A significant increase in the incidence of Candida-mediated infections has been observed in the last decade, mainly in susceptible individuals associated with high mortality and morbidity (Katsipoulaki et al., 2024). To be a potentially pathogenic yeast, it has to show a wide range of virulence factors and physical characteristics (Colombo et al., 2013; Mayer et al., 2013). Candida albicans is the most prevalent agent causing invasive candidiasis (McCarty et al., 2021), and it is also the most isolated species pertaining to these infections from different anatomical sites (Hiller et al., 2011; Kolaczkowski et al., 2010), affecting organs and tissues, such as the lungs, brain, kidneys, bladder, joints, liver, heart, and eyes (Peixoto et al., 2014), with the kidney being the most affected organ (Balk et al., 1978).

For the treatment of invasive candidiasis, amphotericin B (AmB), a drug representing the polyenic class, has significant importance in clinical practice because of its broad spectrum of activities (Ellis, 2002; Wang et al., 2021). Nevertheless, treatment failure is common in infectious diseases (Tøttrup et al., 2014). Insufficient concentration of the used drug at the infection site is one of the leading causes of pharmacological treatment failure (Lagler and Zeitlinger, 2014). For therapeutic success, drug penetration must be sufficient to achieve concentrations that can eliminate or stop the growth of the microorganisms (Felton et al., 2014; Niemirowicz et al., 2017; Ray et al., 2015).

Most infection sites are extravascular, and pharmacological treatment depends on drug diffusion out of the bloodstream and into the interstitial space and intracellular fluid. The penetration capacity of a drug depends on tissue-related factors, such as perfusion, surface area, vascular characteristics, pH, inflammation, protein composition modification, drug transport through efflux pumps (Felton et al., 2014), drug-related factors, and their physicochemical characteristics such as lipophilicity, molecular size, pKa, and plasma protein binding (Hurtado et al., 2014). For these reasons, the use of plasma drug concentrations to determine pharmacokinetic parameters will only yield satisfactory results if blood levels are an appropriate substitute for the levels at the site of infection (Levison and Levison, 2009; Müller et al., 2004).

In this context, the microdialysis technique has important applications, becoming valuable for free drug sampling in the interstitial fluid distribution study (Lagler and Zeitlinger, 2014; MacVane et al., 2014; Lock et al., 2023) in preclinical and clinical pharmacokinetic studies in various tissues such as the bone (Tøttrup et al., 2014), prostate (Hurtado et al., 2014), brain (Levison and Levison, 2009), pancreas (Liu et al., 2014), and pulmonary epithelium (Rottbøll et al., 2015). In addition, this technique allows the comparison of drug penetration in healthy and infected tissues (Joukhadar et al., 2001; Lock et al., 2023).

To date, no study has been conducted that describes the relationship between the total plasma AmB concentrations and free tissue concentrations using microdialysis. Thus, considering the scenario of invasive candidiasis caused by C. albicans and the common use of AmB for its treatment, the importance of a pharmacokinetic study in an in vivo experimental model emerges to evaluate the influence of infection on pharmacokinetic parameters and in kidney penetration of AmB to relate whether plasma concentrations can predict drug-free tissue levels in the target tissue of invasive candidiasis, with this relationship being established by a population pharmacokinetic (PopPK) model.

The relative recovery rates of AmB determined in vitro by the dialysis methodology (RR_D) were 8.1% ± 4.1%, 22.5% ± 8.5%, and 8.0% ± 3.0%, and those determined by the retrodialysis methodology (RR_RD) were 12.9% ± 2.3%, 20.0% ± 6.0%, and 9.2% ± 2.7%, for the 1.0, 1.5, and 2.0 μL/min flow rates, respectively. For the same flow rates, no statistically significant difference was observed between the two methods (P > 0.05), showing that AmB does not bind to microdialysis tubing and that retrodialysis can be used to determine the relative recovery in vivo; the flow rate of 1.5 μL/min is used for subsequent tests as it presents greater relative recovery.

The influence of the drug concentration on the microdialysis recoveries of AmB was evaluated by using a 1.5 μL/min flow rate. The average relative recovery of the analyzed concentrations (2.5, 3.0, and 4.0 μg/mL) obtained by dialysis was 33.1% ± 7.6%, and by retrodialysis, it was 34.5% ± 2.4%. No statistically significant differences (p > 0.05) were observed in the values obtained between the methods or between the AmB concentrations tested. The RR_RD in vivo was determined to be 10.9% ± 3.7%.

AmB protein binding was determined to be 56.4% ± 4.6% for 2 μg/mL and 25.5% ± 3.9% for 8 μg/mL in the plasma of healthy rats, and it was 60.9% ± 12.1% for 2 μg/mL and 41.8% ± 1.7% for 8 μg/mL in the plasma of C. albicans-infected rats, demonstrating it to be nonlinear and dependent on the AmB plasma concentration.

AmB total plasma and free kidney concentrations versus time profiles after the administration of a 2.5 mg/kg i.v. dose to study the plasma pharmacokinetics and kidney penetration in healthy and C. albicans-infected rats are shown in Figure 1. The AmB kidney penetration was 0.77 in healthy rats and 0.71 in infected rats. The pharmacokinetic parameters showed no statistical difference (P > 0.05) for healthy and C. albicans-infected animals and between the plasma and kidney. The mean AUC_{0- ∞} was 96.3 ± 54.8 μg h/mL for plasma and 74.1 ± 31.7 μg h/mL for tissue in healthy animals and 59.8 ± 30.1 μg h/mL for plasma and 42.3 ± 36.0 μg h/mL for tissue in C. albicans-infected rats.

In the population pharmacokinetic analysis, a two-compartment model with linear elimination for both the plasma total and free kidney concentrations was selected as the final model (Figure 2). The system of differential equations for the final model is given in Equations 5–8: d A 1 d t = A 2 × k 21 − A 1 × k 12 − A 1 × k 10 . (5) d A 2 d t = 0.56 × A 1 × k 12 − A 2 × k 21 . (6) C p l a s m a   = A 1 V 1 . (7) C k i d n e y = A 2 V 2 . (8) Here, A 1 and A 2 are the drug amounts in the central and kidney compartments, respectively, t is time, k 12 and k 21 are the first-order distribution rate constants, C p l a s m a is the observed plasma concentration, C k i d n e y is the free concentration measured in the kidney, V 1 represents the volume of the central compartment, and V 2 is the volume of distribution in the kidney compartment.

The estimated parameter values are presented in Table 1. The best residual error model for the total plasma and free kidney concentrations was the proportional model, which was based on goodness-of-fit and lower AIC. In the final model, inter-individual variability (IIV) was incorporated in V₁, V₂, k₁₀, and k₁₂. The model validation methods showed predictive performance for the AmB model. The VCPs for total plasma and free kidney concentrations are shown in Figure 3, confirming the adequacy of the model. The VPCs confirmed that the model was adequate to simultaneously describe the data of all the evaluated groups. The categorical covariate infection by C. albicans did not influence the model performance (p > 0.05). The stability of the final model was assessed by bootstrap analysis, with parameters estimated by the model agreeing with the respective confidence intervals estimated for 200 bootstrap replicates (Table 1).

This is the first study that assesses the renal penetration of AmB using microdialysis in healthy and C. albicans-infected rats. The sites of candidiasis systemic infections are extravascular, so the pharmacological treatment depends on the diffusion of the antifungal drug out of the bloodstream and into the interstitial space and intracellular fluid (Levison and Levison, 2009). Drug penetration must be sufficient to achieve concentrations that eliminate or stop the growth of the fungus, having therapeutic success, and, ideally, the plasma concentration should reflect tissue concentrations (Felton et al., 2014).

Microdialysis has proven to be a powerful tool to investigate the tissue distribution of antimicrobial agents. Compared with traditional sampling methods, microdialysis is the only technique that allows the collection of the protein unbound drug since the unbound fraction of the drug is available for absorption, distribution, metabolism and elimination, and delivery to the target sites for pharmacodynamic actions. Furthermore, microdialysis allows systemic and local administration of drugs, collection of small-molecular weight substances from the extracellular space, and conducting of the experiment in awake animals because the perfusion of the probe is not painful. Regarding limitations, microdialysis is an invasive procedure, and there are analytical difficulties with some molecules and low membrane recoveries with high-molecular-weight compounds (Parent et al., 2001; Lietsche et al., 2014; Tsai, 2003).

In vitro calibration of microdialysis probes is considered an essential step for further in vivo studies. Although in vitro calibration does not eliminate the need to assess drug recovery in vivo, it provides essential information on drug gain and loss and the feasibility of in vivo calibration (MacVane et al., 2014). It was possible to define the microdialysis workflow by evaluating the influence of the different test flow rates on the relative recovery of AmB and drug concentration. Among all flow rates tested, the relative recovery of AmB was greater in the flow rate of 1.5 μL/min (the flow being chosen for this work) with RR_D 22.5% ± 8.5% and RR_RD 20.0% ± 6.0%, with no statistical difference between rates of recovery determined by the two methods used.

There is a change in recovery for different flow rates. In general, lower flow rates result in higher relative recovery values, and high flow rates result in lower recovery rates (Plock and Kloft, 2005). The flow rate of 1 μL/min showed less relative recovery when compared to the flow rate of 1.5 μL/min. This phenomenon can occur in perfusion flows below or equal to 1 μL/min due to membrane adsorption processes (Kjellström et al., 1998). Low flow rates also increase the collection interval due to the sample volume and the quantification limit of the analytical method. On the other hand, high flow rates can be conducive to the ultrafiltration process due to the pressure accumulated in the microdialysis tube, resulting in liquid flow out of the probe (Plock and Kloft, 2005), as was the case with the 2 μL/min flow rate, which had a less relative recovery.

In assessing the influence of AmB concentration on recovery from microdialysis, the results showed that there was no statistical difference in the relative recovery among the concentrations and methods evaluated, showing that the recovery of AmB is independent of the investigated concentration range and will not change in concentrations in floating tissue.

The RR_RD in vivo was lower and statistically different from the recovery determined in vitro by retrodialysis at the same rate. This phenomenon is due in large part to the reduced volume of fluid and the increase in tissue tortuosity, with an increase in the diffusion length caused by the impediment imposed by cellular structures as to the connectivity of spaces. In addition, tissue diffusion can be further delayed by binding the analyte to cell surface proteins along the diffusion path, indicating that the in vivo recovery of an analyte must be less than the in vitro recovery of the same analyte in a solution (Shippenberg and Thompson, 2001), as previously reported for other drugs (Torres et al., 2017).

AmB total plasma and free kidney concentration–time profiles were obtained after the administration of 2.5 mg/kg dose. This dose, still in AmB linearity (Walsh et al., 1998), was used in the pharmacokinetic study due to the low in vivo relative recovery of AmB and the quantification limit of the validated analytical method. After the administration of this dose, no change was observed in the pharmacokinetic parameters and renal distribution of AmB during C. albicans infection in rats using a bi-compartmental model with kidney/plasma redistribution and linear elimination. The final model adequately describes the data with adequate diagnostic parameters and graphics. Observations versus model predictions, GOF, and VPC plots (Figure 3) support the adequacy of the PK model in describing the experimental data.

Some studies demonstrate that an infection can alter the physiological characteristics of the infected tissue due to the inflammatory process developed by the host to eliminate the invasive organism, as observed in soft tissue infections (Joukhadar et al., 2001), septic shock (Joukhadar et al., 2002), or infections (Tegeder et al., 2002) with several antibiotics. Moxifloxacin (Joukhadar et al., 2003), imipenem (Tegeder et al., 2002), piperacillin, and cefpirome (Joukhadar et al., 2002) had less penetration into inflamed, infected tissues, and critical volunteers. There is no clear possible explanation for the poor distribution of tissues and organs. It is expected that the local inflammation itself increases capillary permeability, but studies suggest that because of the systemic inflammatory response of the infection, there is capillary leakage, generating interstitial edema. Likewise, other studies indicate that an infection, mainly caused by Candida sp, does not interfere with the renal penetration of drugs, as with voriconazole (de Araujo et al., 2009) and fluconazole (Azeredo et al., 2012), and therefore, free kidney concentration is predicted by total plasma concentration. Thus, it is suggested that the phenomenon observed with AmB in this study is not exclusive to this drug but is probably shared by other anti-infective agents.

AmB concentrations decreased rapidly during the first 4 h after administration and remained almost constant until 36 h, revealing a phase of slow elimination, as previously reported in Fielding et al., 1991; Fielding et al., 1991. Studies using microdialysis to determine the free tissue concentrations of AmB were not found in the literature. Population parameters showed a considerable tissue distribution of AmB, with V₂ of 328.58 mL/kg and k₁₂ 91.58 h^-1. The entire free concentration of AmB in the plasma can be distributed to the kidneys.

AmB protein binding determined for two plasma concentrations (2 and 8 μg/mL) in vitro using both healthy and C. albicans-infected rat plasma demonstrated that protein binding was concentration-dependent. In the literature, there are no data on the connection to plasma proteins of AmB in rats, although this drug has been used and studied for a long time. The human data demonstrate that AmB is mainly bound to plasma lipoproteins (Brajtburg et al., 1984) and also human serum albumin and human alfa-1-acid glycoprotein (Bekersky et al., 2002); Bekersky et al. (2002) evaluated the in vitro binding of AmB to human plasma using increasing concentrations of drugs. AmB was shown to be distinctly nonlinear and dependent on the concentration, which is not expected for most drugs due to the saturation of the binding site with increased plasma drug concentration (Bekersky et al., 2002). Nonlinear protein binding has been studied and discussed by several authors after some tetracyclines presented this type of behavior and showed the importance of the exposure and penetration of the drug in the tissues (Deitchman et al., 2018). Still, regarding the results of AmB binding to proteins, there was no difference in the binding of the plasma of healthy and C. albicans-infected rats. The concentration dependency in our study did not interfere drastically since the plasma concentrations of the drug only reached the concentration of 8 μg/mL in a few moments of the experiment, as can be seen in the concentration versus time curve (Figure 1).

In conclusion, a population PK model capable of simultaneously describing plasma and kidney concentrations of AmB that was generated and estimating the parameters associated with tissue distribution in the rats’ model, demonstrating that there is no statistical difference in drug penetration in the kidneys in healthy and C. albicans-infected groups. In addition, the data from the present study generate the hypothesis that AmB has a nonlinear protein binding. Therefore, infection by C. albicans does not interfere with kidney penetration of AmB, which easily penetrates, and plasma AmB concentrations are considered a good predictor for free kidney concentrations and can be used to optimize AmB regimens to treat disseminated candidiasis considering the drug–protein binding.