Carbapenem-Containing Combination Antibiotic Therapy against Carbapenem-Resistant Uropathogenic Enterobacteriaceae

The increasing global prevalence of carbapenem-resistant Enterobacteriaceae (CRE) combined with the decline in effective therapies is a public health care crisis. After respiratory tract infections, urinary tract infections and associated urosepsis are the second most affected by CRE pathogens.

ABSTRACT

The increasing global prevalence of carbapenem-resistant Enterobacteriaceae (CRE) combined with the decline in effective therapies is a public health care crisis. After respiratory tract infections, urinary tract infections and associated urosepsis are the second most affected by CRE pathogens. By using checkerboard analysis, we tested eight different antibiotics in combination with carbapenems in CAMHB (cation-adjusted Müller-Hinton broth) and artificial urine against seven CRE strains and three susceptible strains. To further determine whether these combinations are also effective in a dynamic model, we have performed growth curves analyses in a dynamic bladder model with three uropathogenic CRE strains. In this model, we simulated the urinary pharmacokinetic after application of 1,000 mg intravenous (i.v.) ertapenem alone or in combination with 500 mg i.v. levofloxacin, 1,000 mg oral rifampin, or 3,000 mg oral fosfomycin. Bacterial growth was measured for 48 h, simulating voiding of the bladder every 3 h. According to the median fractional inhibitory concentration indices (ΣFICIs), the values we found were additive to synergistic results across all tested CRE strains for combinations of carbapenems with colistin sulfate, levofloxacin, fosfomycin, rifampin, and tigecycline in CAMHB and artificial urine. In the dynamic bladder model, all three CRE strains tested showed regrowth after treatment with ertapenem up to 48 h. Regrowth could be prevented by combination with levofloxacin, fosfomycin, or rifampin. Carbapenem-containing combination therapy with fosfomycin or rifampin could be an option for better treatment of urinary tract infections (UTIs) caused by CRE strains. This should be further investigated in clinical studies.

INTRODUCTION

Community- and hospital-acquired urinary tract infections (UTIs), which are caused by Gram-negative bacteria, especially Enterobacterales, are among the most common reasons for medical consultations and antibiotic prescriptions. We have experienced a continuous emergence of antibiotic resistance in UTIs and a shortage of new antibiotics under development (1). Resistances against β-lactam antibiotics, fluoroquinolones, and aminoglycosides are among the most widespread ones in uropathogenic bacteria in Europe (2). A resulting increased use of carbapenems in turn leads to a rise of resistance to this antibiotic class (3). It is mandatory to find solutions to break established resistance to antibiotics or enhance the antimicrobial activity of antibiotics to reduce emergence of novel resistance. Here, the combination of established antibiotics can be useful to combat existing resistances and to reduce resistance development. In addition, antibiotic combinations are used to broaden the spectrum or, in some cases, to immunomodulate or attenuate virulence. Several combinations of antibiotics were already used in the clinics, especially for resistant Pseudomonas aeruginosa, Acinetobacter baumannii, and Klebsiella pneumoniae carbapenemase (KPC). However, the clinical evidence is often ambivalent or not reliable (4). Combination therapies are mainly applied on the basis of knowing the effectiveness of the individual antibiotic without knowing whether a combination of both substances is beneficial. Recently, a large-scale study showed that most combinations of antibiotics are by no means additive or synergistic, but in some cases even antagonistic, thus even decreasing the effects of the individual antibiotics (5). In the worst case, a wrongly chosen combination can lead to an increased resistance formation in addition to a failed therapy, since the use of multiple antibiotics increases the antimicrobial pressure on the bacteria. Even in cases where there are positive in vitro data, these are often difficult to transfer to the patient situation since standard nondynamic antimicrobial susceptibility tests do not allow measurement of bacterial growth over time in accordance to the antibiotic-specific pharmacokinetic (PK). Furthermore, recommended broth for antimicrobial susceptibility tests (cation-adjusted Müller-Hinton broth [CAMHB]) differs in composition (e.g., cationic concentration) from a physiological medium such as urine. Thus, we recently showed that mcr1-positive Escherichia coli strains are significantly higher resistant to colistin in artificial urine compared to CAMHB or human serum (M. Loose, K. G. Naber, A. Coates, F. Wagenlehner, Y. Hu, submitted for publication).

The aim of this study was to combine carbapenems with eight other antibiotics to screen via checkerboard assays for synergistic effect against six carbapenem-resistant and three carbapenem-susceptible Enterobacteriaceae strains in CAMHB and artificial urine. Furthermore, promising combinations were tested for efficacy in a dynamic bladder model.

RESULTS AND DISCUSSION

MIC determination.

The three employed carbapenem-susceptible strains were in addition sensitive to all other tested antibiotics according to EUCAST breakpoints (Table 1) (6). For rifampin, no breakpoints are available for Enterobacteriaceae. The six carbapenem-resistant strains were in general also resistant against fluoroquinolones (ciprofloxacin and levofloxacin) and the aminoglycoside tobramycin (Table 1). Recently, the European Centre for Disease Prevention and Control (ECDC) gave a warning that E. coli and K. pneumoniae strains showing resistance against carbapenems in combination with fluoroquinolone and aminoglycoside are increasing (7). The majority of the carbapenem-resistant Enterobacteriaceae (CRE) strains showed also increased nitrofurantoin MIC values compared to the carbapenem-susceptible strains. An increase in nitrofurantoin resistance in CRE strains is worrying as nitrofurantoin has been reintroduce alongside with carbapenems and colistin to combat antibiotic-resistant infections. MIC determinations showed a reduced activity (P < 0.05 for 2/3 strains) of tobramycin in artificial urine (AU) compared to CAMHB against the carbapenem-susceptible strains (Table 1). Recently, we already showed a pH-dependent activity of tobramycin in urine (8), which most likely results from a reduced uptake owing to the fall of the proton motive force under acidic conditions (9). Some of the CRE-strains showed a significant increased fluoroquinolone MIC in AU compared to CAMHB (Table 1). Fluoroquinolones had been previously described as less active in artificial urine (10). Both more acidic pH and high cation concentration in the artificial urine compared to CAMHB interfere with antibiotic activity (16, 17). Using AU generally reduced tigecycline activity compared to CAMHB (P < 0.05 for 4/9 strains). Tigecycline is more prone to nonenzymatic epimerization at a lower pH, which could result in loss of activity in urine with acidic pH (18). In contrast, carbapenems were more potent in AU compared to CAMHB for some of the strains. Lemaire et al., who showed increased activity of meropenem against methicillin-resistant Staphylococcus aureus (MRSA) strains at acidic pH values (19), assumed an enhanced binding of carbapenems to the penicillin binding proteins at low pH. The mcr-1-positive, colistin-resistant E. coli NRZ14408 strain showed a significant increase of the colistin MIC in AU compared to CAMHB and increased rifampin MIC values compared to the other strains in both media (Table 1). We recently showed that mcr-1-mediated resistance against colistin was significantly increased in AU due to lower pH and higher cation concentrations compared to CAMHB (Loose et al., submitted).

TABLE 1.

Median MICs in CAMHB and artificial urine

Antibiotica	MIC (mg/liter) in CAMHB/AU forb :
	Carbapenem-susceptible strains			Carbapenem-resistant strains
	E. coli ATCC 25922	K. pneumoniae 595	E. cloacae CHD57	E. coli IR3	E. coli NRZ14408	E. coli BAA2469	K. pneumoniae BAA1705	K. pneumoniae BAA2146	E. cloacae BAA2468
MEPM	0.063/0.063	0.25/0.25	0.25/0.5	32/16	16/16	16/16	128/64	64/128	128/8*
ERT	0.016/0.031	0.031/0.016	0.25/0.25	256/64*	64/16*	64/32	256/128	256/128	128/32*
CIP	0.5/0.25	0.5/0.5	0.5/0.5	16/256*	16/256	128/128	128/256	256/256	64/256*
LVX	0.5/0.5	0.5/0.5	0.25/0.5	16/128	16/64	16/128*	128/128	128/128	32/128*
CS	1/1	2/4	4/2.5	2/16	8/64*	1/16	1/2	2/4	2/2
FOS	32/64	256/512	128/512	256/64	16/128*	32/256	512/512	256/512	1,024/1,024
NFT	16/8	16/32	64/128	64/16	64/32	16/8	128/128	128/128	128/128
RIF	8/16	8/48	32/48	32/32	256/256	8/16	32/64	32/32	32/64
TGC	0.25/1	1/2	4/8	0.25/1*	0.5/2*	0.5/2*	2/4	2/16*	4/8
TOB	2/8*	2/8	1/2*	128/128	128/128	128/128	128/128	128/128	128/128

^aMEPM, meropenem; ERT, ertapenem; CIP, ciprofloxacin; LVX, levofloxacin; CS, colistin sulfate; FOS, fosfomycin; NFT, nitrofurantoin; RIF, rifampin; TGC, tigecycline; TOB, tobramycin.

^b*, P < 0.05 over MIC in CAMHB.

Assumptions suggesting that antibiotic concentrations reached in urine are high enough to overcome even resistant pathogens are often based only on pharmacokinetic and pharmacodynamic results using MIC values obtained in CAMHB. Pharmacokinetic parameters, like AUC (area under the concentration-time curve)/MIC and C_max (maximum concentration of the drug in urine)/MIC ratios will often be much less favorable under physiologically relevant conditions. Therefore, not only pharmacokinetic but also pharmacodynamic studies in urine are as important as in serum or plasma.

Combination screening.

Using checkerboard assays, meropenem was tested in combination with the eight other noncarbapenem antibiotics in CAMHB and AU against all used strains. For the carbapenem-susceptible strains, most of the tested combinations had no effect in CAMHB and only minor effects in AU on the meropenem MIC values—except for fosfomycin, which resulted in a median 8-fold reduction of the meropenem MIC in both media with an additive median fractional inhibitory concentration index (ΣFICI) (Tables 2 and 3). In the case of the carbapenem-resistant strains, combinations with levofloxacin, colistin, fosfomycin, rifampin, and tigecycline showed an average reduction of meropenem MIC values of at least 4-fold compared to meropenem alone in CAMBH. Here, ΣFICI values showed synergism for the combination with levofloxacin, colistin, fosfomycin, and rifampin (Table 2). Similar results were obtained testing combinations in AU, except for the combination with levofloxacin where the median meropenem MIC fold change reduction is clearly reduced from 14 in CAMHB to 2.5 in AU (Table 3). Antibiotics showing synergistic effects in combination with meropenem (namely, levofloxacin, colistin, fosfomycin, rifampin, and tigecycline) were also tested with the carbapenem ertapenem. Here, similar, partly better, results were achieved as with meropenem (Table 4). Surprisingly, combination of meropenem/ ertapenem and levofloxacin showed beneficial combination effects, although all CRE strains showed an additional increased fluoroquinolone resistance. Since the exact mechanism of the fluoroquinolone resistance of the CRE strains used is not known, it is only possible to speculate about the underlying mechanism of this combination.

TABLE 2.

Fold change of reduction of meropenem MIC in combination over meropenem alone and ΣFICI values in CAMHB

Strain	Meropenem in combination witha :
	CIP		LVX		CS		FOS		NFT		RIF		TGC		TOB
	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI
Carbapenem susceptible
E. coli ATCC 25922	1	2	1	2	1	2	8	0.188	2	0.875	1	2	1	2	1.5	2
K. pneumoniae 595	1	2	1	2	1	2	16	0.563	1	2	1	2	1	2	1	2
E. cloacae CHD57	1	2	1	2	1	2	3	0.532	1	2	1	2	8	0.375	16	0.313
Median	1	2	1	2	1	2	8	0.531	1	2	1	2	1	2	1.5	2
Carbapenem resistant
E. coli
IR3	1	2	8	0.5	8.5	0.313	10	0.407	1	2	12	0.375	2	1	2	1
NRZ14408	8	0.625	8	0.625	8	0.625	6	0.328	4	0.5	5	0.813	16	0.563	4	0.375
BAA2469	2	1	6	0.563	8	0.438	18	0.180	16	0.563	12	0.376	5	0.688	1	2
K. pneumoniae
BAA1705	48	0.523	20	0.0578	24	0.422	24	0.251	2	0.563	40	0.525	1	2	32	0.53
BAA2146	1.5	1	20	0.453	8	0.375	8	0.375	2	1	4	0.407	8	0.625	1	2
E. cloacae BAA2468	3	0.875	20	0.328	48	0.289	4	0.625	8	0.625	18	0.391	4	0.375	2	0.75
Median	2.5	0.938	14	0.477	8.25	0.398	9	0.352	3	0.594	12	0.396	4.5	0.625	2	0.875

^aFC, fold change; CIP, ciprofloxacin; LVX, levofloxacin; CS, colistin sulfate; FOS, fosfomycin; NFT, nitrofurantoin; RIF, rifampin; TGC, tigecycline; TOB, tobramycin.

TABLE 3.

Fold change of reduction of meropenem MIC in combination over meropenem alone and ΣFICI values in AU

Strain	Meropenem in combination witha :
	CIP		LVX		CS		FOS		NFT		RIF		TGC		TOB
	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI
Carbapenem susceptible
E. coli ATCC 25922	1	2	1.5	0.565	1	2	8	0.25	4	0.75	2	0.75	5	0.813	2	0.625
K. pneumoniae 595	2	0.563	2	0.625	1	2	2	0.75	2	1	32	0.516	1	2	2	0.625
E. cloacae CHD57	1	2	1	2	2	1	64	0.516	2	1	1	2	2	0.531	32	0.531
Median	1	2	1.5	0.625	1	2	8	0.516	2	1	2	0.75	2	0.813	2	0.625
Carbapenem resistant
E. coli
IR3	8	0.375	8	0.5	12	0.141	4	0.5	4	0.313	12	0.375	6	0.438	1	2
NRZ14408	4	0.75	3	0.625	8	0.625	10	0.407	8	0.625	3	0.875	8	0.625	1	2
BAA2469	2	1	12	0.407	12	0.157	24	0.297	4	0.75	8	0.265	8	0.625	1	2
K. pneumoniae
BAA1705	1	2	1	2	17	0.766	20	0.328	1	2	20	0.578	2	1	16	0.563
BAA2146	1.5	0.75	1	2	5	0.813	4	0.5	1	2	6	0.438	16	0.188	1	2
E. cloacae BAA2468	2.5	0.75	2	0.438	4	0.344	2.5	0.75	1	2	3	0.563	4	0.75	1	2
Median	2.25	0.75	2.5	0.563	10	0.485	7	0.453	2.5	1.375	7	0.5	7	0.625	1	2

^aFC, fold change; CIP, ciprofloxacin; LVX, levofloxacin; CS, colistin sulfate; FOS, fosfomycin; NFT, nitrofurantoin; RIF, rifampin; TGC, tigecycline; TOB, tobramycin.

TABLE 4.

Fold change of reduction of ertapenem MIC in combination over ertapenem alone and ΣFICI values in CAMHB and AU

Strain	Ertapenem in combination with antibiotic showna
	In CAMBH										In AU
	LVX		CS		FOS		RIF		TGC		LVX		CS		FOS		RIF		TGC
	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI	FC	ΣFICI
Carbapenem susceptible
E. coli ATCC 25922	1	2	1	2	4	0.281	1	2	1	2	1	2	1	2	8	0.25	2	1	2	0.625
K. pneumoniae 595	1	2	1	2	4	0.375	1	2	2	0.531	1	2	2	1	1	2	1	2	1	2
E. cloacae CHD57	1	2	1	2	2	0.563	2	1	16	0.078	1	2	4	0.75	8	0.25	4	0.75	16	0.188
Median	1	2	1	2	4	0.375	1	2	2	0.531	1	2	2	1	2	0.625	2	1	2	0.625
Carbapenem resistant
E. coli
IR3	128	0.508	32	0.531	4	0.5	32	0.281	4	0.75	4	0.5	32	0.156	8	0.5	32	0.281	4	0.75
NRZ14408	12	0.313	32	0.75	32	0.281	6	0.438	16	0.563	4	0.75	8	0.625	16	0.5	4	0.75	32	0.313
BAA2469	8	0.375	1	2	16	0.125	16	0.188	1	2	1	2	4	0.5	32	0.188	16	0.563	8	0.375
K. pneumoniae
BAA1705	16	0.313	1	2	8	0.375	1	2	32	0.531	16	0.563	8	0.25	8	0.25	16	0.313	16	0.563
BAA2146	16	0.563	16	0.25	4	0.375	16	0.313	4	0.5	4	0.5	8	0.375	32	0.531	32	0.156	2	1
E. cloacae BAA2468	2.5	0.75	8	0.625	32	0.531	64	0.141	64	0.53	2.5	0.75	8	0.25	4	0.313	4	0.75	1	2
Median	14	0.442	12	0.688	12	0.375	16	0.297	10	0.547	4	0.657	8	0.313	12	0.407	16	0.438	6	0.657

^aFC, fold change; LVX, levofloxacin; CS, colistin sulfate; FOS, fosfomycin; RIF, rifampin; TGC, tigecycline.

Although different growth media could affect the MIC of various antibiotics, their impact on enhancing effects of combinations of two antibiotics was minimal for most of the antibiotics tested. One exception is levofloxacin, which still shows positive effects in combination with carbapenems in AU, but these are much weaker than in CAMHB.

Dynamic bladder model.

To further test the effectivity of combination antibiotic treatment after a single dose according to their urinary concentrations, we used a dynamic bladder model to simulate the pharmacokinetics of ertapenem, levofloxacin, rifampin, and fosfomycin. Here, the growth of three carbapenem-resistant strains (one strain each of E. coli, K. pneumoniae, and E. cloacae) was observed. Measured antibiotic concentrations in the bladder compartment of the dynamic model over 48 h were similar to the concentration curve predicted by the mathematical simulations (Fig. 1).

FIG 1.

Antibiotic concentrations measured in the bladder compartment of the dynamic model over time following a single dose. Single-dose application of ertapenem (1,000 mg), levofloxacin (500 mg), rifampin (600 mg), and fosfomycin (3,000 mg) were simulated mathematically and were measured at different time points in the bladder compartment of the dynamic model.

Adding ertapenem alone to the model initially decreased bacterial numbers below the detection level in the bladder compartment, but all three carbapenem-resistant strains showed regrowth after 24 h at the latest (Fig. 2). Levofloxacin and fosfomycin alone showed similar results for the E. coli and K. pneumoniae strains used. In contrast, for E. cloacae BAA2468, levofloxacin and fosfomycin alone were able to keep the bacterial numbers under the detection limit over 48 h. Initially, rifampin alone reduced the bacterial numbers for all three strains, but the detection limit was never underestimated (Fig. 2). Combination of ertapenem with levofloxacin and fosfomycin, as well as rifampin, decreased the bacterial numbers of the E. coli and K. pneumoniae strains at 24 and 48 h more than 100-fold compared to the antibiotics alone. Thus, the combinations synergistically improved the bactericidal effects of the single antibiotics by reducing the bacterial numbers below the detection limit after 8 h at the latest, up to the full 48 h. Compared to the checkerboard assays (Table 2), there is a partial improvement in the combination efficiency of additive to synergistic effects under dynamic conditions compared to the static ones for the tested strains E. coli NRZ14408 and K. pneumoniae BAA1705. In contrast, the PK simulation revealed synergistic activity only for the combination with rifampin after 24 h for E. cloacae, but regrowth to the level of bacterial numbers after rifampin alone treatment was detected thereafter (Fig. 2). In addition, unlike the checkerboard assay determinations, no further beneficial effect was detected for the combination with levofloxacin or fosfomycin for E. cloacae BAA2468.

FIG 2.

Bacterial growth in the dynamic model following single dose of antibiotics alone or in combination. Urine pharmacokinetic of ertapenem (1,000 mg), levofloxacin (500 mg), fosfomycin (3,000 mg), and rifampin (600 mg) alone or in combination were simulated according to the respective plasma elimination constant and urinary recovery rate at an assumed average urine production rate of 1.5 liters/24 h with a simulated bladder voiding every 3 h. Bacteria were inoculated with 1 × 10⁶ CFU/ml in the bladder compartment, and bacterial growth was measured over 48 h by sample dilution and plating.

Recently Wijma and colleagues described a discrepancy between clinically observed bactericidal effects compared to statically determined in vitro efficacy of antibiotics. Indicating a major role of dynamic models assessing PK/PD parameters to predict clinical/ in vivo microbiological success (20). Thus, our data revealed regrowth for the tested CRE strains already after 24 h if treated with ertapenem, although the ertapenem concentration in the bladder was above the determined MIC values for approximately 20 h. This is in contrast to the description that carbapenem regimens which provide a time above the MIC (T_MIC) target of 40% of the dosing interval are sufficient to cause maximal microbial killing (21). Bacterial regrowth may be an indication of the selection of higher resistant bacteria. Thus, emergence of resistance could occur even with longer T_MIC targets. Nevertheless, in case of the tested E. coli and K. pneumoniae strains, this could be prevented by combining the single dosage of carbapenem with a single dosage of rifampin, levofloxacin, or fosfomycin. Due to the increased fluoroquinolone resistance in the CRE strains and the decreased combination activity of levofloxacin in AU over CAMHB, the combination with rifampin and fosfomycin would be preferable to levofloxacin, despite similar activities in the dynamic model. Although the tested E. cloacae and E. coli strains showed similar results in the checkerboard assay for the combination with levofloxacin or rifampin, these combinations did not show the same effect against these strains in the dynamic bladder model. These differences between static and dynamic models underline the statement by Wijma et al. (20) that in addition to static models, dynamic models should be included to improve the predictability of clinical success using in vitro data. Therapies containing colistin in combination with a second antibiotic (tigecycline or aminoglycoside) are preferred used for the treatment of carbapenem-resistant Gram-negative bacteria. However, the diversity of available clinical data prevents the possibility to make a solid recommendation regarding their use. In addition, nephrotoxicity is a major dose-limiting adverse effect of polymyxins. Inadequate treatment with colistin also poses the risk of selecting additional resistance to colistin, thereby rendering two of the last-resort antibiotics ineffective. Recent data showed promising results of carbapenem-containing combinations especially with high-dose carbapenems. For a better understanding and for a meaningful use of such therapies, more in vitro studies should necessarily be carried out. In order to better predict the clinical efficacy of a combination therapy according to the PK/PD parameters and thus to increase the clinical success. It is important to consider the focus of action of the therapy and to integrate it into the in vitro tests, e.g., using physiologically like media, to further increase predictability.

Further studies should address the underlying molecular mechanism of the synergistic activity of the combination.

Conclusion.

Our in vitro studies using checkerboard and a dynamic bladder model simulating PK in urine showed that treatment of CRE strains could still be successful using carbapenem-containing combination therapy with fosfomycin or rifampin. Combination therapy may also be beneficial for the treatment of UTIs caused by CRE strains, and therapy may also be a better option for treatment of UTIs caused by susceptible strains to prevent regrowth and thus possible emergence of resistance. For further investigations, studies using preclinical infection models as well as well as designed and controlled clinical studies should be performed.

MATERIALS AND METHODS

Bacterial strains and antibiotics.

Three carbapenem-susceptible strains (Escherichia coli ATCC 25922, Klebsiella pneumoniae 595, and Enterobacter cloacae CHD57) and six carbapenem-resistant strains [E. coli strains IR3 (bla_NDM-1), ATCC BAA2469 (bla_NDM-1), and NRZ14408 (bla_KPC-2 mcr-1), K. pneumoniae ATCC BAA2146 (bla_NDM-1), K. pneumoniae ATCC BAA1705 (bla_KPC), E. cloacae ATCC BAA2468 (bla_NDM-1)] were included in this study. E. coli IR3 and NRZ14408 and K. pneumoniae 595, as well as E. cloacae CHD57, are clinical isolates from UTI patients.

Determination of MICs.

MIC determinations were performed according to the CLSI and EUCAST standards (22, 23). A broth microdilution assay was used for determination of the MICs of fosfomycin, levofloxacin, meropenem, rifampin (TCI Deutschland GmbH, Eschborn, Germany), ciprofloxacin, colistin sulfate (Merck, Darmstadt, Germany), ertapenem (Melford Laboratories, Ltd., Ipswich, United Kingdom), nitrofurantoin (Cayman Chemical, Ann Arbor, MI, USA), and tigecycline (LKT Laboratories, St. Paul, MN, USA), as well as tobramycin (Tocris, Wiesbaden, Germany) in CAMHB (Merck) and artificial urine (AU) (24). The inoculums, confirmed by plating, ranged from 3.4 to 8.5 × 10⁵ CFU/ml, and the MIC was defined as the lowest concentration inhibiting visible growth (optical density at 600 nm [OD₆₀₀] of <0.1) after incubation at 37°C for 20 ± 2 h. MIC determinations were repeated at least three times.

Checkerboard assays.

A two-dimensional, two-agent broth microdilution checkerboard titration method was used to study the interaction between the antibiotic combinations in CAMHB and AU (25). The final inoculum, confirmed by plating, ranged from 3.1 to 9.5 × 10⁵ CFU/ml. After 20 ± 2 h of incubation at 37°C, the MIC was determined. Checkerboard determinations were performed one to two times. Interactions between two antibiotics were then evaluated using the fractional inhibitory concentration indices (ΣFICs), calculated as the sum of the FICs as ΣFIC = FIC_{antibiotic A} + FIC_{antibiotic B}, where FIC is the MIC of the substance in combination/MIC of the substance alone. The correlation between ΣFIC and the effect of the combination was calculated as follows: synergy of ≤0.5, additive effect of >0.5 to 1, indifference of >1 to <4, and antagonism of ≥4.

Pharmacokinetic simulation/dynamic bladder infection in vitro model.

The dynamic bladder model used in this study was adapted to previous described models (26, 27). Autoclavable PVC tubing (Ismatec, Wertheim, Germany) and glassware (VWR, Darmstadt, Germany) were connected to peristaltic pumps (Ismatec), which enabled simulation of three individual bladder compartments to be run in parallel that were set within water baths maintained at 37°C. Urine concentrations following administration of 1,000 mg ertapenem (intravenous [i.v.]), 500 mg levofloxacin (i.v.), 600 mg rifampin (oral), and 3000 mg fosfomycin (oral) alone or in combination were simulated. Antibiotics were added to an AU-containing vessel upstream of the bladder compartments at time point 0. From this, the antibiotics were pumped into the bladder compartments according to their plasma elimination rate constant (K_ε). A constant influx of antibiotic-free AU into the upstream vessel simulates antibiotic washout of the plasma into the urine.

Urine pharmacokinetics were fitted from normal human PK parameters following administration of a single dose (Table 5) at an assumed average urine production rate of 1.5 liters/24 h. In combination with a simulated bladder voiding pattern every 3 h, this procedure allows the simulation of continuously fluctuating urine concentrations. Individual test pathogens were added to each bladder compartment at an inoculum of ∼1 × 10⁶ CFU/ml. Simulations were performed over 48 h. Samples for PD assessment were taken directly from each bladder compartment. Samples were diluted and plated on CAMHB agar plates for CFU/ml determination. PK simulations were performed in AU. The effects of the combination were defined according to White et al. (14): (i) synergy as ≥100-fold (2-log₁₀) decrease of CFU/ml at 24/48 h compared to single agent, (ii) indifference as ≤10-fold change of CFU/ml at 24/48 h compared to single agent, and (iii) antagonism as ≥100-fold increase of CFU/ml at 24/48 h compared to single agent.

TABLE 5.

PK parameters of antibiotics used in this study as simulated in the dynamic bladder model

Antibiotic	Simulated dose	Plasma t1/2 (h)a	Kεb	Urine recovery (%)a	Reference(s)
Ertapenem	1,000 mg i.v.	4.8	0.144	49.7	28–30
Levofloxacin	500 mg i.v.	6.7	0.103	61.5	31–34
Rifampin	600 mg oral	3.1	0.224	15	11
Fosfomycin	3,000 mg oral	4.5	0.154	39	12, 13

^aShown are the mean values of the specified reference.

^bPlasma elimination rate constant.

Measurement of antibiotic concentrations.

All antibiotic concentrations were quantified by spectrophotometric methods. Urine samples were diluted in water and transferred to UVettes (Eppendorf, Hamburg, Germany), and absorbance was measured at 296 nm (ertapenem), 298 nm (levofloxacin), and 472 nm (rifampin) using the MultiskanGO spectrophotometer (Thermo Scientific, MA, USA). For fosfomycin determination, a method based on formation of a 1:1 charge transfer complex with alizarin in ethanol solution was used (15). Urine samples were diluted in absolute ethanol and mixed with 5 mM alizarin. Afterwards absorbance at 545 nm was measured. Concentration calculations were based on standard curves.

Statistical analysis.

Differences between two groups were assessed using the Student's t test. Statistical calculations were performed by using Microsoft Excel 2016.