Unexpected Activity of Oral Fosfomycin against Resistant Strains of Escherichia coli in Murine Pyelonephritis

Fosfomycin tromethamine activity is well established for oral treatment of uncomplicated lower urinary tract infections, but little is known about its potential efficacy in pyelonephritis. Ascending pyelonephritis was induced in mice infected with 6 strains of Escherichia coli (fosfomycin MICs, 1 μg/ml to 256 μg/ml). The urine pH was 4.5 before infection and 5.5 to 6.0 during infection.

ABSTRACT

Fosfomycin tromethamine activity is well established for oral treatment of uncomplicated lower urinary tract infections, but little is known about its potential efficacy in pyelonephritis. Ascending pyelonephritis was induced in mice infected with 6 strains of Escherichia coli (fosfomycin MICs, 1 μg/ml to 256 μg/ml). The urine pH was 4.5 before infection and 5.5 to 6.0 during infection. Animals were treated for 24 h with fosfomycin (100 mg/kg of body weight subcutaneously every 4 h), and the CFU were enumerated in kidneys 24 h after the last fosfomycin injection. Peak (20.5 μg/ml at 1 h) and trough (3.5 μg/ml at 4 h) levels in plasma were comparable to those obtained in humans after an oral dose of 3 g. Fosfomycin treatment significantly reduced the bacterial loads in kidneys (3.65 log₁₀ CFU/g [range, 1.83 to 7.03 log₁₀ CFU/g] and 1.88 log₁₀ CFU/g [range, 1.78 to 5.74 log₁₀ CFU/g] in start-of-treatment control mice and treated mice, respectively; P < 10⁻⁶). However, this effect was not found to differ across the 6 study strains (P = 0.71) or between the 3 susceptible and the 3 resistant strains (P = 0.09). Three phenomena may contribute to explain this unexpected in vivo activity: (i) in mice, the fosfomycin kidney/plasma concentration ratio increased from 1 to 7.8 (95% confidence interval, 5.2, 10.4) within 24 h in vitro when the pH decreased to 5, (ii) the fosfomycin MICs for the 3 resistant strains (64 to 256 μg/ml) decreased into the susceptible range (16 to 32 μg/ml), and (iii) maximal growth rates significantly decreased for all strains and were the lowest in urine. These results suggest that local fosfomycin concentrations and physiological conditions may favor fosfomycin activity in pyelonephritis, even against resistant strains.

INTRODUCTION

Over the last 2 decades, resistance to β-lactams among the Enterobacteriaceae has emerged as a major public health threat. Isolates of Escherichia coli producing extended-spectrum β-lactamases are currently responsible for a large proportion of urinary tract infections (UTIs) in the community as well as in the health care setting (1–3). In the current era of the increasing prevalence of antibiotic resistance, fosfomycin has attracted renewed interest for the treatment of infections caused by multidrug-resistant pathogens and especially UTIs. Indeed, it has broad-spectrum antimicrobial activity and a favorable safety profile (4, 5). Fosfomycin tromethamine, a soluble salt with improved bioavailability over that of fosfomycin, is currently recommended at a single dose as the first-line drug for the treatment of uncomplicated lower UTIs in Europe (6) and in the United States (7). However, it is unknown whether fosfomycin tromethamine would be useful for the treatment of pyelonephritis in humans, raising the questions of kidney diffusion and the breakpoints to be used in pyelonephritis compared with cystitis. Indeed, the current susceptibility breakpoint of fosfomycin for Enterobacteriaceae is an MIC of 32 μg/ml, according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (8).

We previously demonstrated that fosfomycin resistance in E. coli strains from various genetic backgrounds was associated with a decrease of in vitro fitness and in vivo virulence in a murine model of pyelonephritis (9). In addition, some mutations conferring fosfomycin resistance have been shown to decrease pilus biosynthesis and bacterial adhesion to epithelium cells (10, 11). Furthermore, Martín-Gutiérrez et al. recently demonstrated that anaerobiosis and acidic pH values in urine decreased fosfomycin MICs in strains harboring chromosomal resistance mutations (12). Finally, it has previously been shown that high concentrations (1,000 to 4,000 μg/ml) were achieved in urine after oral administration of 3 g of fosfomycin tromethamine in humans and remained above 100 μg/ml for at least 30 to 40 h (5), but data on the concentrations in kidneys are scarce.

However, possible limitations for the use of the oral formulation of fosfomycin tromethamine in pyelonephritis may be anticipated. These are related to the potentially low concentrations in kidneys compared to those in urine, which in turn might be associated with limited bactericidal activity and a risk of selection of resistant mutants. Indeed, the selection of spontaneous fosfomycin-resistant mutants occurs at a very high rate in vitro (between 10⁻⁷ and 10⁻⁶ cells among Gram-negative bacteria) (13).

Therefore, the aim of the present study was to investigate the activity of fosfomycin in a murine model of pyelonephritis due to E. coli, using a dosing regimen that reproduced plasma concentrations comparable to those obtained in humans with the oral formulation of fosfomycin tromethamine. Different strains with increasing MICs were used in order to define the in vivo activity of fosfomycin tromethamine according to strain susceptibility in this specific infection.

RESULTS

Bacterial study strains.

The characteristics of the bacterial strains used in the study are shown in Table 1. Three strains were susceptible to fosfomycin, and three strains were resistant, according to EUCAST breakpoints (32 μg/ml). Mutations in genes known to be involved in fosfomycin resistance (uhpB, uhpC, glpT, and cyaA) were detected in all strains, including those categorized as susceptible (B175 and B56) according to the current EUCAST breakpoint (Table 1). All but one strain (E. coli B56, producing an extended-spectrum β-lactamase) were susceptible to other antibiotic families.

TABLE 1.

Susceptibility to fosfomycin of E. coli strains used in this study, according to the pH

Strain	Origin	Phylogenetic group	Fosfomycin MIC (μg/ml) at pH:			Susceptibility categorization at pH 7.2/pH 5a	Fosfomycin resistance mutation(s) detected
			7.2	6	5
CFT073	Urine clinical isolate	B2	1	1	1	S/S	None
B56	Urine clinical isolate	B2	8	8	16	S/S	UhpB (P169S)
B175	Urine clinical isolate	B2	32	32	32	S/S	CyaA (L125F), UhpC (T72I)
MUT2	Mutant from CFT073	B2	64	16	16	R/S	UhpB (G469R)
C05	Urine clinical isolate	B2	128	64	32	R/S	UhpC (Q210X)
C114	Urine clinical isolate	B2	256	32	32	R/S	UhpC (Q132X), GlpT (G144D)

^aSusceptibility (S) and resistace (R) were determined according to EUCAST breakpoints (8).

Fosfomycin concentrations in plasma and kidneys.

Fosfomycin concentrations in plasma and kidneys after a single injection of 100 mg/kg of body weight subcutaneously (s.c.) are shown in Table 2. While the fosfomycin concentrations were comparable in the plasma and kidneys at 1 h following the injection, fosfomycin concentrations were higher in the kidneys than in the plasma over time, with mean concentrations at 24 h being <1 mg/liter in plasma and 6.8 mg/liter in kidneys, and the kidney/plasma concentration ratio was 7.8 (95% confidence interval [CI], 5.2, 10.4) (Fig. 1). After a single injection, the area under the concentration-time curve (AUC) from 0 to 24 h (AUC_0–24) in the kidney was estimated to be 166.6 mg · h/liter, being 2-fold higher than that in plasma, where it reached 85.0 mg · h/liter. The fosfomycin AUC_0–24 in plasma and kidneys after a 100-mg/kg injection every 4 h for 24 h were 357.1 and 614.5 mg · h/liter, respectively, and the AUC in plasma from 0 to 44 h, which was the time of sacrifice (AUC_0–44), was 470.3 mg · h/liter.

TABLE 2.

Fosfomycin concentrations in plasma and kidneys from CBA female mice after a single injection of 100 mg/kg given subcutaneously

Time (h)	Mean (range) fosfomycin concn (μg/ml) ina :
	Plasma	Kidneys
1	20.5 (9.0 to 33.5)	20.8 (11.4 to 27. 5)
4	3.5 (2.8 to 4.8)	8.4 (6.7 to 11.4)
24	<1 (<1 to <1)	6.8 (4.0 to 10.1)

^aEach set of values for a given sample time corresponds to the mean (range) for 4 mice.

FIG 1.

Ratio of the kidney/plasma fosfomycin concentrations in CBA mice after a single injection of 100 mg/kg s.c. The curve represents the mean ratio predicted by the model. The dark area represents the 95% confidence interval of the prediction. Each point represents the observed ratio for each sample.

Fosfomycin antimicrobial effect in murine pyelonephritis.

The bacterial loads in kidneys according to fosfomycin treatment and the infective strain are presented in Fig. 2 and Table 3. Fosfomycin treatment significantly reduced the bacterial loads in the kidneys (P < 10⁻⁶). The median bacterial loads in the kidneys were 3.7 log₁₀ CFU/g (range, 1.8 to 7.0 log₁₀ CFU/g) in start-of-treatment control mice and 1.9 log₁₀ CFU/g (range, 1.8 to 5.7 log₁₀ CFU/g) in fosfomycin-treated mice. However, there was no significant difference in the kidney bacterial loads according to the infective strain (P = 0.71) or between the 3 susceptible and the 3 resistant strains (P = 0.09). The interaction between the infective strain and the fosfomycin effect was not significant (P = 0.53). Similarly, the proportion of sterile kidneys in treated mice did not differ between the 6 strains (P > 0.8) (Table 3). No fosfomycin-resistant mutant was detected in kidneys at the time of sacrifice for any of the 6 strains.

FIG 2.

Bacterial counts (number of log₁₀ CFU per gram of kidney) of E. coli in mice with pyelonephritis treated or not for 24 h with fosfomycin at 100 mg/kg s.c. every 4 h according to study strain. The no-fosfomycin-treatment groups corresponded to the start-of-treatment control mice. Each circle represents a mouse; horizontal bars represent median values. Fosfomycin treatment significantly reduced the bacterial loads in the kidneys (P < 10⁻⁶). The effect of the infective strain on the bacterial load in the kidney was not significant (P = 0.53), nor was the interaction between the type of strain and the fosfomycin effect (P = 0.71).

TABLE 3.

PK/PD of fosfomycin in murine pyelonephritis due to E. coli after a 24-h treatment^a

Strain	FOS MIC (μg/ml)	FOS AUC0–24/MIC		Start-of-treatment control mice		FOS-treated mice
		Plasma	Kidney	Median (range) log10 CFU count/g of kidneys	No. of sterile mice/total no. of mice tested	Median (range) log10 CFU count/g of kidneys	No. of sterile mice/total no. of mice tested
CFT073	1	357.1	614.5	3.9 (1.9–5.3)	0/11	2.0 (1.8–5.1)	4/9
B56	8	44.6	76.8	3.0 (2.5–6.5)	0/11	1.8 (1.8–2.8)	4/10
B175	32	11.2	19.2	3.4 (2.1–7.0)	0/11	1.9 (1.8–4.9)	4/10
MUT2	64	5.6	9.6	3.7 (2.2–6.8)	0/12	1.9 (1.8–5.5)	5/10
C05	128	2.8	4.8	3.7 (2.3–4.9)	0/8	2.7 (1.9–4.1)	3/10
C114	256	1.4	2.4	3.7 (1.8–5.8)	0/14	1.9 (1.8–5.7)	6/10

^aFOS, fosfomycin. The median bacterial loads in kidney were 3.7 log₁₀ CFU/g (range, 1.8, 7.0 log₁₀ CFU/g) in the start-of-treatment control mice and 1.9 log₁₀ CFU/g (range, 1.8, 5.7 log₁₀ CFU/g) in the fosfomycin-treated mice (P < 10⁻⁶). The lower limit of detection was 1.8 log₁₀ CFU/g of kidney.

Pharmacokinetic (PK)/pharmacodynamic (PD) analysis.

The fosfomycin AUC_0-24/MIC ratios in plasma and kidneys according to the strain are shown in Table 3. The relationship between the bacterial loads in kidneys and the AUC_0-24/MIC ratios in plasma or kidney and the AUC_0-44/MIC ratios in plasma was not significant (P = 0.45 for both plasma and kidney).

Effect of pH on fosfomycin activity and bacterial growth rate.

In order to explain the unexpected activity of fosfomycin against both susceptible and resistant strains, investigations were performed to test the influence of acidic pH, as 85% of patients with a UTI due to E. coli had a urine pH of ≤6.5 (12). Indeed, the pH in uninfected urine from 5 CBA mice before experimental pyelonephritis was 4.5 and ranged from 5.5 to 6.0 in the same mice after 48 h of experimental pyelonephritis.

(i) Effect of pH on in vitro fosfomycin activity.

Low pH values increased fosfomycin activity among fosfomycin-resistant strains (Table 1). Indeed, when the pH was decreased from 7.2 to 5, fosfomycin MICs against the 3 resistant strains decreased from 64 to 256 μg/ml to 16 to 32 μg/ml, which corresponded to susceptibility according to EUCAST (8).

(ii) Effect of pH and urine on in vitro bacterial growth rate.

At pH 7, the maximal growth rates (MGRs) differed among the 6 studied strains, with median values ranging from 3.29 to 4.20 h⁻¹ (P < 0.01). For each of the 6 strains, the MGR was reduced as the pH was lower (Fig. 3). This reduction was statistically significant (P < 0.05) between pH 7 and pH 5 for all strains except C05. The lowest MGR values were observed in urine, with a significant reduction compared with the MGR determined at pH 7 being found for all strains (P < 0.05). For each of the 6 strains, the time to achieve MGR was prolonged at pH 5 compared with pH 7, and this difference was statistically significant (P < 0.05) for all strains except C05 (Fig. 3).

FIG 3.

In vitro maximal growth rates (top) and the time to achieve the maximal growth rate (bottom) for each study strain according to pH in Luria-Bertani or in urine. Each value is the median for three independent experiments. The brackets on the bars represent standard deviations. MGR was significantly reduced (P < 0.05) in urine compared with that at pH 7 for all strains and at pH 5 for all strains except C05. The time to achieve MGR (T_max) was significantly prolonged (P < 0.05) at pH 5 compared with that at pH 7 for all strains except CO5.

DISCUSSION

Fosfomycin has been approved for decades as an injectable antibiotic for the treatment of systemic infections due to both Gram-positive cocci and Gram-negative bacilli, including severe infections (4, 14, 15). More recently, an oral formulation, fosfomycin tromethamine, has rapidly become the first-line empirical treatment recommended worldwide as a single-dose treatment for acute uncomplicated lower UTIs (6, 7).

Here we show that, when given in a regimen generating peak and trough levels in plasma in the range of those obtained in humans after a single dose of the oral formulation of fosfomycin tromethamine (approximately 20 μg/ml and 5 μg/ml, respectively) (16, 17), fosfomycin produced a significant reduction of the bacterial load after 24 h of treatment in kidneys from mice with pyelonephritis. However, the unexpected result was that fosfomycin efficacy was similar whether the study strains were susceptible or resistant to fosfomycin according to EUCAST breakpoints (8).

The fosfomycin activity observed in infected kidneys may result from the combination of local specific physiological conditions in urine and fosfomycin concentrations in the kidneys.

Indeed, several authors have already shown that low pH values increased the fosfomycin in vitro activity (12, 18, 19). In an acidic environment, such as urine, fosfomycin is partially protonated, thus being in a more lipophilic state, allowing fosfomycin entry into bacteria and resulting in higher antimicrobial activity (19). Our results support these findings, as low pH values were associated with a significant decrease in fosfomycin MIC values for resistant strains into the susceptible range (Table 1). Factors other than acidic pH may alter the growth of E. coli in urine, such as the lack of essential sources of metabolic elements, like urea, creatinine, iron, citric acid, d-serine, or ammonia (20).

In the present study, we also confirmed in vitro that low pH values were associated with a decreased in vitro fitness for fosfomycin-susceptible or -resistant strains, as previously reported (12), and showed that this decrease was maximal in urine for all the tested strains (Fig. 2).

Chromosomal mutations conferring fosfomycin resistance have been associated with a high biological cost, entailing a reduced fitness (11). This fitness cost is of particular interest in uropathogenic E. coli strains, because if the biological cost is high, the resistant bacteria will not grow at the minimal rate needed to establish infection or to invade the kidney (21, 22). We have recently shown in the same model of ascending murine pyelonephritis due to uropathogenic E. coli strains belonging to the B2 phylogenetic group that there was a significant reduction in kidney infection rates with fosfomycin-resistant isolates compared with susceptible ones (9). This phenomenon may, of course, have favored fosfomycin antimicrobial activity in mouse pyelonephritis and may have contributed to limit the selection of fosfomycin-resistant mutants, as observed in the present study in mice, with fosfomycin resistance not being detected for any strain tested.

From a pharmacokinetic point of view, fosfomycin concentrations over time were higher in kidneys than in plasma. This increase in drug exposure obviously favored fosfomycin activity in pyelonephritis. From a PK/PD point of view, recent studies done in mice demonstrated that the PK/PD index that best predicted fosfomycin efficacy in a thigh model and in a UTI model was the AUC/MIC ratio (23, 24). More specifically, in the thigh infection model, for E. coli, net stasis was observed at a median AUC/MIC ratio value of 19.3, and 1-log kill was observed at a median AUC/MIC ratio value of 97.5 (23). According to these data, since a similar dosing regimen was used in the present study in all mice and generated in plasma the same AUC_0–44 of 470.3 mg · h/liter at the time of sacrifice, the target value for net stasis would be achieved only for strains with an MIC of 24 μg/ml or lower and the 1-log kill target would be achieved only for strains with an MIC of 4 μg/ml or lower. The higher than expected fosfomycin activity observed in our model (Table 3) is probably the consequence of the overall favorable local physiological conditions discussed above, which reduced the MIC, fitness, and virulence of resistant strains and which was associated with kidney concentrations favoring fosfomycin activity in pyelonephritis.

The lack of correlation between the AUC/MIC ratio in plasma and the number of CFU in kidneys that we observed was related to the fact that, on the one hand, all the mice were treated with the same dosing regimen and were therefore exposed to the same AUC and that, on the other hand, the local MICs of resistant strains in urine became very similar to those of susceptible strains. Thus, as the AUC was similar for all strains and the MICs under local conditions were not different from those under standard conditions (Table 1), the AUC/MIC ratio reached a plateau that precluded investigation of the PK/PD relationship and did not allow us to determine an MIC breakpoint.

Several limitations of our study, due to experimental conditions, should be taken into account when extrapolating the data, as they could limit the probability for selection of fosfomycin-resistant mutants in vivo: (i) the bacterial loads in kidneys before fosfomycin treatment were moderate, and no spontaneous mortality was observed in this model; (ii) the duration of fosfomycin treatment was limited to 24 h, a short duration of time to select for resistant mutants; however, it must be acknowledged that this is very similar to the situation in humans after an oral single dose of fosfomycin; (iii) the fosfomycin dosing regimen used in the present study generated an AUC_0–24 that corresponded approximately to what would be obtained in humans after two doses of oral fosfomycin tromethamine (14, 15, 17), as repeated oral doses is a dosing regimen that is under investigation for situations other than cystitis in women (18); and (iv) finally, we did not include a group of end-of-treatment control mice, which would have helped us to analyze the relative part of the spontaneous bacterial clearance from the kidneys and the antibacterial effect of fosfomycin.

In conclusion, our results suggest that the fosfomycin tromethamine oral formulation might be of interest in humans for the treatment of UTI with parenchymal infections, such as pyelonephritis, due to the favorable local physiological conditions and kidney concentrations. Our results suggest that repeated dosing of fosfomycin tromethamine should be investigated for the treatment of uncomplicated pyelonephritis in women.

MATERIALS AND METHODS

Bacterial strains.

Six bacterial strains of E. coli were used (Table 1), with the MICs of fosfomycin for these strains ranging from the susceptible to the resistant range (1 to 256 μg/ml). The reference wild-type E. coli CFT073 (O6:K2:H1) strain (24) was previously used to set up a murine model of pyelonephritis by our group (25, 26). The other strains were clinical isolates from urinary tract infections (UTIs), except for E. coli MUT2, which was a fosfomycin-resistant mutant selected in vitro from E. coli CFT073. We selected strains belonging to phylogenetic group B2 since such strains are most frequently responsible for UTIs in humans and carry the most important virulence factors (27, 28).

In vitro fosfomycin activity.

The MICs of fosfomycin (Sanofi-Aventis, Paris, France) were determined by the dilution method in Mueller-Hinton agar (pH 7.2), in accordance with EUCAST guidelines (9), with 25 μg/ml glucose-6-phosphate (G6P; Sigma-Aldrich, Saint-Quentin Fallavier, France) added in the medium. The MICs of fosfomycin were also determined at pH 5 and 6. Each in vitro experiment was replicated in at least three independent experiments, and the median values are reported for each strain.

Mechanisms of fosfomycin resistance.

Known mechanisms of fosfomycin resistance were determined for each strain exhibiting an MIC of fosfomycin of ≥8 μg/ml, as several of them have been found in strains with MICs from the susceptible range (9). Mutations in the genes involved in fosfomycin chromosomal resistance (i.e., murA, glpT, uhpT, cyaA, ptsI, uhpA, uhpB, uhpC) were determined by nucleotide sequencing after amplification by PCR, as previously described (9). The amino acid sequences were compared with those of E. coli CFT073 and K-12.

In vitro bacterial growth rate.

Growth rates at 37°C were measured in Luria-Bertani (LB) broth with various pHs (pH 5, 6, and 7) and in sterile-filtered pooled human male urine (pH 6), as previously described (9). For each strain and condition, the maximal growth rate (MGR) and the time to achieve MGR (T_max) were measured in three independent experiments, and the median values are reported for each strain.

Murine pyelonephritis model.

We used the ascending, unobstructed UTI mouse model previously developed by our group (25, 26). The pyelonephritis protocol (no. APAFIS#4950-2016021211417682 v4) was approved by the French Ministry of Research and by the ethical committee for animal experiments. Eight-week-old immunocompetent CBA female mice (weight, 20 to 23 g) were used. Bacterial inocula were obtained by overnight incubation in LB broth, followed by centrifugation at 8,000 × g for 15 min. Pellets were suspended in 1 ml of sterile saline solution to a final inoculum of 10⁹ CFU/ml. Pyelonephritis was induced after general anesthesia (with intraperitoneal administration of 150 mg/kg of body weight of ketamine and 0.5 mg/kg xylazine) by injecting 50 μl (10⁷ CFU of E. coli) into the bladder through a urethral catheter. Urine was sampled for pH determination in 5 mice just before bacterial inoculation and at 48 h after infection. For each strain, 18 to 24 mice were infected. At 2 days after inoculation, 8 to 14 mice were sacrificed before treatment (start-of-treatment controls) and 9 to 10 mice were treated over 24 h by subcutaneous injections (fosfomycin at 100 mg/kg every 4 h for 24 h). In order to avoid unnecessary killing of mice, we did not constitute a group of untreated mice (end-of-treatment controls) because a previous study with this model showed that the bacterial counts in the kidney were stable for at least 5 to 10 days (25). Treated mice were sacrificed at 24 h after the last antibiotic injection to avoid a carryover effect. Kidneys were aseptically removed and were homogenized in 1 ml of saline solution. Then, 100 μl of the solution or its dilution was spread onto Mueller-Hinton (MH) agar plates and incubated at 37°C for 24 h. Selection of resistant mutants after in vivo exposure was sought by plating 100 μl of kidney homogenates onto MH agar containing fosfomycin at a concentration of 4 times the MIC for fosfomycin-susceptible strains and 2 times the MIC for fosfomycin-resistant strains. The kidneys were considered sterile if no colony grew on the agar plate. In the absence of bacterial growth, the number of log₁₀ CFU per gram of kidney was calculated by considering the growth of one colony and the weight of the kidney as the method detection limit. This corresponded to approximately 1.8 log₁₀ CFU/g of kidney. Assessment criteria for each strain were as follows: (i) the bacterial load in the kidneys, expressed as the number of log₁₀ CFU per gram, and (ii) the percentage of sterile kidneys.

Fosfomycin dosing regimen.

Single-dose plasma pharmacokinetic studies were performed on 8-week-old CBA female mice (weight, 20 to 23 g) in order to determine the therapeutic regimen that best reproduced the peak and trough plasma levels obtained in humans with a daily oral dose of 3 g of fosfomycin tromethamine (20 and 5 μg/ml, respectively) (14, 17, 18). The dosing interval was chosen accordingly to reproduce the area under the concentration-time curve (AUC) in the range of that obtained in humans with a single oral dose of 3 g of fosfomycin tromethamine (i.e., up to 228 mg · h/liter, compared with a range of 1,400 to 1,800 mg · h/liter in human with the intravenous route) (5, 14, 17), since the fosfomycin AUC/MIC ratio is the PK/PD index most closely linked to in vivo efficacy (23). This regimen was determined to be 100 mg/kg every 4 h (14, 17, 18). Total drug concentrations were utilized in the PK/PD analyses, as fosfomycin is not bound to plasma protein (14, 17).

Fosfomycin sampling in plasma and kidney.

Blood samples of at least 500 μl were obtained by intracardiac puncture from 4 anesthetized mice at 5 different intervals after a single subcutaneous injection of fosfomycin (100 mg/kg): 1 h, 2 h, 4 h, 6 h, and 24 h. After blood collection, plasma was separated by centrifugation. Kidneys were also removed at the same sampling times.

Fosfomycin assays in plasma and kidneys.

The concentrations of fosfomycin were determined using a triple-quadrupole mass spectrometer, AquityÒ TQD (Waters, St, Quentin en Yvelines), operated with negative electrospray ionization. Instrument parameters were optimized for fosfomycin (137 → 79 m/z) and propylphosphonic acid (the internal standard; 123 → 79 m/z) transitions. Fosfomycin was extracted from samples (plasma and kidneys) via protein precipitation with acetonitrile. The chromatographic separation took place on an AcquityÒ UPLC BEH HILIC column with dimensions of 100 mm by 2.1 mm and a particle size of 1.7 μm (Waters, St. Quentin en Yvelines). The lower limit of quantitation was 1 mg/liter. Intraday and interday coefficients of variation in plasma were 8.5 and 11.1%, respectively, at concentrations ranging from 5 to 250 μg/ml. The possible influence of coextracted matrix compound on the detectability of target analyses was checked.

Fosfomycin PK/PD analysis.

For the PK analysis, concentration data for fosfomycin in plasma and kidney were separately fitted to 2-compartment models with extravascular administration (29), pooling the data for all mice and using the optim package of R statistical software (v3.4.0). For each fit, the pharmacokinetic model could thus be written as C = Ae^−αt + Be^−βt − (A + B)e^−k_at. The parameters to be estimated were A and B, the first and second macroconstants, respectively. C is concentration; α and β are the first and second rate constants, respectively; t is time; and k_a is the absorption rate constant. Data below the lower limit of quantitation (LLOQ) were imputed to the LLOQ. We derived the AUC_0–24 in plasma (AUC_0-24,plasma) and kidney (AUC_0-24,kidney) and the AUC in plasma to the time of sacrifice (AUC_0–44) from the model with administration of fosfomycin at 100 mg/kg every 4 h. Finally, AUC_0–24/MIC ratios in plasma and kidneys were determined for each strain used in the murine pyelonephritis model. We studied the relationship between the bacterial load in kidney and the AUC_0–24/MIC ratios in plasma or kidney using linear regressions.

Statistical analysis.

The MGRs of the bacterial strains at pH 7 were determined, and the effect of pH on MGRs and the time to achieve MGR were studied using a Kruskal-Wallis nonparametric test.

A 2-way analysis of variance was performed to study the effect of the infective strains and fosfomycin treatment on the bacterial loads in the kidneys and to test the interaction between the infective strain susceptibility and the treatment effect. We also compared the proportion of sterile kidneys in fosfomycin-treated mice according to the infective strain using a Fisher exact test. The bacterial loads in the kidneys from treated mice were compared between the 3 susceptible strains and the 3 resistant strains by the Mann-Whitney nonparametric test. The relationship between the bacterial load in the kidney and the AUC_0–24/MIC and AUC_0–44/MIC ratios in plasma and the AUC_0–24/MIC ratio in kidneys was studied using linear regression.

The type I error was set at 0.05, and two-tailed tests were used. All analyses were performed using R statistical software.