Modeling the antimicrobial pharmacodynamics for bacterial strains with vs. without acquired resistance to fluoroquinolones or cephalosporins

Abstract

Objectives:

Antimicrobial resistance threatens therapeutic options for human and animal bacterial diseases worldwide. The current antimicrobial treatment regimens were designed against bacterial pathogen strains that were fully susceptible to them. To expand the useable lifetime of existing antimicrobial drug classes by modifying the treatment regimens, data are needed on the antimicrobial pharmacodynamics (PD) against strains with the reduced susceptibility. In this study, we generated and mathematically modelled PD of the fluoroquinolone ciprofloxacin and cephalosporin ceftriaxone against nontyphoidal Salmonella enterica subsp. enterica strains with varying levels of the acquired resistance.

Methods:

We included Salmonella strains across categories of reduced susceptibility to fluoroquinolones or cephalosporins reported to date, including isolates from human infections, food-animal products sold in retail, and food-animal production. We generated the PD data for each drug and strain via the time-kill assay. Mathematical models were compared in their fit to represent the PD. The best-fit model’s parameter values across the strain susceptibility categories were compared.

Results:

The inhibitory baseline sigmoid I_max (or E_max) model was best-fit for the PD of each antimicrobial against a majority of the strains. There were statistically significant differences in the PD parameter values across the strain susceptibility categories for each antimicrobial.

Conclusions:

The results demonstrate predictable multi-parameter changes in the PD of these first-line antimicrobials depending on the Salmonella strain’s susceptibility phenotype and specific genes conferring the reduced susceptibility. The generated PD parameter estimates could be used to optimize the treatment regimens against infections by the reduced susceptibility strains.

1. Introduction

Nontyphoidal Salmonella enterica subsp. enterica are the number one bacterial foodborne pathogen linked to hospitalizations and death in cases of foodborne illness in the U.S. [1, 2]. Antimicrobial drug resistant (AMR) nontyphoidal Salmonellae, specifically, cause ~212,500 infections in the U.S. each year [2]. The challenge of treating infections by AMR strains is not easily remedied, because the time-line for development of new effective antimicrobials is uncertain [3–5]. Devising approaches to optimize the treatment regimens by existing antimicrobials is of paramount importance, to expand the useable lifetime of these effective antimicrobial drug classes. Developing these approaches requires understanding the antimicrobial pharmacokinetics in the target hosts, and the antimicrobial pharmacodynamics (PD) against the bacterial populations at the sites of infection in the hosts; the latter is approximated using in vitro PD assays and mathematical modeling [6–10].

The minimum inhibitory concentration (MIC) of an antimicrobial for a bacterial strain is a measurement of the strain’s phenotypic susceptibility. Currently, it is used as a single parameter of the PD, along with the projected drug pharmacokinetics (PK) in the target host, to design treatment regimens. In the integrated PK-PD models, the duration of time the drug concentration at the site of infection remains above the MIC, or the area under the concentration-time curve (AUC) above the MIC, is projected [7, 10, 11]. In this framework, the regimen of the antimicrobial administration is designed to achieve the desired duration or AUC above MIC [7, 10]. Multi-parameter PD models, however, further capture the antimicrobial PD against the bacterial population depending on the drug concentration that inevitably varies at the sites of infection during the treatment, and thus are superior tools for the regimen design compared to the MIC alone [8, 11–14]. Another aspect is that the current antimicrobial treatment regimens have been designed using the PD parameter values (whether the MIC or those from multi-parameter PD models) for pathogen strains that were fully susceptible to the drugs. Less susceptible strains are considered to differ only by the MIC of the antimicrobial [7, 12]. However, evidence is emerging that all the antimicrobial’s PD parameter values, not only the MIC, are changing between the fully susceptible strains and strains with acquired resistance to the drug [11, 14, 15]. It is unknown whether the changes are predictable and how they relate to the strain’s susceptibility phenotype or genotype. It is also unknown if for a given antimicrobial the same mathematical model captures the PD across the strain susceptibility categories. Several multi-parameter models have been shown to capture various forms of the PD relationships between an antimicrobial drug concentration and its effects on the bacterial population [12, 13, 16, 17].

In this study, we investigated the PD against nontyphoidal Salmonella strains that are susceptible or have reduced susceptibility to the first-line treatment choices for serious salmonellosis in adults, the fluoroquinolone ciprofloxacin and cephalosporin ceftriaxone. The study included the strains carrying majority of resistance reported in the U.S. gene families of acquired resistance to these drug classes (Supplementary Tables 1–2). We investigated whether and how the PD parameter values change depending on the strain phenotypic susceptibility to the antimicrobial or specific resistance gene type (chromosomal vs. plasmidic) or family carried.

2. Materials and methods

2.1. Bacterial strains

The nontyphoidal Salmonella enterica subsp. enterica strains (n=53) were isolated and shared by the U.S. National Antimicrobial Resistance Monitoring System (NARMS), Centers for Disease Control and Prevention (CDC), and colleagues at the University of Nebraska Medical Center (strains acquired within the NARMS activities in Nebraska). Those organizations also shared data on the strain serotype and AMR gene content determined via the whole genome sequencing and annotation. Based on the data, strains with majority of reported in the U.S. gene families conferring reduced susceptibility to fluoroquinolones or extended spectrum cephalosporins were selected for the study (Supplementary Tables 1–2). In the ciprofloxacin PD study n=24 strains and in the ceftriaxone PD study n=29 strains were used.

2.2. MIC estimation

We measured the phenotypic susceptibility of each strain to ciprofloxacin or ceftriaxone. The strains were stored in Brucella broth with 15% glycerol (Remel™, Lenexa, KS, USA) at −80 °C. Tryptic soy agar supplemented with 5% sheep blood (BAP, Remel™) was used for the plating. High purity (analytical standards) forms of ceftriaxone and ciprofloxacin were purchased from Sigma-Aldrich, Inc. (St. Louis, MO, USA). The antimicrobial’s MIC for a bacterial strain was determined by the broth microdilution assay aerobically in cation-adjusted Mueller-Hinton broth (CA-MHB, Remel™), following the procedures recommended by the Clinical and Laboratory Standards Institute (CLSI) [18]) with the following modification. To get a more precise MIC estimate, for each drug and strain, multiple broth microdilution assays were performed with different starting drug concentrations. The midpoint method was used to estimate the MIC; it was assumed that in each assay the strain’s MIC was in the middle of the drug double-dilution interval identified and the average of the midpoints was accepted as the MIC estimate. The categorization of a strain as susceptible, intermediately susceptible, or resistant was based on the CLSI recommended clinical interpretive breakpoints for ceftriaxone or ciprofloxacin MIC for nontyphoidal Salmonellae treatments in humans.

2.3. Time-kill experiments

Time-kill experiments were conducted for n=24 strains with ciprofloxacin and n=29 with ceftriaxone, following previously described procedures [11, 19]. In brief, an overnight culture of the strain was inoculated in Mueller Hinton 2 broth (MHB, Remel™) and incubated aerobically at 37 °C with shaking (200 rpm). In 1 hour (once in the exponential growth phase), the culture was diluted 1:40 into flasks containing MHB with the drug concentration of 0.5, 0.75, 1, 2, 3, 5, 8, or 10 multiples of the MIC for the strain. Each culture was incubated as specified above and sampled at 0, 1, 2, 3, 4, 5, 6, 7, 8, 12, and 24 hours of incubation. The colony forming units (CFU) in the culture were estimated as previously described [11, 19]. The mean of the log₁₀(CFU/mL) from all the dilutions plated from a culture at a time-point provided the bacterial population density estimate. The time-kill experiment for the drug and strain was replicated on different days.

2.4. Pharmacodynamic modeling and statistical analysis

The bacterial population density estimates for the strain, drug, drug concentration (a MIC multiple or no-drug control), and time-point were averaged between the time-kill experiment replicates and plotted against the experiment time. The plotted curve reflected the strain’s population density under that drug concentration or in the control throughout the 24-hour experiment. The curve’s hourly slope was estimated for each time-window between 0 and 1 to 23 hours, and for each time-window between 1 (beginning of the exponential growth phase) and 2 to 22 hours, and the maximal slope identified. The curve’s maximal hourly slope provided an estimate of the maximal hourly bacterial population growth or decline rate when exposed to that drug concentration or in the absence of antimicrobial exposure. These estimates collated for the strain yielded the antimicrobial concentration-bacterial population growth/decline rate PD data-set. Six candidate models were compared in capturing the antimicrobial’s PD for the strain. Each model was fitted as a non-linear regression model using the least-squares method, by regressing the maximal hourly bacterial population growth/decline rate on the drug concentration for the strain. The models were:

1. Inhibitory baseline sigmoid I_max model – form 1:

   $$E(C)=E_o-\frac{I_{\max }\times C^H}{I{C}_{50}^H+C^H}$$
2. Inhibitory baseline sigmoid I_max model – form 2:

   $$E=E_0-\frac{I_{\max }}{I{C^H}_{50}+C^H}$$
3. Inhibitory fractional model:

   $$E=E_0\times \frac{1-C}{I{C}_{50}+C}$$
4. Inhibitory fractional sigmoid model:

   $$E(C)=E_0\times \frac{1-C^H}{I{C^H}_{50}+C^H}$$
5. Inhibitory fractional I_max model:

   $$E=E_0\times \frac{1-I_{\max }\times C}{I{C}_{50}+C}$$
6. Inhibitory fractional sigmoid I_max model:

   $$E=E_0\times \frac{1-I_{\text{max}}\times C^H}{I{C^H}_{50}+C^H}$$

where:

• E(C) – maximal bacterial population growth rate (log₁₀(CFU/mL)/hour) when exposed to the antimicrobial concentration C

• C – antimicrobial concentration (μg/mL)

• E₀ – bacterial population growth rate in the absence of antimicrobial exposure (log₁₀(CFU/mL)/hour)

• H – Hill-coefficient reflecting steepness of the relationship between an increase in the antimicrobial concentration and an increase in the inhibition of the bacterial population growth (dimensionless)

• IC₅₀ – Drug concentration at which 50% of the maximal inhibition in the bacterial population growth occurred (μg/mL)

• I_max – Maximal inhibition of the bacterial population growth at high antimicrobial concentrations (dimensionless)

To identify a best-fit PD model for each antimicrobial and strain, the six models were ranked on their relative fit to the PD curve based on the Akaike information criterion (AIC) values (with a lower AIC indicating a better fit model), and on the adjusted coefficient of determination, R², values (with a higher R² indicating a better fit model). For each ciprofloxacin and ceftriaxone, the PD model ranked “best” fit most often across for the strains was identified. Statistical analysis was performed to test for trends in values of the best-fit PD model’s parameters depending on the strain’s phenotypic susceptibility or the resistance gene family carried for the antimicrobial. The PD modeling and statistical analysis were performed in the SAS® University version 9.4 Edition software for Windows (SAS Institute Inc., Cary, NC).

3. Results and Discussion

3.1. Relative fit of pharmacodynamic models

Relative fit of the six candidate models to the antimicrobial concentration-maximal bacterial population hourly growth/decline rate curve for individual strains of nontyphoidal Salmonellae is summarized for ciprofloxacin in Supplementary Table 5 and for ceftriaxone in Supplementary Table 6. For ciprofloxacin, the inhibitory baseline sigmoid I_max model – form 1 (equation (1)) was best-fit to the PD curve for 96% of the strains (i.e., for 23 out of 24 strains). The inhibitory fractional I_max model (equation (5)) was best-fit for the other strain’s PD curve in the ciprofloxacin study. For ceftriaxone, the inhibitory baseline sigmoid I_max model – form 1 (equation (1)) was best-fit to the PD curve for 86% of the strains (i.e., for 25 out of 29 strains). The inhibitory fractional I_max model (equation (5)) was best-fit for the other 4 strain PD curves in the ceftriaxone study. To enable statistical analysis of trends in the PD parameter values, the inhibitory baseline sigmoid I_max model – form 1 (equation (1)) was accepted as the best-fit model for all the strains in both the ciprofloxacin and ceftriaxone studies.

3.2. Trends in the ciprofloxacin pharmacodynamic parameter values

The time-kill experiment results with ciprofloxacin are included in Supplementary Figures 1–2. These demonstrated the ciprofloxacin PD varies depending on the strain’s phenotypic susceptibility (measured by the drug MIC) and genotype –carrying chromosomal or plasmidic gene of fluoroquinolone resistance. For all the strains, the PD parameter value estimates of the inhibitory baseline sigmoid I_max model – form 1 and the model fit statistics are shown in Supplementary Table 3. Examples of the model projections are plotted against the observed PD curves for the individual strains in Figure 3.

Figure 3.

Projections of the bacterial population growth/decline rate depending on the ciprofloxacin concentration. The projections are of the inhibitory baseline sigmoid I_max model, which demonstrated best-fit out of six candidate modes to the time-kill data for nontyphoidal Salmonella enterica subsp. enterica strains investigated. The projections for the strains with the ciprofloxacin MIC ranging from 0.012 μg/mL (A) to 10.1 μg/mL (F) are plotted in the increasing order of the MIC (left to right, and top to bottom). These strains are categorized as intermediatelysusceptible to ciprofloxacin and resistant, based on the ciprofloxacin MIC clinical interpretative breakpoints set by the Clinical and Laboratory Standards Institute.

The strain’s maximal population growth rate in the absence of antimicrobial exposure, E₀, represents the strain fitness or intrinsic growth potential. This is also a parameter that impacts the PD of antimicrobials acting on the replicating bacterial cells, as has been shown by us for fluoroquinolones [20] and by others for β-lactams (e.g., [21]). We categorized the Salmonella strains in their phenotypic ciprofloxacin susceptibility based on the ciprofloxacin MIC and the corresponding CLSI-set clinical interpretative breakpoints for salmonellosis treatment in humans. The categories were the susceptible, intermediately susceptible, and resistant to ciprofloxacin strains. The mean intrinsic growth rate of the ciprofloxacin resistant strains was statistically significantly lower, and less variable compared to that of the susceptible or intermediately susceptible strains (Kruskal-Wallis one-way analysis of variance p-value=0.060, n = 24 strains; Figure 1A). Further analysis showed that the mean E₀ tended to be lowest for the strains with chromosomal fluoroquinolone resistance genes, increasing for the strains with plasmidic fluoroquinolone resistance genes, and further increasing for the susceptible strains (Kruskal-Wallis one-way analysis of variance p-value=0.346, n=24; Figure 1D). The mean maximal inhibition of the bacterial population growth, I_max, was the highest for the ciprofloxacin susceptible strains, decreasing for the intermediately susceptible, and further decreasing for the resistant strains (Kruskal-Wallis one way analysis of variance p-value=0.077, n =22; Figure 1B). However, the I_max was highly variable across the susceptible strains, while it was highly predictable across the resistant strains (Figure 1B). The mean Hill-coefficient estimate was statistically significantly lower for the ciprofloxacin susceptible strains, increasing for the intermediately susceptible, and further increasing for the resistant strains (the Kruskal-Wallis one-way analysis of variance p-value=0.039, n =22; Figure 1C). As well, the Hill-coefficient was highly predictable for the susceptible strains, while it was highly variable across the intermediately susceptible and resistant strains (Figure 1C). Further analysis showed a stepwise and statistically significantly increase in the Hill-coefficient from the susceptible strains to the strains with chromosomal to the strains with plasmidic genes conferring fluoroquinolone resistance (Kruskal-Wallis one-way analysis of variance p-value=0.010, n=20).

Figure 1.

Distributions of the parameter estimates of the inhibitory baseline sigmoid I_max model fitted to the time-kill data for nontyphoidal Salmonella enterica subsp. enterica strains (n=24) in the study of ciprofloxacin pharmacodynamics. Of six compared pharmacodynamic models, this model fitted the data best across the strains. (A-C): The strains are categorized by ciprofloxacin susceptibility based on the ciprofloxacin MIC and the clinical interpretative breakpoints set by the Clinical and Laboratory Standards Institute. (D-F): Strains are categorized by the content of genes encoding reduced susceptibility to quinolones, based on the whole genome sequencing and annotation.

The ciprofloxacin study results confirmed our earlier hypothesis that fluoroquinolone PD against a bacterial strain is significantly associated with its intrinsic growth potential [20]. As E₀ significantly declined from the fluoroquinolone susceptible to intermittently resistant to resistant Salmonellae, I_max also declined, while the Hill-coefficient value increased (the Kruskal-Wallis one-way analysis of variance p-value=0.060, 0.077, and 0.039, respectively; Figure 1A–C). For the ciprofloxacin PD, there was a statistically significant correlation between a higher rank in E₀ and a lower rank in the Hill-coefficient value across the strains (Spearman’s nonparametric rank-order correlation coefficient ρ=−0.617, p-value=0.001, n=24; Table 1). The I_max and Hill-coefficient parameter value changes could be collectively described as a trend from the concentration-dependent to time-dependent pharmacodynamics with a decreasing intrinsic growth potential of the pathogen strain. This potential declines with the strain’s phenotypic susceptibility (Figure 1A); this implies that while the fluoroquinolone PD is concentration-dependent for susceptible strains, it is trending towards time-dependency through the intermediately susceptible to resistant strains (Supplementary Figures 1–2; Supplementary Table 1). However, the impacts of the intrinsic growth potential may be interacting with other influences of the susceptibility phenotype on the ciprofloxacin PD. Specifically, there were a statistically significant negative correlation between the I_max and MIC:IC₅₀ ratio ranks across the strains (Spearman’s nonparametric rank-order correlation ρ=−0.659, p-value ≤0.001, n=24), and a positive correlation between the ranks in the Hill-coefficient value and MIC:IC₅₀ ratio (ρ=0.538, p-value=0.010, n=24). Moreover, the intrinsic growth rate and PD parameter value changes further depended on whether the strains carried chromosomal or plasmidic genes of fluoroquinolone resistance (Figure 1D–F). That is, the strain’s intrinsic growth rate and PD parameter value ranges may be associated with specific mechanisms of fluroquinolone resistance.

Table 1:

Pair-wise correlations between the ciprofloxacin pharmacodynamics parameter values based on the inhibitory baseline sigmoid I_max model fitted to the data on ciprofloxacin pharmacodynamics for individual strains of nontyphoidal Salmonella enterica subsp. enterica (n=24).

Parameter	I max	E 0	Hill-coefficient
I max	-	0.258	−0.298
E 0	0.258	-	−0.617*
Hill-coefficient	−0.298	−0.617*	-
MIC: IC50 ratio	−0.659*	−0.249	0.538*
MIC	−0.318	−0.155	0.546

Spearman correlation coefficient values,

^*indicates a correlation coefficient for which p-value ≤0.05

3.3. Trends in the ceftriaxone pharmacodynamic parameter values

The time-kill experiment results with ceftriaxone are included in the Supplementary Figures 3–5. The results demonstrated the ceftriaxone PD varies depending on the strain’s phenotypic cephalosporin susceptibility and further on the bla-gene family that confers the reduced susceptibility. For all the strains, the PD parameter value estimates of the inhibitory baseline sigmoid I_max model – form 1 and the model fit statistics are shown in Supplementary Table 4 and the other five models in Supplementary Table 6. Examples of the model projections are plotted against the observed PD curves for individual strains in Figure 4.

Figure 4.

Projections of the bacterial population growth/decline rate depending on the ceftriaxone concentration. The projections are of the inhibitory baseline sigmoid I_max model, which demonstrated best-fit out of six candidate modes to the time-kill data for nontyphoidal Salmonella enterica subsp. enterica strains investigated. The projections for the strains with the ceftriaxone MIC ranging from 0.012 μg/mL (A) to 631 μg/mL (F) are plotted in the increasing order of the MIC (left to right, and top to bottom). These strains are categorized as susceptible, intermediately susceptible and resistant to ceftriaxone, based on the ceftriaxone MIC clinical interpretative breakpoints set by the Clinical and Laboratory Standards Institute.

The strains were categorized in their phenotypic cephalosporin susceptibility based on the ceftriaxone MIC and the corresponding CLSI-set clinical interpretative breakpoints for salmonellosis treatment in humans. The categories were the susceptible and resistant to ceftriaxone strains. The mean intrinsic growth rate of the resistant strains was statistically significantly lower compared to that of the susceptible strains (Student’s t-test p-value=0.005, n=27; Figure 2A). Interestingly, the intrinsic growth rate of the cephalosporin resistant strains was highly variable, in opposite to the highly predictable intrinsic growth rate of the fluoroquinolone resistant strains (Figure 2A vs. Figure 1A). A statistically significantly higher mean maximal inhibition of the bacterial population growth, I_max, was estimated for the ceftriaxone resistant strains compared to the susceptible strains (Welch’s t-test p-value≤0.001, n = 26; Figure 2B). The mean Hill-coefficient estimates were similar between the ceftriaxone susceptible and resistant strains (Student’s t-test p-value=0.156, n=26; Figure 2C).

Figure 2.

Distributions of the parameter estimates of the inhibitory baseline sigmoid I_max model fitted to the time-kill data for nontyphoidal Salmonella enterica subsp. enterica strains (n=29) in the study of ceftriaxone pharmacodynamics. Of six compared pharmacodynamic models, this model fitted the data best across the strains. (A-C): The strains are categorized by ceftriaxone susceptibility based on the ceftriaxone MIC and the clinical interpretative breakpoints set by the Clinical and Laboratory Standards Institute. (D-F): The strains are categorized by the content of genes encoding reduced susceptibility to cephalosporins, based on the whole genome sequencing and annotation.

The ceftriaxone study results confirmed the cephalosporin PD against a bacterial strain is significantly associated with the strain’s intrinsic growth potential [20]. For the ceftriaxone PD, there was a statistically significant correlation between a lower rank in the intrinsic growth rate, E₀, and a higher rank in I_max across the strains (Spearman’s nonparametric rank-order correlation coefficient ρ=−0.448, p-value=0.020, n=25; Table 2). Lower ranks in E₀ and MIC:IC₅₀ ratio were also significantly correlated (ρ=0.414, p-value=0.040, n=25; Table 2) across the strains. Correspondingly, there was a negative correlation between I_max and MIC:IC₅₀ ratio ranks across the strains (ρ=−0.583, p-value=0.002, n=22; Table 2). The strain’s intrinsic growth rate significantly decreased from the cephalosporin susceptible to resistant strains (Student’s t-test p-value=0.005; Figure 2A). Taken together, the results imply that while the ceftriaxone PD is more often time-dependent for the susceptible strains, it is trending towards concentration-dependent dynamics with the decreasing phenotypic susceptibility (Supplementary Figures 3–5; Supplementary Table 2). Moreover, carriage of specific bla-gene families, which confer a reduced cephalosporin susceptibility by encoding production of different β-lactamases, significantly affected the ceftriaxone PD parameter value ranges, and, interestingly, was associated with the intrinsic growth rate of the strains. Specifically, there were statistically significant differences in the mean intrinsic growth rate, E₀, the maximal inhibition of the bacterial population growth by ceftriaxone, I_max, and the mean Hill-coefficient estimates among the strain groups carrying individual bla-gene families (Kruskal-Wallis one-way analysis of variance p-value=0.025, n=28; p-value=0.003, n=26; and p-value=0.028, n=27, respectively) (Figure 2D–F).

Table 2:

Pair-wise correlations between the ceftriaxone pharmacodynamics parameter values based on the inhibitory baseline sigmoid I_max model fitted to the data on ceftriaxone pharmacodynamics for individual strains of nontyphoidal Salmonella enterica subsp. enterica (n=29).

Parameter	I max	E 0	Hill-coefficient
I max	-	−0.448*	−0.354
E 0	−0.448*	-	−0.193
Hill-coefficient	−0.354	−0.193	-
MIC: IC50 ratio	−0.583*	0.414*	0.308
MIC	0.564*	−0.473*	−0.641*

Spearman correlation coefficient values,

^*indicates a correlation coefficient for which p-value ≤0.05.

4. Conclusion

The study results demonstrated statistically significant changes in the pharmacodynamics against nontyphoidal Salmonellae of each of the first-line antimicrobials for treating serios salmonellosis in adults, the fluoroquinolone ciprofloxacin and 3rd generation cephalosporin ceftriaxone. The changes are predictably linked to the strain’s phenotypic susceptibility to the antimicrobial and associated intrinsic growth potential. While the ciprofloxacin PD against nontyphoidal Salmonellae is understood to be time-dependent, the results demonstrate this only applies to fluroquinolone susceptible strains. The ciprofloxacin PD predictably changes to concentration-dependent dynamics as the strain phenotypic susceptibility and intrinsic growth rate decrease. Moreover, the PD changes further depend on whether the reduced susceptibility to fluoroquinolones is conferred by chromosomal or plasmidic genes or on the associated resistance mechanisms. While the ceftriaxone PD against nontyphoidal Salmonellae is understood to be time-dependent, the results demonstrate this only applies to cephalosporin susceptible strains. The ceftriaxone PD predictably changes to concentration-dependent dynamics as the strain phenotypic susceptibility and intrinsic growth rate decrease. Moreover, importantly, specific bla-gene families conferring the reduced cephalosporin susceptibility, or the β-lactamases produced by the strains, may be predictive of the ceftriaxone PD parameter value ranges. The predictability of the PD changes depending on the strain phenotype and genotype of the reduced antimicrobial susceptibility strongly support the premise of a future investigation of whether the treatment regimens by these antimicrobials for salmonellosis in adults could be safely modified to achieve efficacy to treat the infections by the reduced susceptibility strains. Such modified treatment regimens effective against bacterial strains with the acquired resistance to the antimicrobials could reduce recourse to the limited resource newer antimicrobial drug classes, and expand the useable lifetime of the critically important existing classes of fluoroquinolones and cephalosporins.

Highlights.

• Predictable changes in the pharmacodynamics against nontyphoidal Salmonellae

• PD changes are related to fitness cost for the strains associated with resistance

• Pharmacodynamic changes may depend on the gene families conferring the resistance.