and poultry [52]. Valnemulin may thus be an interesting therapeutic alternative as it has been shown to be effective for the treatment of calves experimentally infected with M. bovis [53].
Overall, low in vitro susceptibility was observed for doxycycline (MIC 90 = 4 µg/mL). Analysis of 16S rRNA genes revealed that isolates with MIC values ≥ 1 µg/mL were characterized by mutations A965T and A967T in both rrs alleles. Previous studies have concluded that this double mutation causes decreased susceptibility to other antimicrobials from the same group, such as oxytetracycline and tetracycline [43,51]. Mutations C1192A and T1199C were previously described in French isolates [43], although they did not further modify MIC values as it occurs with Spanish isolates. However, the mutation C1192A has been described both in Hungarian and Japanese isolates and was associated with high MIC values for spectinomycin [47,48]. As expected, mutations C335T and C859T, which have never been associated with antimicrobial resistance, had no influence on the susceptibility of the Spanish isolates. Finally, our results were also consistent with data suggesting that after macrolides, the highest resistances of the main veterinary mycoplasmas species are observed for tetracyclines [22].
In conclusion, our study revealed the extended circulation of M. bovis in Spanish beef cattle herds and its implication in mastitis cases. Circulating isolates are divided into two groups, ST2 and ST3, both being resistant to macrolides, lincosamides and tetracyclines. Most ST3 isolates circulating in Spain are resistant to FLQ, a situation which illustrates the remarkable capacity of ST3 to accumulate mutations in QRDR and the selective pressure imposed by the indiscriminate use of these antimicrobials. Valnemulin has been shown to be very effective against both STs in vitro, and its effectiveness in vivo should be further investigated.

Animal Sampling
All animal procedures were performed following the EU Directive 2010/63/EU for animal experimentation and had the authorization of the Ethics Committee on Animal Testing of the University of Murcia (Number: 307/2017).
In this study, 260 animals from 10 Spanish regions were sampled over a 4 year period (2016-2019). A total of 433 samples were collected from beef and dairy cattle.
Among beef cattle, 183 calves were raised in feedlots and 22 pasture-raised animals were sampled. Healthy animals (n = 125) and animals with clinical symptoms of respiratory disease or arthritis (n = 80) were both considered. In total, 331 samples were obtained from beef cattle. The sample collection was composed of nasal swabs (n = 278), auricular (n = 27) and conjunctival swabs (n = 3), synovial fluid (n = 3), as well as a number of swabs from tissues (lung, n = 16; liver, n = 2; spleen, n = 1; and mediastinal lymph node, n = 1). Those samples were obtained from 30 farms and 8 different regions ( Figure S1).
Among dairy cattle, 39 cows with mastitis, and 16 calves with clinical signs of respiratory disease (n = 5) or asymptomatic (n = 11) were sampled. In total, 105 samples were obtained from dairy cattle. The sample collection was composed of mastitic milk (n = 66), bulk tank milk (BTM) (n = 9), and nasal (n = 27), auricular (n = 1), and conjunctival swabs (n = 2). Those samples were obtained from 7 farms and a milk analyses laboratory that provided samples and they were taken from 5 different regions ( Figure S1).

Mycoplasma Isolation and Subtyping
For mycoplasma isolation from animal samples, swabs or mastitic milk samples (200 µL) were incubated at 37 • C for 24 h in 2 mL of SP4 medium [54] with modifications (Appendix A). Cultures were filtered through a 0.45 µm membrane filter (LLG-Labware, UK) and further incubated for 48 h before plating 5 µL onto solid SP4 medium. Agar plates were grown at 37 • C and examined daily under the microscope for the presence of mycoplasma colonies with the typical fried egg morphology.
The DNA extraction was performed from 200 µL of culture [55]. M. bovis detection was performed by PCR amplification of the membrane protein 81 gene [56]. M. bovis PCR positive cultures were three times cloned by picking single colonies and the identity of the final isolate was confirmed again by PCR.
M. bovis subtyping was performed by sequence analysis of a 520 bp region of the polC gene, as previously described [41]. Amplicon sequencing was performed at the molecular biology service of the University of Murcia and sequence analyses were conducted using MEGA 6 [57]. . Stock solutions (1 mg/mL; 0.1 mg/mL for valnemulin hydrochloride) and two-fold dilutions were prepared in sterile distilled water. For preparing enrofloxacin, marbofloxacin, and danofloxacin, 0.1 M HCl was added dropwise until dissolution occurred and the volume was adjusted with sterile distilled water. A final range from 128 µg/mL to 0.0625 µg/mL was tested except for valnemulin, for which a final range from 12.8 µg/mL to 0.00625 µg/mL was studied.

MIC Assays
Stationary-phase cultures of 95 M. bovis isolates and the reference strain PG45 were used for MIC assays. Mycoplasma cultures were carried out in PH medium [58] without antimicrobials, supplemented with sodium pyruvate (0.5%) and phenol red (0.005%), and mycoplasma titers were determined as previously described [59]. MIC assays were carried out in 96-well microtiter plates using the microbroth dilution method [23]. Briefly, 25.6 µL of each antimicrobial dilution and 25 µL of the diluted M. bovis inoculum (10 3 -10 5 CFU/mL) were added to 150 µL of culture medium. Additionally, a positive control (well without antimicrobial) and a negative control (well without neither antimicrobial nor inoculum) were included in each essay. After 48 h of incubation at 37 • C, plates were examined for color change. MIC was defined as the lowest concentration of antimicrobial capable of completely inhibiting the growth of M. bovis. For each antimicrobial, the MIC range, MIC 50 (lowest concentration of antimicrobial capable of inhibiting the growth of 50% of the isolates), and MIC 90 (lowest concentration of antimicrobial capable of inhibiting the growth of 90% of the isolates) were calculated. All the assays were performed in duplicate. For accepting the results, MIC values of the duplicate tests had to be within one dilution, with the higher MIC value being used. If not, a third assay was performed, and the final MIC value was the mode of the three values.

Statistical Analysis
MIC values were transformed to a continuous variable by calculating their Log2 values. Log2MIC means values of ST2 and ST3 isolates were compared for each antimicrobial. Statistical analyses were run using the EpiInfo software [60] using ANOVA or Mann-Whitney/Wilcoxon Two-Sample Test (Kruskal-Wallis test for two groups) according to the inequality of population variances and with the significance level set at 0.01.

Whole-Genome Sequencing
Genomic DNA was extracted from a selection of 36 isolates (Table S1) from 15 mL of mycoplasma culture using a High Pure PCR Template Preparation Kit (Roche, Bâle, Suisse) according to the manufacturer's instructions. Whole-genome sequencing was performed using Illumina technology Hiseq (paired-end, 2 × 150pb) by Novogene Europe (Cambridge, UK). Bioinformatics analyses were performed on Galaxy platform (Genotoul, Toulouse, France). Quality controls of reads were performed using FastQC tool [61]. Alignments were carried out with BWA-MEM using PG45 as the reference [62], and alignments quality controls were checked with QualiMap BamQC [63]. SNP identification was done by alignment visualization with Integrative Genomics Viewer (IGV 2.7.0) [64] or by variant calling analysis with breseq [65]. All sequence files are available from the European Nucleotide Archive database (ENA), under study accession number PRJEB38707.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-0817/9/7/545/s1, Figure S1: Map of Spain showing the autonomous communities (AC) and the origin of the samples, Table S1: Epidemiological background, polC characterization and minimum inhibitory concentration (MIC) values of the 95 Mycoplasma bovis isolates, Table S2: Partial sequences (520 pb) types of the polC gene.  Acknowledgments: The authors are grateful to: Ángel García Muñoz, Juan Seva Alcaraz, and Juan Alcazar Triviño for providing us the contacts for taking samples; to Mercè Lazaro for providing us the mastitic milk samples received in her laboratory; to the veterinarians that collected samples in the field; to the farmers for allowing the collection of the samples; to the Genotoul bioinformatics platform Toulouse Midi-Pyrenees for providing help and storage resources.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A
The medium SP4 was prepared following previous recommendations [54] but with some modifications.The modified medium is composed of three parts (A, B, and C). Part A is composed of 4.2 g of Difco PPLO broth (BD), 6.4 g of Bacto Peptone (BD), 12 g of Bacto Tryptone (BD) and 724 mL of deionized water. The solid medium includes 7 g of European Bacteriological Agar (Conda-Pronadisa). The pH is adjusted to 7.8 and then part A is autoclaved at 121 • C for 20 min. Part B is composed of 60 mL of RPMI-1640 (Sigma-Aldrich), 21 mL of fresh yeast extract 50% w/v, 2.4 g of yeast extract (Conda-Pronadisa), 4.8 mL of phenol red 0.5%, (Sigma-Aldrich) and 0.642 g of ampicillin sodium salt (Fisher bioreagents). The pH is adjusted to 7.2 and then part B is filter-sterilized through a 0.2 µL pore size filter. Part C is composed of 251 mL of heat-inactivated horse serum (Hyclone) for 30 min at 56 • C.