Population Pharmacokinetic Modeling of Free Plasma and Free Brain Concentrations of Ceftaroline in Healthy and Methicillin-Resistant Staphylococcus aureus-Infected Wistar Rats

ABSTRACT

A population pharmacokinetic model was developed to describe alterations in ceftaroline brain disposition caused by meningitis in healthy and methicillin-resistant Staphylococcus aureus (MRSA)-infected rats. Blood and brain microdialysate samples were obtained after a single bolus dose of ceftaroline fosamil (20 mg/kg) administered intravenously. Plasma data were modeled as one compartment, and brain data were added to the model as a second compartment, with bidirectional drug transport between plasma and brain (Q_in and Q_out). The cardiac output (CO) of the animals showed a significant correlation with the relative recovery (RR) of plasma microdialysis probes, with animals with greater CO presenting lower RR values. The Q_in was approximately 60% higher in infected animals, leading to greater brain exposure to ceftaroline. Ceftaroline brain penetration was influenced by MRSA infection, increasing from 17% (Q_in/Q_out) in healthy animals to 27% in infected animals. Simulations of a 2-h intravenous infusion of 50 mg/kg every 8 h achieved >90% probability of target attainment (PTA) in plasma and brain for the modal MRSA MIC (0.25 mg/L), suggesting that the drug should be considered an option for treating central nervous system infections.

INTRODUCTION

Staphylococcus aureus is one of the most common pathogens causing health care-associated infections, which vary from skin and soft tissue infections to life-threatening infections such as bloodstream infections (1). Albeit uncommonly, it can also cause central nervous system (CNS) infections, which typically occur from complications of invasive neurosurgical procedures (2). A retrospective study by Aguilar et al. (3) at Henry Ford Hospital revealed an incidence of 4.9% for S. aureus meningitis, and 48% of those cases were associated with methicillin-resistant S. aureus (MRSA).

Vancomycin is recommended for the treatment of MRSA meningitis (4). However, studies have shown high interindividual variability (IIV) and poor penetration into the cerebrospinal fluid (CSF) of patients with suspected or proven CNS infections (5). As an alternative, some clinicians have been considering ceftaroline because it presents activity against MRSA (6, 7).

Studies have shown that ceftaroline presents penetration ratios in the CSF in humans ranging from 2.4 to 9.4% (8, 9). A recent population pharmacokinetic (popPK) analysis of plasma and CSF concentrations in neurosurgical patients suggested that this penetration could increase up to 62% in the presence of meningeal inflammation (10). Also, studies with rabbits have shown that ceftaroline increases its CSF penetration from 3% to up to 15% in the presence of meningeal inflammation (11).

Although reports have indicated that ceftaroline penetrates the CSF and inflammation could impact this penetration, no study has investigated drug penetration into the extracellular fluid (ECF) space. Given that the interstitium is the site of most bacterial infections, knowing free antimicrobial concentrations in the ECF is critical for correlation with the effects (12). Considering that repeated CSF sampling could interfere with the physiology of the CSF and present technical difficulties for drug extraction due to usually low volumes of the samples in the preclinical setting, brain microdialysis can be employed as a semi-invasive technique that enables the quantification of unbound drug in the ECF over a continuous period (13, 14).

Our research group has successfully applied microdialysis to measure differences in the tissue penetration of antimicrobials and antifungals between infected and healthy individuals (15–19). The distribution process affected by the presence of the infection has been characterized through the employment of popPK analyses, which has given important insights into how the infection status could impact tissue exposure to different drugs (17–20). In this context, the present study aimed to evaluate ceftaroline penetration into the ECF, as determined by microdialysis, to investigate the influence of MRSA brain infection on drug penetration and to develop a popPK model that describes the rate and extent of brain penetration in both healthy and infected individuals.

RESULTS

Preclinical MRSA infection model.

Experimental meningitis was confirmed by bacterial count in brain tissue homogenate after the end of each pharmacokinetic (PK) experiment, showing the presence of viable MRSA (2.92 ± 0.88 log₁₀ CFU/g of tissue). A histopathological analysis revealed moderate multifocal areas of inflammatory infiltrate of neutrophils, lymphocytes, plasma cells, and macrophages in the leptomeninges, in addition to discrete fibrin deposition, with rare myriad basophilic coccoid bacteria (see Fig. S1 in the supplemental material).

PK model.

A total of 655 observations (58 plasma probe recovery observations, 64 brain probe recovery observations, 262 plasma concentrations, and 271 brain concentrations) from 23 rats were included in the popPK analysis. Plasma concentrations were best described by a one-compartment model parameterized in terms of clearance (CL) and central volume of distribution (V₁). The model-building process is summarized in Table S1 in the supplemental material. Brain concentrations were best described by a two-compartment model, with the reversible transfer of unbound drug concentration between the central compartment and the brain through the blood-brain barrier (BBB) described with a bidirectional transport, parameterized as intercompartmental clearance in and out (Q_in and Q_out, respectively), brain volume of distribution (V₂), intercompartmental clearance from the brain compartment to the brain peripheral compartment (Q₃), and brain peripheral volume of distribution (V₃). Figure 1 provides a schematic representation of the selected PK model.

FIG 1.

Ceftaroline structural popPK model. V₁, volume of the central compartment; V₂, volume of the brain compartment; V₃, volume of the brain peripheral compartment; CL, clearance from the central compartment; Q_in, intercompartmental clearance into the brain tissue; Q_out, intercompartmental clearance out of the brain tissue; Q₃, intercompartmental clearance between the brain and brain peripheral compartments; RR_plasma and RR_brain, RRs from the plasma and brain microdialysis probes, respectively. A₁, A₂, and A₃, are amounts of the central compartment, brain compartment, and brain peripheral compartment, respectively.

Free plasma and brain concentrations were calculated with equations 1 and 2:

$$\mathit{Plasma}\hspace{0.17em}\mathit{concentration}=\frac{A_1}{V_1}\hspace{0.17em}.\hspace{0.17em}R{R}_{\mathit{plasma}}$$ (1)

$$\mathit{Brain}\hspace{0.17em}\mathit{concentration}=\frac{A_2}{V_2}\hspace{0.17em}.\hspace{0.17em}R{R}_{\mathit{brain}}$$ (2)

where RR_plasma and RR_brain are the relative recoveries (RRs) from the plasma and brain microdialysis probes, respectively, and A₁, A₂, V₁, and V₂ are amounts and volumes of ceftaroline in the central and brain compartments, respectively. The RR values were calculated as follows (equation 3):

$$C_{\mathit{out}}=C_{\mathit{in}}-(C_{\mathit{in}}\hspace{0.17em}.\hspace{0.17em}R{R}_{\mathit{plasma}\hspace{0.17em}\mathit{or}\hspace{0.17em}\mathit{brain}})$$ (3)

where C_in and C_out are the ceftaroline concentration entering the probe and the ceftaroline concentration leaving the probe, respectively. IIV was included in RR_plasma, V₁, Q_in, and V₂.

Covariate analysis.

Table S2 in the supplemental material shows the stepwise covariate model-building process. With the available covariates, the RR_plasma parameter showed good correlation with the cardiac output (CO) (Fig. 2), which is described with equation 4.

$$R{R}_{\mathit{plasma}}={\theta }_{R{R}_{\mathit{plasma}}}\hspace{0.17em}.\hspace{0.17em}{e}^{({\theta }_{\mathit{CO}}\hspace{0.17em}.\hspace{0.17em}\left(\mathit{CO}-C{O}_{\mathit{median}}\right))}$$ (4)

where CO_median is the median value for the observed CO. The IIV on RR_plasma was explained by the insertion of this continuous covariate (CO), decreasing from initially 49% to 31.6%, which resulted in an objective function value (OFV) reduction of 19.16 points (P < 0.01).

FIG 2.

Correlation between the model-predicted individual RR_plasma and the CO of the animals.

The parameter Q_in was significantly impacted by the infection status (P < 0.01), when estimated separately for healthy and infected animals (0.0370 L/h and 0.0598 L/h, respectively). This parameter was explained by a reduction in the IIV from 44% to 37.5% and an OFV reduction of 7.18 points. The inclusion of this covariate also resulted in a better fit of the prediction-corrected visual predictive check (pcVPC) plots.

The higher Q_in in infected animals resulted in a higher penetration ratio (Q_in/Q_out) for infected animals, compared with healthy animals (27% versus 17%).

Model parameter estimates and model evaluation.

The parameters of the final popPK model are shown in Table 1, together with uncertainties and distributions describing the IIV in model parameters. The low relative standard errors (RSEs) and the 95% confidence intervals (CIs) calculated from the bootstrap analysis not lower than or equal to zero demonstrated that the parameters were precisely estimated by the model.

TABLE 1.

Parameter estimates for the final ceftaroline popPK model

Parametera	Estimate (RSE [%])	Mean bootstrap value for estimate 95% CIb	CV (%) for IIV (RSE [%])	Mean bootstrap value for IIV (IQR) (%)b	Shrinkage (%)
RRplasma (%)	15.8 (9.2)	16.1 (12.4 to 20.8)	31.8 (15.8)	30.6 (19.8–39.5)	1.2
RRbrain (%)	14.7 (5)	14.6 (13.0 to 16.3)
CL (L/h)	0.646 (6.5)	0.657 (0.513 to 0.854)
V1 (L)	0.359 (7.1)	0.363 (0.284 to 0.477)	11.9 (17.3)	12.2 (7.2–17.5)	5.4
Qin (L/h)	0.037 (16.9)	0.0423 (0.0101 to 0.0948)	37.5 (15.2)	35.6 (26.4–42.5)	0.1
Qin, infected (L/h)	0.0598 (14.5)	0.0670 (0.0181 to 0.137)
Qout (L/h)	0.224 (10.6)	0.241 (0.0711 to 0.449)
V2 (L)	0.0291 (16.5)	0.0323 (0.00885 to 0.0679)	101 (18.3)	98.4 (73.4–123.1)	13.1
Q3 (L/h)	0.0307 (15.8)	0.0345 (0.00924 to 0.0731)
V3 (L)	0.0258 (11.5)	0.0280 (0.00879 to 0.0506)
θCO	−0.477 (18.4)	−0.492 (−0.715 to −0.340)
Plasma additive error (mg/L)	0.193 (5.1)	0.192 (0.163 to 0.227)
RRplasma additive error (mg/L)	0.113 (8.9)	0.108 (0.0697 to 0.147)
Brain additive error (mg/L)	0.168 (4.9)	0.166 (0.139 to 0.194)
RRbrain additive error (mg/L)	0.0721 (8.6)	0.0707 (0.0552 to 0.0842)

^aRR, relative recovery for the referred microdialysis probe; CL, clearance; V₁, central volume of distribution; Q_in, intercompartmental clearance from the central compartment to the brain compartment; Q_out, intercompartmental clearance from the brain compartment to the central compartment; V₂, brain volume of distribution; V₃, brain peripheral volume of distribution; Q₃, intercompartmental clearance from the brain compartment to the brain peripheral compartment; CV, coefficient of variation; 95% CI, 2.5th–97.5th confidence interval; θ_CO: θ_CO, covariate effect of cardiac output on RR_plasma.

^bBootstrap analysis was performed with 1,000 datasets, with 956 successful runs.

The pcVPCs (Fig. 3) stratified by plasma, brain from healthy animals, and brain from infected animals indicated adequate goodness of fit and good predictive performance of the final popPK model for both plasma and brain concentration data. Goodness-of-fit plots are presented in Fig. 4 and 5 and showed a good agreement between the observed and predicted data, demonstrating a lack of model misspecification.

FIG 3.

pcVPC plots for ceftaroline in plasma (A), brain of healthy animals (B), and brain of MRSA-infected animals (C) based on 1,000 simulated data sets. Circles represent the observed data. The shaded areas are 95% CIs for the 2.5th percentile (gray), 50th percentile (green), and 97.5th percentile (gray) prediction intervals, based on the simulated data. The solid black lines represent the median observed concentrations, and the dashed black lines represent observed 2.5th and 97.5th percentiles.

FIG 4.

Observed versus populational predicted (left) and individual predicted (right) ceftaroline concentrations in plasma (A) and brain (B). The solid gray lines represent the lines of identity (x = y), and the dashed black lines represent the linear regression lines of fit. Individual data points are indicated by filled circles (healthy animals) or open circles (MRSA-infected animals).

FIG 5.

Conditional weighted residual errors (CWRES) versus populational predicted concentrations (left) and time (right) for plasma (A) and brain (B). The dashed black lines represents smooths, and the horizontal solid gray lines are the zero lines. Individual data points are indicated by filled circles (healthy animals) or open circles (MRSA-infected animals).

Simulations and PTA analysis.

The probability of target attainment (PTA) versus MIC profiles corresponding to simulations of the healthy and infected animals after humanized dosing (50 mg/kg every 8 h as a 2-h intravenous infusion) for the pharmacodynamic (PD) targets of the percentage of the dosing interval that the unbound concentrations remained above the MIC (%fT_>MIC) of 26.8%, 30.7%, and 34.7% are represented in Fig. 6. For both healthy and infected groups, the percentage of simulated animals achieving the plasma PK/PD target was >90% with a MIC of 2 mg/L for the targets of %fT_>MIC of 26.8% and 30.7% and a MIC of 1 mg/L for the higher target of 34.7%. Considering brain concentrations, assuming the same PK/PK index reported for plasma, the PTAs are lower for the noninfected animals than for the MRSA-infected group. For a MIC of 0.5 mg/L, the percentages of infected animals achieving the targets of %fT_>MIC of 26.8% and 30.7% are 96.8% and 93.5%, respectively; for healthy animals, the percentages are 79.4% and 62.9%, respectively. Regarding the higher target, 82.6% of the infected animals had unbound concentrations above 0.5 mg/L for 34.7% of the dosing interval. Both groups achieved all the PK/PD targets for the modal MIC (0.25 mg/L), according to European Committee for Antimicrobial Susceptibility Testing (EUCAST) guidelines (https://mic.eucast.org/search/).

FIG 6.

PTAs for 1,000 simulated healthy rats (plasma represented as red continuous lines and brain as blue-green dashed lines) and 1,000 simulated infected rats (plasma represented as black dot-dashed lines and brain as gray dashed lines) achieving %fT_>MIC targets of 26.8% (A), 30.7% (B), and 34.7% (C) for MRSA by MIC following the administration of ceftaroline at 50 mg/kg every 8 h as a 2-h intravenous infusion overlaid with MIC distribution for MRSA from EUCAST (light gray bars). The horizontal gray lines represent the 90% PTA.

DISCUSSION

Infections of the CNS are medical emergencies associated with high morbidity and mortality rates (21). Optimal antimicrobial treatment requires adequate drug exposure, which demands sufficient penetration through the barriers of the CNS. Infections could lead to different degrees of drug penetration due to pathophysiological alterations of the brain barriers (22). In the present study, unbound ceftaroline plasma and brain concentrations were determined in healthy and MRSA-brain-infected Wistar rats. By applying the popPK approach, we were able to describe the influence of infection on the penetration of ceftaroline into the brain, resulting in higher exposure levels during an active meninges infection. This higher exposure led to increased percentages of animals reaching the PK/PD targets, as shown through simulations and PTA analysis. Also, we were able to identify an important inverse correlation between the RR_RD of the plasma microdialysis probe and the CO of the animals, revealing an interesting cause for the high variability usually found in plasma microdialysis experiments.

The ceftaroline plasma free concentration was adequately described by a one-compartment model, with IIV in the parameters of central volume of distribution (11.9%) and a higher variability found in the RR for the plasma microdialysis probes (31.8%). No differences were found in the plasma disposition of the drug between healthy and infected animals, showing that probably this animal model of infection does not cause systemic alterations in the rats.

The variability found in the RR_RD for the plasma microdialysis probes could be partially explained by the animal CO. A significant inverse correlation was identified, demonstrating that animals with higher CO values showed lower RR_RD values for the plasma microdialysis probe. Considering that the recovery of an analyte is strongly influenced by the processes in the tissue that remove the drug from the extracellular space (23), this correlation is justifiable. Animals with a lower CO, in other words, with a smaller volume of blood being pumped by the heart by minute, have less capability of clearing the drug from the surroundings of the probe, resulting in greater relative drug recoveries by the probes. This is not the first work that identified important covariates in the RR_RD of microdialysis probes. Busse and colleagues (24) analyzed the RR_RD of subcutaneous microdialysis probes for seven different drugs and discovered that the obesity status of the patient was associated with lower RR_RD for six of them. They were able to characterize an inverse correlation between calculated fat mass and RR (24).

Ceftaroline brain disposition was adequately described by two compartments, with bidirectional transport between the plasma and brain compartments, parameterized as CL_in and CL_out. IIVs were included in the parameters of Q_in (37.5%) and volume of the brain compartment (101%). Since low variability was found between the RR_RD of the brain microdialysis probes, the IIV encountered in this parameter was small, and we decided on its exclusion from the model. This decision was also supported by the low unexplained residual variability (0.0721), much lower than that found for the plasma microdialysis probes (0.113). Also, the high variability found in the volume of distribution of the brain compartment could not be explained by the covariates available for the analysis.

The infection led to significant differences in the exchange of ceftaroline from blood to brain ECF. Infected rats had an increase of about 60% in the parameter Q_in, in comparison to healthy rats. This difference resulted in greater brain exposure in the group of infected rats. This phenomenon has already been described by Alves et al. for both voriconazole (17) and fluconazole (20) in experimental cryptococcal meningitis in Wistar rats. For both antifungals, the infection resulted in greater exposure to free drugs, as determined by microdialysis, compared to the healthy state. It is thought that the inflammatory response by bacteria/fungi leads to alterations in the BBB, which becomes leaky due to opening of the intercellular tight junctions of the vessel walls (22).

With regard to ceftaroline CSF penetration, scarce data are available. Stucki et al. (11) compared the total levels of ceftaroline in the CSF of rabbits with and without meninges inflammation caused by Gram-negative bacterial meningitis. They found increased CSF penetration into inflamed meninges (15%), in comparison to uninflamed meninges (3%), as determined by the ratio of the areas under the concentration-time curves (AUCs) for CSF and serum (11). In a study with a rabbit MRSA meningitis model, the authors compared the penetration of ceftaroline and vancomycin by measuring the CSF and total serum concentrations 1 h after the sixth dose. They reported a mean penetration of 51 ± 15% for ceftaroline and a lower penetration for vancomycin, between 14.7% and 35% (25). Those findings support the idea that ceftaroline has increased penetration through inflamed meninges and probably better penetration than vancomycin (4).

In humans, some case reports have shown penetration of ceftaroline in CSF. A patient with a ventriculopleural shunt infection caused by MRSA received ceftaroline, and total serum and CSF concentrations at steady state were determined simultaneously. Based on the AUC_CSF/AUC_serum ratio, the estimated penetration was 8.4% (8). Roujansky and colleagues (26) reported a case of an elderly patient who presented a series of ventriculostomy-related infections, with one of the infections being caused by multidrug-resistant Staphylococcus epidermidis. Penetration values of 2.6 to 4.8% were reported, corresponding to the ratio of trough concentrations in plasma and CSF at steady state (26). In another study, nine neurosurgical patients who required insertion of an external ventricular drain received one dose of 600 mg of ceftaroline in a 1-h intravenous infusion, and plasma and free CSF samples were obtained. The patients showed limited meningeal inflammation, and a ceftaroline mean penetration value (AUC_CSF/AUC_plasma) of 6.4% was reported (9). Interestingly, those authors found an inverse correlation between clearance into the CSF and the glucose CSF concentration, suggesting greater penetration through inflamed meninges (9). This correlation was further explored by our group (11) by developing a popPK model; through Monte Carlo simulations, it was possible to show that ceftaroline CSF penetration could reach up to 62% when an inflamed meningeal state was considered.

However, it is noteworthy that all of these reports describe ceftaroline penetration into CSF, thus reflecting drug passage through the blood-CSF barrier (BCSFB). Although the CSF is in close contact with brain ECF and thus is expected to reflect the brain ECF concentrations, this is not always the case (13, 27). There are pronounced differences between the BBB and the BCSFB, with the latter being considered leakier (28). Additionally, the clearance from the CSF is higher than that from brain ECF, which is determined by diffusion back to the blood through BCSFB and by bulk flow of the CSF into the venous blood (22). Consequently, the concentrations in the CSF could be higher or lower than those found in the brain ECF (27), without a straightforward correlation between them.

Considering that greater-than-expected penetration of ceftaroline in the brain ECF for healthy animals (mean of 17%, based on the Q_in/Q_out ratio) was found in the present study, it can be hypothesized that the CSF concentrations might be lower than the concentrations in brain ECF. This was already reported by Hosmann et al. (29) for meropenem in humans by comparing the drug concentrations at steady state in blood, CSF, and cerebral microdialysis fluid of neurointensive care patients. The authors found an AUC_CSF/AUC_ECF ratio of 0.41 ± 0.16, suggesting greater exposure in the brain ECF than in the CSF. Therefore, we think that the discrepancies found between our work and previous work with rabbits could be explained by different penetrations in the CSF and brain ECF compartments.

In order to investigate whether ceftaroline concentrations achieved in the brain were effective for the treatment of MRSA infections, we performed PTA analysis using simulated unbound plasma and unbound brain concentrations after administration of a humanized dose of 50 mg/kg to rats, which is equivalent to a 2-h intravenous infusion of 600 mg every 8 h in humans. This dose was proposed following a PTA analysis of CSF concentrations in humans (10). The PTAs from plasma were identical for healthy and MRSA-infected animals, which was expected, since we did not find differences in the plasma PK between the two groups. Considering the previous analysis of PTAs in humans, the plasma exposure in rats resulted in lower PTA values, with the 90% target (%fT_>MIC of >34.7%) being reached for MICs up to 1 mg/L, in comparison to up to 4 mg/L estimated for humans (10). We think that this difference could be related to the short half-life in rats and greater binding to plasma proteins, in comparison to humans, leading to lower concentrations than in humans (30). Also, we determined the actual unbound plasma concentrations, as opposed to values corrected after measurement of the total concentration in humans. Regarding the brain, both infected and noninfected animals reached sufficient concentrations for the treatment of infections, when the modal MIC (0.25 mg/L) according to EUCAST was considered. The greater brain penetration in infected animals resulted in differences in the PTA when a MIC of 0.5 mg/L was considered. For %fT_>MIC targets of 26.8% and 30.7%, more than 90% of the MRSA-infected animals would attain the targets, while less than 80% of the healthy animals would attain the same targets if the ceftaroline MIC was 0.5 mg/L. These results suggest that, in the presence of an active infection, there is increased penetration of ceftaroline into the brain, as already hypothesized for humans (10), leading to more effective exposures for the treatment of CNS infections.

In conclusion, the popPK model developed was able to successfully describe unbound plasma and unbound brain ceftaroline concentrations in healthy and MRSA-infected Wistar rats. We were able to identify significant sources of variability in plasma microdialysis, which could aid further studies. Our results also demonstrated that infected animals presented greater brain exposure to the drug and presented promising results in the PTA analysis. Therefore, we postulate that ceftaroline should be considered an option for treating CNS infections. Nevertheless, further studies are necessary to confirm that such alterations also occur in the patient population.

MATERIALS AND METHODS

Chemicals and reagents.

Ceftaroline hydrochloride was provided by Pfizer (USA), and ceftaroline fosamil was purchased from Wyeth (Brazil). Acetonitrile and methanol (high-performance liquid chromatography [HPLC] grade) were obtained from Merck (Brazil). All other chemicals and reagents were of analytical grade and were purchased from commercial sources. Ringer’s solution (147 mM NaCl, 2.3 mM CaCl₂, 4 mM KCl) was prepared in-house. Artificial CSF (ACF) solution (147 mM NaCl, 2.7 mM KCl, 1.2 mM CaCl₂, 0.85 mM MgCl₂) was also prepared in-house. The solutions were filtered through a 0.47-μm filter prior to use.

Preclinical MRSA infection model.

This study was approved by the Ethics Committee in Animal Use from the Federal University of Rio Grande do Sul (CEUA/UFRGS number 35635) and was conducted in compliance with the principles of laboratory animal care from the National Council for Animal Experimentation (CONCEA) (Brazil). Male Wistar rats (150 to 350 g) were obtained from the Center for Laboratory Animals Reproduction and Experimentation of the Federal University of Rio Grande do Sul (Porto Alegre, Brazil).

The experimental meningitis was induced in the animals using MRSA (ATCC 43300), by direct intracisternal injection of 10 μL of ACF solution containing 1× 10⁸ CFU/mL of the MRSA strain with a 30-gauge insulin syringe in anesthetized animals.

Bacterial counts in the brains of the animals were determined to ensure development of infection. Briefly, animals were euthanized 3 or 5 days after inoculation, and the brains were aseptically excised and homogenized with sterile saline (0.5 g of brain/mL of saline). The homogenates were serially diluted and plated on Mueller-Hinton and mannitol salt agar, and colonies were counted after overnight incubation at 37 ± 1°C.

Histopathological analysis of the brain (n = 2) 3 days after inoculation was conducted. The tissues were fixed in a 10% neutral buffered formalin solution, embedded in paraffin blocks, sliced, stained with eosin and hematoxylin dyes, and microscopically analyzed.

Surgical procedure.

Forty-eight hours before plasma and brain microdialysis experiments, the animals were anesthetized with ketamine and xylazine (100 and 10 mg/kg, respectively) administered intraperitoneally. For plasma microdialysis, a CMA 20 probe (4 mm, 20-kDa cutoff; CMA, Sweden) was inserted in the right jugular vein, and its outlet was passed subcutaneously and exteriorized at the posterior surface of the neck. To avoid clotting, the probes were perfused for 10 min with heparinized Ringer’s solution (100 IU/mL heparin), at a flow rate of 2 μL/min, before and after insertion. For brain microdialysis, a CMA 12 guide cannula was placed in the primary motor cortex of animals (A, +2.2 mm; L, +2.8 mm; V, −3.6 mm relative to bregma) (31) and fixed with two screws and dental cement (JET, Brazil). Animals were allowed to recover in individual polypropylene boxes with food and water available ad libitum until the experiment.

In vivo microdialysis probe recovery.

The ceftaroline in vivo RR was determined by retrodialysis (RR_RD) in all animals, for both plasma and brain probes. On the day of the microdialysis experiments, the animals were placed in a CMA 120 system for freely moving animals, and the guide cannula was replaced with the CMA 12 probe (3 mm, polyaryl ether sulfone-polyvinylpyrrolidone [PAES] membrane, 20-kDa cutoff; CMA). The brain and plasma probes were continuously perfused with ACF and Ringer’s solution, respectively, both containing ceftaroline at 100 ng/mL, at a flow rate of 2 μL/min for 1 h for equilibration. Microdialysate samples were collected every 15 min up to 1 h, immediately frozen at −20 ± 1°C, and then stored at −80 ± 2°C until the analysis. After this period, the perfusion solutions were replaced with ACF or Ringer’s solutions at a flow rate of 3 μL/min for 1 h to allow probe wash out. The flow rate was adjusted to 2 μL/min for a 30-min equilibration before the start of the experiment.

Ceftaroline unbound plasma PK and unbound brain distribution.

To determine ceftaroline brain penetration, male Wistar rats were divided into three groups, i.e., healthy (n = 8), 3 days of infection (n = 9), and 5 days of infection (n = 6). After determination of the RR_RD, all animals received ceftaroline fosamil (20 mg/kg) by intravenous bolus injection into the lateral tail vein. Microdialysis samples were collected from plasma and brain every 15 min up to 3 h, frozen immediately at −20 ± 1°C, and then stored at −80 ± 2°C until analysis.

A previously validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) method was used to quantify ceftaroline in microdialysate samples, which were injected directly into the system without sample processing (32). The mobile phase consisted of water (A) and acetonitrile (B), both with 5 mM ammonium formate and 0.1% formic acid, and elution was performed with a C₁₈ column (Phenomenex Luna, 50 by 2.0 mm, with a particle size of 3 μm; Phenomenex, USA). Elution at a flow rate of 0.5 μL/min with gradient, starting with 5% of B increasing to 95% in 0.5 min, maintained in 95% of B until 2.3 min, returning to initial conditions at 2.4 min, and maintained until 4 min; 2 min equilibrium time between analysis was applied. Ceftaroline was monitored with the transition of m/z 604.89 to m/z 209.3 (positive-ion electrospray ionization), in a concentration range of 0.5 to 500 ng/mL for brain microdialysate and 0.5 to 2500 ng/mL for plasma microdialysate (r² ≥ 0.997). The interassay and intraassay precision and accuracy were within ±15% (33).

Data analysis.

Experimental data were analyzed using the nonlinear mixed-effect modeling approach in NONMEM (34) with the first-order conditional estimation method with interaction. Model management was done in Pirana (35). Perl-speaks-NONMEM (36) was used for automating and controlling the runs. Graphical analysis was conducted with the ggplot2 (37) and Xpose4 (35, 38) libraries on R.

None of the data was below the quantification limit; missing data represented <5% of the total observations and were not included in the data set. Values of observed concentrations were logarithmically transformed for the analysis. The microdialysate concentration data were described by the integral over each collection interval and with the integrated model, estimating the blood and brain probe recoveries (39). IIV was modeled exponentially, and residual variability was described with an additive error model on a logarithmic scale.

Structural model building was performed in a sequential manner. A model for the plasma microdialysate data was developed, and then the brain microdialysate data were added. Selection between model candidates was performed based on changes in the minimum value of the OFV, as well as visual exploration of the standard goodness-of-fit plots.

The effects of the continuous covariates body weight and CO (40) were investigated for all parameters. The impact of the categorical covariate infection was also tested for inclusion in all parameters. They were included in the model based on statistical significance, following a stepwise forward addition (P < 0.05) and backward elimination (P < 0.01) procedure (41).

Evaluation of the selected popPK model was performed according to visual inspection of the goodness-of-fit plots, condition number, precision of model parameters (reflected as the RSE), and pcVPC. For each condition (healthy and infected, both 3 days and 5 days after inoculation), 1,000 simulated profiles were generated and displayed graphically together with the experimental data. Parameter precision was further evaluated from the analysis of 1,000 nonparametric bootstrap data sets, and the medians and CIs were determined.

Simulations and PTA analysis.

To investigate whether ceftaroline concentrations reached in the brain were effective for the treatment of brain infections, the developed popPK model was used to simulate free brain and plasma concentrations in 1,000 healthy and 1,000 infected animals following administration of an allometrically scaled dose of 50 mg/kg every 8 h as a 2-h intravenous infusion (41). The simulated concentrations were used to analyze the PTA of the PK/PD index %fT_>MIC. The %fT_>MIC was calculated for each group at MIC values ranging from 0.06 to 8 mg/L. The proportions of animals reaching the MRSA PK/PD targets of 26.8% for bacteriostasis, 30.7% for a 1-log₁₀ CFU reduction, and 34.7% for a 2-log₁₀ CFU reduction were determined and compared with the MIC frequency distributions reported on the EUCAST MIC distribution website (https://mic.eucast.org/search/).