Dosing Regimen of Enrofloxacin Impacts Intestinal Pharmacokinetics and the Fecal Microbiota in Steers

Objective: The intestinal concentrations of antimicrobial drugs that select for resistance in fecal bacteria of cattle are poorly understood. Our objective was to associate active drug concentrations in the intestine of steers with changes in the resistance profile and composition of the fecal microbiome. Methods: Steers were administered either a single dose (12.5 mg/kg) or 3 multiple doses (5 mg/kg) of enrofloxacin subcutaneously every 24 h. Enrofloxacin and ciprofloxacin concentrations in intestinal fluid were measured over 96 h, and the abundance and MIC of E. coli in culture and the composition of the fecal microbiota by 16S rRNA gene sequencing were assessed over 192 h after initial treatment. Results: Active drug concentrations in the ileum and colon exceeded plasma and interstitial fluid concentrations, but were largely eliminated by 48 h after the last dose. The concentration of E. coli in the feces significantly decreased during peak drug concentrations, but returned to baseline by 96 h in both groups. The median MIC of E. coli isolates increased for 24 h in the single dose group, and for 48 h in the multiple dose group. The median MIC was higher in the multiple dose group when compared to the single dose group starting 12 h after the initial dose. The diversity of the fecal microbiota did not change in either treatment group, and taxa-specific changes were primarily seen in phyla commonly associated with the rumen. Conclusions: Both dosing regimens of enrofloxacin achieve high concentrations in the intestinal lumen, and the rapid elimination mitigates long-term impacts on fecal E. coli resistance and the microbiota.


INTRODUCTION
Some studies have demonstrated an association between antimicrobial administration and antimicrobial resistance (AMR) in fecal bacteria of cattle (Kanwar et al., 2013;Zaheer et al., 2013;Hog and Korsgaard, 2017). Yet others demonstrate little impact of antimicrobial therapy on AMR in fecal microbiota of cattle (Checkley et al., 2010;Morley et al., 2011;Schmidt et al., 2013).
Methodological variation in identifying and characterizing antimicrobial resistance in the studies investigating AMR associated with treatment in feedlots prevents meaningful comparison of their results. Additionally, conclusions are often drawn from single bacterial isolates, selected resistance genes, or over varying sampling times, which further confounds interpretation and comparison between studies (Benedict et al., 2013(Benedict et al., , 2014. Parenteral antibiotics administered to treat infections diffuse into the intestine and influence the population and susceptibility of the intestinal bacteria (Lindecrona et al., 2000;Ferran et al., 2013;Foster et al., 2016). However, there is little known about the duration of this effect, particularly on potential foodborne pathogens. Measurement of active, unbound drug concentrations in the intestinal lumen is necessary to correlate pharmacokineticpharmacodynamic (PK-PD) indices with microbial changes. Based on evidence that the exposure relationships are important for development of antimicrobial resistance (Martinez et al., 2012), it is possible that resistance can be affected by the dose and frequency of drug administration (Enne, 2010).
In the US, there are two approved dosing regimens for enrofloxacin for treating respiratory disease in cattle, but little is known about the relative impact of these dosing regimens on AMR. It is presumed that each dosing regimen has similar efficacy; therefore, impact of each regimen on AMR may provide a rational basis for choosing one regimen over the other. Further, understanding the relationships between dosing regimen of enrofloxacin and AMR is critical due to the importance of fluoroquinolones in human health. The objective of this study was to model the PK of the active concentrations of enrofloxacin and its metabolite, ciprofloxacin, in the intestine of steers, and correlate these PK indices with changes in the MIC of E. coli and fecal microbiome. Our hypothesis is that the single 12.5 mg/kg dose of enrofloxacin will achieve a greater maximum concentration of active drug within the intestine resulting in less selection for AMR bacteria and changes in the microbiota due to the brief drug exposure.

Animals and Treatments
This study was approved by the North Carolina State University Institutional Animal Care and Use Committee. Twelve healthy, 6-months-old Holstein steers (125 to 241 kg) were obtained as in previous studies (McKellar et al., 1999;TerHune et al., 2005;Davis et al., 2007;Foster et al., 2016). Sample size was determined based on previous PK-PD studies in cattle demonstrating differences in drug concentrations in different sampling locations (Davis et al., 2007;Foster et al., 2016). At 24-48 h post-probe placement (described below), steers received either a single subcutaneous injection of 12.5 mg/kg enrofloxacin (Baytril 100 R ; Bayer Animal Health, Shawnee Mission, KS) or three subcutaneous injections of 5.0 mg/kg enrofloxacin (Baytril 100 R ) administered once each day on 3 sequential days. Treatments were administered in a crossover design, with each calf receiving both treatments separated by a minimum washout period of 10 days, accounting for at least six drug half-lives from last dose. Half of the steers (n = 6) received the single dose regimen first, while the other half received the multiple dose regimen first.

Interstitial Fluid Collection
An in-vivo ultrafiltration probe (UF-3-12, BAS; Bioanalytical Systems, West Lafayette, IN, USA) was inserted in the subcutaneous space above the shoulders as described previously (Davis et al., 2007;Messenger et al., 2012). The interstitial fluid (ISF) was collected at time 0 and appropriate intervals for each drug, which will account for approximately three drug half-lives post-administration. The collected fluid was immediately frozen at −80 • C until assayed.

Placement of Intestinal Ultrafiltration Probes
Surgical procedures took place 24-48 h prior to enrofloxacin administration over 4 days with three surgeries per day as previously described (Warren et al., 2014) with the only significant modification being the use of standing-restraint and local anesthesia instead of general anesthesia. Local anesthesia was provided by infusion 15 ml of 2% lidocaine above and below the transverse processes of lumbar vertebrae L1, L2, L3, and L4 in order to anesthetize the dorsal and ventral spinal nerves T13, L1 and L2. Adequacy of anesthesia was determined by pricking the skin of the flank prior to the incision. Steers received 2.2 mg/kg of flunixin meglumine intravenously immediately prior to surgery and 24 h later. The collecting loops of an ultrafiltration probe (UF-3-12, BAS; Bioanalytical Systems, West Lafayette, IN, USA) were inserted into the lumen of the ileum and spiral colon. The free ends of the probes were exteriorized through the abdominal wall. To collect the ultrafiltrate, a 3 mL evacuated tube with no additive (Vacutainer R , Becton-Dickinson) was inserted onto the needle of the vacuum vial needle holder at the end of the external tubing. Samples were collected from probes placed in the ileum and spiral colon at 0, 2, 4, 6, 8, 12, 24, 36, and 48 h post high-dose drug administration, and additionally 26, 28, 30, 32, 50, 52, 54, 56, 60, 72, 84, and 96 h after the first lowdose drug administration by changing the collection tubes at the predetermined time points. The collected fluid was aliquoted into cryogenic vials and immediately frozen at −80 • C until assayed.

Feces Collection
Feces was collected directly from the rectum of each steer 0, 12, 24, 36, 48, 72, 96, 120, 144, and 168 h post single-dose drug administration, and additionally 60, 84, and 192 h after the first multiple-dose drug administration. Samples were placed into bags (Whirlpak R , Nasco, Fort Atkinson, WI) and stored briefly on ice before microbiological analysis. An aliquot of each sample was stored in a 2 ml cryogenic vial and frozen at −80 • C. To avoid the residual effects of enrofloxacin on fecal bacteria between crossover arms of the study, only fecal samples collected from the first arm of each crossover study were analyzed and reported here.

Determining Active Drug Concentration
For plasma samples, we used solid-phase extraction, identical to the method in our previous papers (Davis et al., 2007). Plasma and tissue fluid samples were analyzed by reversephase high-pressure liquid chromatography with fluorescence (enrofloxacin and its metabolite ciprofloxacin) detection to determine the active concentrations of each drug as previously described (Davis et al., 2007;Foster et al., 2016). All drug concentrations were determined from calibration curves made from fortified blank plasma, intestinal and interstitial fluid collected from the experimental calves prior to antibiotic administration. Calibration curves were prepared from fortifying the blank matrix with reference drug standards of enrofloxacin and ciprofloxacin (United States Pharmacopeia, Rockville, MD) to validate the HPLC analysis and perform quality control assessments during the assay.

Pharmacokinetic Analysis
The drug concentrations were analyzed using standard pharmacokinetic methods to determine the drug disposition for each drug in each calf. A computer program (Phoenix, V. 6.1; Pharsight Corporation, Certara, St. Louis MO) was used to determine pharmacokinetic parameters as well as deriving statistical values.
Plasma, ISF, and intestinal drug concentrations were plotted on linear and semi-logarithmic graphs for analysis and for visual assessment of the best model for pharmacokinetic analysis. Specific models (e.g., one, two, etc. compartments) were determined for best fit based on visual analysis for goodness of fit and by visual inspection of residual plots. The best model fit was based on the equation described in the following formula: Where C is the plasma concentration, t is time, k 01 is the non-IV absorption rate, assuming first-order absorption, k 10 is the elimination rate constant, V is the apparent volume of distribution, and D is the non-IV dose. Secondary parameters calculated from the model included the peak concentration (C MAX ), time to peak concentration (T MAX ), area under the plasma-concentration vs. time profile (AUC), and the respective absorption and terminal half-lives (t½). Data from some calves were analyzed using noncompartmental analysis (NCA) due to sparse sampling using the same pharmacokinetic program described above. For the NCA, the area under the plasma concentration vs. time curve (AUC) from time 0 to the last measured concentration (defined by the LOQ), was calculated using the log-linear trapezoidal method. The AUC from time 0 to infinity was calculated by adding the terminal portion of the curve, estimated from the relationship C n /λ Z , to the AUC 0 Cn , where λ Z is the terminal slope of the curve, and C n is the last measured concentration point.
The relative drug transfer from the plasma compartment to the ISF and intestinal fluids was measured by calculation of a penetration factor. The penetration factor was determined by the ratio of AUC for the intestinal fluid to the AUC for plasma:

Quantification of E. coli From Feces
One gram of feces was inoculated into 9 ml EC broth (Oxoid Ltd., Basingstoke, Hampshire, England) and vortexed. One ml was removed and serially diluted 10-fold in sterile phosphatebuffered saline, and 100 µl was plated in triplicate onto HardyCHROM TM ECC Media (Hardy Diagnostics, Santa Maria, CA). Plates were incubated overnight at 37 • C and pinkviolet colonies were counted to determine the CFU/ml of E. coli. Dilutions that yielded between 30 and 300 colonies on each of the three plates were counted to quantify E. coli based on the dilution. These three replicates at the counted dilution were averaged to determine the quantity of E. coli at each time point. The remaining enrichment was incubated overnight at 37 • C, and if no growth was observed on direct plates, the samples were streaked for isolation on ECC plates and incubated overnight at 37 • C. From the quantified or enrichment plate, eight colonies were randomly picked and streaked onto Columbia agar with 5% sheep blood (Remel, Lenexa, KS) and incubated overnight at 37 • C. Following incubation, each isolate was transferred to a 2 ml cryogenic vial containing LB Broth (Sigma-Aldrich, St. Louis, MO) supplemented with 25% glycerol, vortexed, and frozen at −80 • C.

Determination of Minimum Inhibitory Concentration
The minimum inhibitory concentration (MIC) of each E. coli isolate to enrofloxacin was determined using broth microdilution according to Clinical and Laboratory Standards Institute guidelines (Clinical Laboratory Standards Institute, 2013). Each

DNA Extraction and 16S rRNA Sequencing
Fresh feces collected at each time point were frozen in 1 gram aliquots for future analysis. From these samples, 50 mg of feces were extracted individually using a MoBio PowerMag Microbiome kit (Qiagen, Inc., Germanton, MD) optimized for the epMotion 5075 TMX (Eppendorf, Hauppauge, NY). The DNA libraries were prepared as described previously (Seekatz et al., 2015).

Illumina MiSeq Sequencing of Bacterial Communities
The V4 region of the 16S rRNA gene was amplified from each sample using the Dual indexing sequencing strategy (Kozich et al., 2013 (Quail et al., 2008). PhiX and genomes were added in 16S amplicon sequencing to create diversity. Sequencing reagents were prepared according to the Schloss SOP, and custom read 1, read 2 and index primers were added to the reagent cartridge. FASTQ files were generated for paired end reads.

Microbiota Analysis
The median number of sequences per group was 24,866. Analysis of the V4 region of the 16S rRNA gene was done using mothur (version 1.37.4; Schloss et al., 2009). The standard operating procedure at http://www.mothur.org/wiki/ MiSeq_SOP was followed to process the MiSeq data. The pairedend reads were assembled into contigs and then aligned to the SILVA 16S rRNA sequence database (Pruesse et al., 2007;Quast et al., 2013) and classified to the family taxonomic level using the Wang method and an 80% bootstrap minimum (Wang et al., 2007) and the mothur-adapted RDP training set v9. Chimeric sequences were removed using UCHIME (Edgar et al., 2011). Sequences were clustered into operational taxonomic units (OTU) using a 3% sequence similarity threshold. The percentage relative abundance of bacterial Phyla and Family members in each sample was calculated. The inverse Simpson index on OTUs was used as a measure of alpha diversity. Additional high-resolution analysis was performed on exact amplicon sequence variants (ASVs) generated by the DADA2 method (Callahan et al., 2016(Callahan et al., , 2017. ASV generation followed FIGURE 3 | Median MIC of enrofloxacin in fecal E. coli isolates over time after treatment with either a single dose (12.5 mg/kg once) or multiple dose (5 mg/kg once a day for 3 days) of enrofloxacin. * indicates a significant difference from time 0, p < 0.05. the DADA2 tutorial (https://benjjneb.github.io/dada2/tutorial. html), but with samples pooled for the ASV inference step in order to increase sensitivity to rare sequence variants. E. coli ASVs were identified by exact matching to sequenced E. coli genomes.

Data Analysis
Differences in the E. coli concentration between groups were determined by a one way ANOVA. ANOVA on ranks was used to compare median MIC over time within treatment groups. A pvalue < 0.05 was considered significant. Differential abundance testing was performed using the DESeq2 package (Love et al., 2014), and following the recommendations for its application to microbiome sequencing data in McMurdie and Holmes (McMurdie and Holmes, 2014). Briefly, DESeq2 controls for the over-dispersion typical in microbiome sequencing data by fitting the abundance-variance relationship over all ASVs, and using that shrinkage estimator of the variance to test for significance. Testing for the treatment effect was performed on paired samples from each steer, before and after treatment, and after controlling for the steer. False discovery rate (FDR) was controlled using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995).

Deposition of Data
Sequence data will be deposited in the SRA database upon acceptance of the manuscript.

Pharmacokinetic Modeling
Tables 1, 2 depict the results of the PK analysis for enrofloxacin and ciprofloxacin for the single and multiple dose regimens. The ciprofloxacin concentration accounted for 25.8% of total fluoroquinolone concentrations in the single dose regimen and 26.7% of total fluoroquinolone concentrations in the multiple dose regimen. Ciprofloxacin penetration from the plasma into the ISF was two times greater than enrofloxacin in the single dose regimen and four times greater than enrofloxacin in the multiple dose regimen. Enrofloxacin penetration into the ileum from the plasma (single dose = 491.86%, multiple dose = 451.03%) and into the colon from the plasma (single dose = 385.29%, multiple dose = 293.32%) exceeded the penetration into the ISF from the plasma (single dose = 96.03%, multiple dose = 49.57%). Figure 1 depicts the combined concentrations of enrofloxacin and ciprofloxacin over time with the two dosing regimens in plasma, interstitial fluid, ileum and spiral colon.

Concentration of E. coli
The concentration of E. coli recovered at 12 (1.60 log 10 CFU/g), 24 (0.75 log 10 CFU/g) and 36 (1.11 log 10 CFU/g) h was significantly different from 0 h (6.17 log 10 CFU/g) in the single dose group (Figure 2A). By 72 h post-treatment, the E. coli concentration (4.72 log 10 CFU/g) had returned to baseline. In animals administered the multiple dose regimen, the concentrations at 24, 36, 48, 60 and 72 h were significantly different from 0 h (Figure 2A), and then returned to baseline by 96 h post-treatment. The correlation between culturable and sequenced E. coli concentrations was 0.557 over all samples, increasing to 0.808 when considering only those samples with >500,000 CFUs/g, which roughly corresponds to the minimum detectable frequency of 0.0001 for the 16S sequencing in this study (∼10 k reads per sample; Figure 2B). This correlation suggests that the culture results agree with the sequencing findings presented below.

E. coli Minimum Inhibitory Concentration
Prior to treatment, 84% E. coli isolates had an MIC of 0.03 mg/L or less, which is below the epidemiological cutoff value of 0.12 mg/L (Figure 3). The median MIC (0.029 mg/L) was not different between the two groups at time 0. In the single dose group, the median MIC significantly increased at 24 h posttreatment (0.5 mg/L), but returned to baseline by 48 h (0.03 mg/L). The median MIC of E. coli isolates from the multiple dose group significantly increased by 12 h after treatment (0.75 mg/L), and remained increased through 48 h (1.0 mg/L; Figure 3). When comparing individual time points across treatment regimens, the median MIC of E. coli isolates from the multiple dose group was significantly higher than the single dose group starting at 12 h and remained higher through the end of the study (Figure 3).

DISCUSSION
Enrofloxacin is commonly used for treatment of bovine respiratory disease (National Animal Health Monitoring System, 2011), but little is known about the relative impact of the two dosing regimens on AMR in fecal bacteria. In this study, subcutaneous administration of this fluoroquinolone at either dose resulted in high concentrations of active drug in the ileum and colon. Ciprofloxacin penetrates the ISF to a much larger extent than enrofloxacin. Our earlier studies showed that lower protein binding for ciprofloxacin (19 ± 5 vs. 46 ± 4%) might account for this difference in diffusion from plasma to interstitial fluid (Davis et al., 2007). However, this difference in protein binding does not appear to have the same effect on the penetration of enrofloxacin from the plasma to the intestine as enrofloxacin penetration was 450-490% into the ileum and 280-380% into the colon, similar to our previous study (Foster et al., 2016). This suggests that there may be a transporter protein or biliary excretion responsible for the high penetration of these fluoroquinolones into the intestine relative to both the plasma and ISF. Active secretion of fluoroquinolones into the intestine (Griffiths et al., 1993(Griffiths et al., , 1994Dautrey et al., 1999) and other tissues (Pulido et al., 2006;Alvarez et al., 2008;Mulgaonkar et al., 2013) has been demonstrated in other studies. Further, the difference in the AUC in the colon was greater than would be expected based on the different doses, indicating that this transport mechanism may be somewhat concentration dependent. The concentration of E. coli shows an inverse relationship with the concentration of enrofloxacin, as the single dose quickly lowered the E. coli concentration by 5 log 10 CFU/g of feces.
By 96 h post-treatment, E. coli fecal concentrations returned back to baseline for both the single and multiple dose regimens. This precipitous decline and rapid re-establishment of E. coli concentrations with low MIC values toward enrofloxacin is consistent with the "inverted U" concept to describe the effects of fluoroquinolone concentration on bacterial resistance (Tam et al., 2007). With high fluoroquinolone exposure, as we observed in the intestine of these cattle, E. coli were significantly, but temporarily suppressed (Figure 2A). By 96 h after treatment, the median MIC of the isolates from the single dose regimen reverted to the MIC of a wild-type population as the E. coli concentration rebounded. The return to baseline concentration and MIC corresponded with elimination of the drug from the intestine. In the multiple dose group, the median MIC increased to a greater degree than in the single dose group. Again, the change we observed in MIC was temporary, suggesting a lack of fitness for bacteria with a high MIC in the intestine of these cattle. This time for return to a wild-type MIC range is well within the slaughter withdrawal time of 28 days (Weatherbee, 2012) established by the FDA to prevent drug residues. Our results indicates that when FDA-approved slaughter withdrawal times are observed for enrofloxacin, it may mitigate the risk of transfer of resistant E. coli isolates through the food supply at slaughter. As we only assessed the MIC of E. coli in this study, it is unclear if these findings can be extrapolated to other enteric bacteria.
While evaluating changes to the complete fecal microbiome, we demonstrated only minor changes in community structure after treatment, and there was no change in microbial diversity. The transient changes that were seen were primarily in bacteria predominantly associated with the rumen microbiota (Prevotellaceae, Rikenellaceae, Christensenellaceae, and Erysipelotrichaceae; Mao et al., 2015). Further, the relatively small, short-lived changes found in steers treated with enrofloxacin is in contrast to findings in chickens (Li et al., 2017), mice (Choo et al., 2017), and humans (Stewardson et al., 2015;Pop et al., 2016) administered a fluoroquinolone that demonstrate significant changes at the phylum and genus levels which persist for a week or more. Potentially, the large microbial population of the rumen provides a source of microbes to continually repopulate the intestinal environment, leading to minimal changes in the fecal microbiota in ruminants compared to simple-stomached animals. A similar approach to measure active drug concentrations in the rumen and assess the impact of parenteral drugs on the rumen microbiome would be valuable to address this question. Alternatively, differences in the dosing regimens used in these studies compared those used in cattle may also influence these findings as each of these studies administered the drugs for longer durations and by a different route.
The study was conducted in healthy Holstein steers, and the PK of enrofloxacin or associated microbiological changes may not represent that of sick calves. Nonetheless, our results are similar to findings in the National Antimicrobial Monitoring System that suggest resistance to fluoroquinolones in bovine E. coli isolates is rare in cattle at slaughter (Food and Drug Administration, 2018). The PK findings and short duration of changes to fecal bacteria found in this study may help explain the rare incidence of fluoroquinolone-resistant E. coli isolates in cattle at slaughter (1.2% of isolates in beef cattle in 2015). Resistance to fluoroquinolones in Salmonella isolates was similarly uncommon in beef cattle at slaughter, while rates of fluoroquinolone-resistance in Enterococcus spp. and Campylobacter spp. tended to be higher. As we only examined the impact on fecal E. coli, extrapolation of these findings to other fecal bacteria may not be appropriate, and warrants additional investigation.
In conclusion, enrofloxacin penetrates the ileum and colon of cattle at high concentrations with either the single or multiple dosing regimens causing a significant reduction in fecal E. coli concentration that resolved within 4 days of treatment. With the multiple dose regimen, there was a small increase in median MIC. Yet, the median MIC returned to baseline by 2 weeks posttreatment. Assessment of the microbiome after treatment did not show clinically significant changes with either dosing regimen. From this study, it appears that dosing regimens in which antimicrobial concentrations rapidly achieve pharmacodynamic targets and are then quickly eliminated from the GI tract will minimize the effect on the fecal microbiota of cattle.

AUTHOR CONTRIBUTIONS
MJ, CT, MP, and DF contributed to study design, data collection and analysis, and manuscript preparation. KF, BC, and TP contributed to data collection and analysis, and manuscript preparation.