Investigation of In Vitro Susceptibility and Resistance Mechanisms in Skin Pathogens: Perspectives for Fluoroquinolone Therapy in Canine Pyoderma

Fluoroquinolones (FQ) are commonly used in dogs with bacterial skin infections. Their use as first choice, along with the increased incidence of FQ-resistance, represents a risk to animal and public health. Our study determined minimum inhibitory (MIC) and bactericidal (MBC) concentrations of five FQs in Staphylococcus aureus, Staphylococcus pseudintermedius, and Escherichia coli, together with FQ-resistance mechanisms. MICs, efflux pump (EP) overexpression and MBCs were measured in 249 skin infection isolates following CLSI guidelines (CLSI VET01-A4, CLSI M26-A). Chromosomal and plasmid-mediated resistance genes were investigated after DNA extraction and sequencing. FQ-resistance was detected in 10% of methicillin-susceptible (MS), 90% of methicillin-resistant (MR) staphylococci and in 36% of E. coli. Bactericidal effect was observed except in 50% of MRSA/P for ciprofloxacin and in 20% of MRSPs for enrofloxacin. Highest MICs were associated with double mutation in gyrA (Ser83Leu + Asp87Asn), efflux pumps and three PMQR genes in E. coli, and grlA (Ser80Phe + Glu84Lys) in S. aureus. EP overexpression was high among E. coli (96%), low in S. aureus (1%) and absent in S. pseudintermedius. Pradofloxacin and moxifloxacin showed low MICs with bactericidal effect. Since in vitro FQ resistance was associated with MR, FQ use should be prudently guided by susceptibility testing.


Introduction
Canine pyoderma, a cutaneous pyogenic bacterial infection [1], represents one of the most common skin diseases in dogs [2], and is caused, in almost 90% of cases, by Staphylococcus pseudintermedius [3]. Other less-frequently isolated pathogens include Escherichia coli, Pseudomonas aeruginosa, and Streptococcus spp. [1,4]. S. aureus is rarely found on canine skin [5], and the majority of isolates are human-related [6]. However, the increased incidence of infections caused by multidrug-resistant (MDR) and methicillin-resistant (MR) staphylococci (MRSP and MRSA) in dogs is a public health concern [7,8], along with the increased frequency of infections caused by E. coli [9][10][11].
Fluoroquinolones (FQ) are efficacious and licensed for the systemic treatment of bacterial skin infections in dogs [12][13][14][15][16], and their use is advocated in deep pyoderma [17], or widespread and severe superficial pyoderma when antimicrobial susceptibility testing (AST) results are compatible.

Minimum Inhibitory Concentration
MICs distributions were bimodal for the three species ( Figure 1). In staphylococci, MS and MR resistant isolates showed different modes within each distribution.
MICs were lower for each FQ tested in MSSP compared with MRSP (p < 0.0001) and lower MICs were observed between both pradofloxacin, moxifloxacin (p < 0.0001) and the other FQs tested.
E. coli: 41 out of 65 isolates (63%) were FQ-susceptible except for moxifloxacin (36 out of 65, 55%). Highly significant differences in MICs and MBCs distribution between fluoroquinolones (p < 0.0001) were observed, and pradofloxacin showed the lowest MICs and MBC (p < 0.0001) when compared with the other FQs tested.

Minimum Bactericidal Concentration
MBCs distributions were bimodal for the three species ( Figure S1). In staphylococci, MS and MR isolates showed different modes within each distribution.  (Table S4), both for MSSA and MRSA, in comparison with the other FQs (p < 0.0001) and higher values were associated with methicillin resistant isolates (p < 0.0001). MBCs for MRSA were higher than MSSA (p < 0.0001) for all FQs tested.
E. coli: pradofloxacin showed the lowest MBCs (p < 0.0001) when compared with the other FQs tested. MBC/MIC ratios were within four doubling dilutions only with pradofloxacin and moxifloxacin.

Resistance Mechanism Detection
S. aureus: in five selected resistant isolates (two MSSA and three MRSA), gyrA showed either single (Ser84Leu, three out of six isolates) or double (Ser84Leu + Gly90Cys, two out of six isolates) chromosomal mutations. A single mutation (ser80Phe) was observed on grlA (Table 2) in four out of six isolates with pradofloxacin MIC ≤ 4 µg/mL whereas two mutations were observed in one out of six isolates with high pradofloxacin MIC (32 µg/mL). No mutations were detected in one isolate showing efflux pump overexpression, which represented 1% (1 out of 79) of the resistant isolates.

Discussion
Our data indicate that pradofloxacin and moxifloxacin overall show low MICs with bactericidal effect both in staphylococci and E. coli, compared to earlier generation veterinary and human FQs. The results are compatible with and expand those previously published on MICs [31,38,40,42,51,[55][56][57], in isolates collected from canine skin infections. However, this is the first veterinary study that examined a substantial number of MS and MR staphylococci in canine pyoderma. The percentages of FQ-resistances were high among MR staphylococci (90%) and E. coli (36-40%), in line with previous studies, where resistance was also detected among four non-β-lactam classes [40,[58][59][60]. Since the MIC data are obtained from different countries and periods of time, future analysis is needed to investigate the differences in terms of antibiotic pressure and the related resistance development between old and new generation FQs.
This is the first report on MBC of veterinary and human FQs in canine skin isolates. Previous studies showed that the MBC of ciprofloxacin, moxifloxacin, and pradofloxacin in MS and MR S. aureus [61], and pradofloxacin in S. pseudintermedius [54], was within four doubling dilutions of the MIC in all isolates. In E. coli, ciprofloxacin [62], and pradofloxacin [54], showed bactericidal effect in almost all isolates. Our study confirmed the high bactericidal effect in E. coli collected from canine pyoderma and extended to other veterinary and human FQs. However, in staphylococci no bactericidal effect (MBC/MIC ratio > four) was observed for ciprofloxacin in 50% of MRSA and MRSP and in 20% of enrofloxacin in MRSP. A comparison of the MBC 90 of ciprofloxacin in MRSA in our study with results from a similar study by Smith and Eng [63], reflects an alarming increase of more than 512-fold over a 30-year period.
Mutations on S. pseudintermedius isolates were found in codons gyrA84 and grlA80 in all selected FQ-resistant isolates as previously reported [45,64], except in one isolate with intermediate/increased susceptibility to pradofloxacin (0.5 µg/mL) with a single mutation on codon 88 (Glu88GLy). The same mutation was previously identified by Descloux et al. [45], in one FQ-resistant isolate with an MIC of enrofloxacin at the breakpoint (4 µg/mL) and one isolate from Japan [65], with intermediate susceptibility to ofloxacin and resistance to enrofloxacin and levofloxacin. In S. aureus, mutations on gyrA (Ser84Leu) and grlA (Ser80Phe) were identified both in isolates with intermediate susceptibility and resistance to pradofloxacin. However, the presence of an additional mutation on gyrA (Gly90Cys), together with grlA (Glu84Lys) as reported by Hiasa et al. [66], conferred high resistance to pradofloxacin (32 µg/mL) and moxifloxacin (16 µg/mL). Further screening is therefore necessary to understand the molecular mechanisms that confer low resistance profiles in pradofloxacin and moxifloxacin when compared to higher resistant MICs in early generation FQs.
All E. coli had two mutations on gyrA (Ser83Leu, Asp87Asn) and one mutation on parC (Ser80Ile). However, additional mutations on parC (Ala108Val, Glu84Val) were screened in isolates with both moderate and, if associated with plasmid genes, high resistance profiles.
Mutations were found to be within the quinolone resistance determining region (QRDR), which are in the proximity of the FQ-binding sites of the primary targets (Tyr122 in E. coli gyrA and Tyr119 in staphylococcal grlA). Mutations are known to reduce the hydrogen bond with FQs [67], with the addition of loss in negative charge with mutations in parC [68]. However, in silico tools such as molecular docking of FQs into the native and mutated protein are necessary to elucidate the primacy of enzyme targeting in veterinary FQs and their associated reduced resistance emergence.
Resistance conferred by efflux pumps was observed to be highly prevalent in E. coli (96%), compared to in S. aureus (1%) or absent in S. pseudintermedius. Pradofloxacin was the only FQ tested against all three bacterial species to detect efflux pump overexpression. Further research is therefore needed to investigate the susceptibility of other fluoroquinolones. It has been suggested that more lipophilic FQs (such as enrofloxacin, and moxifloxacin) are more readily excreted by efflux pumps in comparison to hydrophilic FQs (such as ciprofloxacin, pradofloxacin, and marbofloxacin) [38,[69][70][71][72][73]. In S. pseudintermedius, the mechanism was not detected in our study, which is in accordance with previous studies [31,32]. Although efflux pump overexpression was observed in only 1 out of 79 of S. aureus isolates, to our knowledge, this is the first report of FQ resistance due to an efflux pump mechanism in S. aureus from canine pyoderma. Similar results were obtained by Schmitz et al. [74], with moxifloxacin, where the change in MIC after exposure to reserpine were negligible (1-2 dilutions), compared to ciprofloxacin (1-4 dilutions). In contrast to staphylococci, almost all the E. coli isolates showed efflux pump overexpression mechanism, and we identified a higher percentage compared to previous studies on E. coli collected from canine otitis or different body sites [37,38].
With regards to PMQR, nearly 50% of FQ-resistant isolates were associated with at least one PMQR gene. Among the genes screened for, aac-(6 )-lb-cr (33%) and QnrS (12.5%) were the most prevalent, followed by QnrB (8.3%) as also demonstrated in other studies [37,38,75], whereas no QnrA was found. The presence of QnrB together with QnrS was associated with high MICs.
Limitations of the study are represented by the lack of veterinary CBPs for moxifloxacin and ciprofloxacin, that were extrapolated from human guidelines. The higher percentage of resistance of moxifloxacin in E. coli may be associated with discrepancies between EUCAST and CLSI breakpoints guidelines. As other authors have highlighted, this addresses the need for harmonized guidelines across countries [76]. Moreover, CBPs were not available for S. pseudintermedius and were extrapolated from S. aureus as adopted by previous investigators [51].

Bacterial Pathogens
A total of 249 isolates from canine skin or wound infections were included in this study.  [77,78], in staphylococci and uidA [79], and uspA [80], in E. coli. Methicillin resistance in staphylococci was investigated by the presence of the mecA gene [81].

Antimicrobial Agents Tested
Enrofloxacin, marbofloxacin, ciprofloxacin, and moxifloxacin were purchased from Merck (Steinheim, Germany). Pradofloxacin was provided by Bayer Animal Health (Monheim, Germany). Stock solutions (1 mg/mL) were dissolved in deionized water, filter sterilized, adjusted for potency according to Clinical and Laboratory Standard Institute guidelines (CLSI VET01-A4, 2013) [82], and stored in darkness at −80 • C for up to one month.

Minimum Inhibitory and Bactericidal Concentrations
MICs were determined by broth microdilution (CLSI VET01-A4) [82]: isolates were recovered from Brain Heart Infusion (BHI, ThermoFisher, Basingstoke, UK) with 25% glycerol at −80 • C, subcultured onto 5% sheep blood (TCS Bioscience, Botolph Claydon, UK) agar plates (BA, Merck, Steinheim, Germany) and incubated at 37 • C for 18-24 h. After incubation, 3-5 colonies were suspended into glass tubes containing phosphate buffer saline (PBS ThermoFisher, Basingstoke, UK). A spectrophotometer (Densichek ® , Biomérieux, Marcy L'Étoile, France) was used to standardize optical density to 0.5 McFarland, equal to 1-2 × 10 8 colony forming unit/mL (CFU/mL) and the bacterial suspension was diluted to achieve a final concentration of 5 × 10 5 CFU/mL in a 96-well MIC plate (Sarstedt ® , Nümbrecht, Germany) containing two-fold dilution series of the antimicrobial agents. Plates were incubated at 37 • C for 18-24 h and after incubation were manually read against a black paper background; MIC was considered as the lowest concentration where no visible growth was observed for each isolate. MIC 50 and MIC 90 (50th and 90th percentiles of the distribution) were calculated as the lowest concentrations that inhibited the growth of 50% and 90% of the isolates, respectively. S. aureus ATCC 29213 and E. coli ATCC 25922 were included for quality control purposes. Their respective ranges of MICs are shown in Table S1. Inoculum density and culture purity were confirmed according to EUCAST guidelines (2022) [83].
MICs data were compared with clinical breakpoints published by CLSI guidelines VET01S (2020) [84], for veterinary fluoroquinolones, and CLSI M100 (2022) [85], for human fluoroquinolones. Isolates with "intermediate" susceptibility were considered susceptible with increased exposure as reported by EUCAST. Epidemiological cut-offs (ECOFFs) were also included, if available from datasets, to distinguish between wild-and non-wildtype isolates. Moxifloxacin clinical breakpoints for E. coli were obtained from EUCAST guidelines (2022) [86]. For S. pseudintermedius, moxifloxacin and ciprofloxacin CBPs were adopted from CLSI for S. aureus as previously described in a veterinary study [51]. Clinical breakpoints are listed in Table S2.
After MIC determination, MBC was measured according to CLSI M26-A "Methods for Determining Bactericidal Activity of Antimicrobial Agents; Approved Guidelines" [87].
MBC 50 and MBC 90 (50th and 90th percentiles of the distribution) were calculated as the lowest concentrations that produced at least a 3log 10 reduction (99.9% bacterial killing) of viable bacterial populations (i.e., <5 × 10 2 CFU/mL) in antibiotic-treated wells where no visible growth was observed in 50% and 90% of the isolates, respectively. Moreover, drugs were considered bactericidal if the MBC/MIC ratio was ≤4.

Efflux Pump Overexpression
Resistant isolates were tested, according to the same MIC method previously described (CLSI VET04), with the addition of EPIs: 20 µg/mL of reserpine and 80 µg/mL of PaβN for E. coli were added to MIC plates containing 2-fold pradofloxacin (used as standard FQ) dilutions. Overexpression of efflux pump was detected if the ratio between MIC in the absence and MIC in the presence of EPI (MIC/MIC EPI ) was ≥4.

Chromosomal Mutations and PMQR Genes
Eighteen FQ-resistant isolates (6 S. aureus, 6 S. pseudintermedius, and 6 E. coli, Table 2) were chosen to represent different resistance profiles based on pradofloxacin MICs (low/susceptible with increased exposure, medium, and high resistance).
DNA was sequenced for the presence of chromosomal mutations on the first subunit of the two topoisomerases (DNA gyrase and topoisomerase IV) targeted by fluoroquinolones, namely gyrA and grlA (parC for E. coli).
DNA was extracted from bacterial cells: 5-6 colonies of an overnight culture on sheep blood agar were suspended into Eppendorf tubes containing 1 mL of PBS and centrifuged at 5000 g for 10 min to pellet cells. The supernatant was removed, cell pellet was resuspended in 100 µL of Tris EDTA (TE) buffer and heated at 100 • C for 10 min. For staphylococci, the heated suspension was incubated on ice for 1 min to degrade the cell wall. The final step involved centrifugation at 5000 g for 30 s to pellet cell debris and collect supernatant into sterile Eppendorf tubes.
GyrA and grlA (parC) genes were amplified with polymerase chain reaction (PCR) using previously published protocols [64,88,89]. PCR products were run with electrophoresis on 2% agarose gel together with a positive and a negative control. Results were read with a transilluminator, and results saved on electronic files.
DNA concentration was measured with a fluorometer (Qubit, Invitrogen) and diluted to a standard concentration of 10 ng/mL. Each primer (both forward and reverse) was separately diluted to a standard concentration of 1.3 nmol/µL. Samples were analyzed by Source bioscience (Cambridge) and genetic sequences were aligned and compared with QC isolates with Bioedit 7.2 version.

Statistical Analysis
MICs and MBCs data were log 2 transformed before statistical analysis and a Shapiro-Wilk test was used to assess normality of the distributions. MS and MR staphylococci were considered as two different bacterial types within the same species.
MICs and MBCs distributions were compared between 5 fluoroquinolones within bacterial type (MS, MR staphylococci and E. coli) using Anderson Darling and Friedman's tests. Dunn's post hoc analysis was carried out for pairwise comparisons.
Wilcoxon signed-rank tests were used to compare the median MICs distributions of pradofloxacin in presence and absence of EPI. Analyses were performed using Graph-Pad Prism version 9.0 statistical software package (San Diego, CA, USA) with p < 0.05 for significance.

Conclusions
Lowest MICs and MBCs were measured with pradofloxacin and moxifloxacin. The presence of methicillin resistance can predict FQ-resistance in more than 90% of MR staphylococci and in 36% of E. coli. Bactericidal effect may not be achieved in MR staphylococci with enrofloxacin and ciprofloxacin.
Since FQ-resistance is multifactorial, further molecular screening of resistance mechanisms and its correlation with antimicrobial susceptibility are necessary.
These findings address the need for prudent use of FQs that should be preventively guided by antimicrobial susceptibility testing prior to systemic treatment.

Conflicts 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. LP acted as a consultant for Zoetis, Vetcare, CEVA and Beckton Dickinson, received research grants from DEFRA, WSAVA, PetPlan, Transpharm, UKRI, BBSRC, Horizon 2020 and is an active member of the VetCAST steering committee. FZ, DT, and LP are members of the European Network for Optimization of Veterinary Antimicrobial Treatment (ENOVAT, CA18217).