Niclosamide (Nc) is an FDA-approved anti-tapeworm drug for use in humans to treat tapeworm infections. It is also included on the World Health Organization’s list of essential medicines. Nc exerts its anti-tapeworm effect by inhibiting oxidative phosphorylation and stimulating adenosine triphosphatase activity in the mitochondria (Jeon et al., 2020; Xu et al., 2020; Shamim et al., 2021). A recent drug repurposing screening has suggested that Nc possesses potent antiviral activity against viruses including Zika virus, MERS-CoV, SARS-CoV, SARS-CoV-2, hepatitis C virus, and human adenovirus (Jeon et al., 2020; Xu et al., 2020; Shamim et al., 2021). It was recently demonstrated that Nc inhibits SKP2 activity, which enhances autophagy and reduces MERS-CoV replication, and a similar mechanism is expected to contribute to the inhibition of SARS-CoV-2 infection (Jeon et al., 2020). Nc exhibits very potent antiviral activity against SARS-CoV-2 with IC₅₀ ranges from 0.06–0.28 μM (Weiss et al., 2021).

Due to its low aqueous solubility, low permeability, and extensive liver and intestine metabolism, Nc has very low oral bioavailability. The solubility of Nc is reported to range from 0.2 to 19 μg/mL, depending on the temperature and pH conditions as well as on salt contents (Andrews et al., 1982). The solubility of the ethanolamine salt of Nc can be improved up to 372 μg/mL (Andrews et al., 1982). Nc also displays poor permeability with the parallel artificial membrane permeability assay (PAMPA) permeability of less than 1 × 10⁻⁶ cm/s (Shamim et al., 2021). Pharmacokinetic (PK) study in rats that received a single oral (PO) dose of 50 mg/kg ¹⁴C-radiolabeled Nc indicated that about one-third of the dose was absorbed while two-thirds of the dose was excreted in feces (Andrews et al., 1982). The oral bioavailability of Nc in rats and dogs is only 5.5% and 0.54% respectively (Choi et al., 2021).

Nc has been investigated in multiple clinical trials for colorectal cancer, prostrate carcinoma, and COVID-19. In an early human clinical study, male and female volunteers received a single oral dose of 2000 mg ¹⁴C-labeled Nc. The fraction of ¹⁴C-activity eliminated with the urine was 2–25% over a 4-day period. The remainder was eliminated in feces. Serum maximal concentration (C_max) equivalent to 0.25–6.0 μg/mL Nc was detected (Andrews et al., 1982). A recent Phase I dose-escalation study reported that oral dose of Nc could not be escalated above 500 mg three times a day (TID). The maximum plasma concentrations could not achieve the threshold to inhibit cell growth in castration-resistant prostate cancer models (Schweizer et al., 2018). A prospective phase II clinical trial of Nc in patients with metastasized colorectal cancer indicated that an oral dose of 2 g once a day (QD) per patient could produce a median plasma C_max of 0.665 μg/mL (Burock et al., 2018), which is still modest for anticancer treatment. In a recently published phase Ib clinical trial of reformulated Nc, an oral dose of 1.2 g TID Nc could achieve a mean peak plasma concentration of 0.159 μg/mL in three male prostate cancer patients (Parikh et al., 2021). Multiple clinical trials for the treatment of COVID-19 and cancer with Nc are still ongoing.

The low solubility, poor bioavailability, and poor plasma exposure of Nc hinder its use in patient treatments (Cairns et al., 2022). Previous efforts have been made to increase its systemic exposure by optimizing injectable formulation for IV administration (Ye et al., 2015) or to increase its local concentrations at the lung by developing inhalation or intranasal formulations (Backer et al., 2021; Jara et al., 2021). However, oral therapeutics have the potential to maximize patient benefit due to their non-invasiveness, patient compliance, and convenience. It is still the first choice of route of administration, especially during a pandemic. In order to achieve high exposures after oral delivery, ester-derivative prodrugs were employed. Few attempts have been made in this approach. An Nc stearate prodrug has been developed for increasing its solubility but was only evaluated for IV administration (Reddy et al., 2020). In another study, five prodrugs of Nc were synthesized, and the in vitro and in vivo efficacy was evaluated (Elkihel et al., 1994). However, the pharmacokinetic profiles were not determined for these prodrugs. In this study, we evaluated in vitro ADME properties and pharmacokinetics (PK) of an Nc prodrug, namely, NCATS-SM4705, in lab animals. The structures of Nc and its prodrug NCATS-SM4705 (PDN) are shown in Figure 1. PDN showed in vitro activity against SARS-CoV-2 in our previous cytopathic effect assay and cytotoxicity counter screen; and in vivo efficacy was tested in a hamster model of SARS-CoV2 infection (Huang et al., 2022).

To predict human pharmacokinetic profiles, modeling methodologies integrating both in vitro absorption, distribution, metabolism, and excretion (ADME) properties and preclinical pharmacokinetic data were applied. Physiologically based pharmacokinetic (PBPK) modeling combines drug information and animal physiology at organ levels and provides mechanistic representations of drug concentrations in biological systems (Bi et al., 2016; Kuepfer et al., 2016). This system pharmacology approach can not only predict the pharmacokinetic profiles of a new chemical entity in various tissues but also analyze the in vivo ADME processes with advanced scientific knowledge at the mechanistic level, supporting further drug candidate optimization and development, and guiding dose selection in clinical trials.

The current study aims to evaluate a novel prodrug of Nc, NCATS-SM4705 (PDN), on improving the in vivo exposure of Nc. In vitro ADME properties and PK of PDN were studied in mice and hamsters. Based on in silico, in vitro and in vivo mouse data, a whole-body PBPK model was developed to characterize the concentration-time profiles of PDN and Nc in systemic circulation and different tissues. The PBPK model was verified using hamster PK profiles and then extrapolated to predict the concentration-time profiles in humans. The simulation was performed at different dose regimens for first-in-human PK prediction and future drug development guidance.

PBPK modeling has been well established for in vitro to in vivo translation and pharmacokinetic modeling during drug discovery in preclinical species (Bi et al., 2016; Zake et al., 2021) as well as in clinical trials (Lukacova et al., 2016; Miller et al., 2019). With the PBPK model, the dynamics of drug distribution in various tissues can be evaluated and predicted, thus leading to a better understanding of the relationship between target tissue exposure and drug efficacy. In addition, PBPK model is a useful tool to analyze and simulate the ADME process at the organ level by incorporating physiology parameters and mechanisms determined in the in vitro model. The use of the PBPK model for human pharmacokinetic predictions is growing and has been considered by regulatory agencies as beneficial to mitigate the uncertainty of a new chemical entity before clinical trials (Wagner et al., 2015; Miller et al., 2019). The present study evaluated a novel prodrug of Nc (PDN) in improving the in vivo exposure of Nc in mice and hamsters, and featured a PBPK model that was able to adequately characterize the plasma and organ concentration-time profiles of PDN and Nc. The PBPK model was then applied to simulate the pharmacokinetics of PDN and Nc in humans, supporting the evaluation of its potential clinical use as well as guiding future development strategies for other Nc prodrugs.

We conducted PK experiments in mice and hamsters, the species commonly used for COVID-19 disease model, through IV and PO administration routes and various dosing regimens. We observed a linear pharmacokinetics profile of PDN and formed Nc within the doses studied in both species. A moderate and high CL_p in mice and hamsters, and a suitable t_1/2 were observed for PDN. We also observed extensive tissue distribution and high Vd_ss of PDN in both species. PDN was preferably distributed to the highly perfused organs such as the lung, spleen, and kidney, with their AUC in these tissues 8–44 folds higher than in plasma. In contrast, Nc has been reported to have very limited tissue distribution in previous studies (Choi et al., 2021; Shamim et al., 2021). Nc presented high plasma protein binding (>99.8%) and was confined mainly in the plasma with limited tissue distribution after intravenous (IV) injection (Choi et al., 2021; Shamim et al., 2021). The volume of distribution for Nc has been reported to be less than 0.4 L/kg in rats (Choi et al., 2021). After an IV dose of Nc (2 mg/kg) in rats, the concentration of Nc was much higher in the intestine followed by the kidney and liver compared to other tissues including the spleen, lung, and heart (Ye et al., 2015). The tissue concentrations of Nc in mice after oral administration of Nc at 40 mg/kg were also high in the kidney and liver, lower in the lung and heart, and very low in the brain (Tao et al., 2014). Consistent with these previous reports, in our study in mice, Nc showed Kp around or less than 1 in most of the tissues, while much higher partitions were found in intestine, kidney, and liver. In the current study, the dosed PDN was found efficiently converted to Nc in both preclinical species tested, improving the systemic exposure of Nc in mice after oral administration compared to the previously reported concentrations after oral administration of Nc. In our study, after a PO dose of 10 mg/kg PDN, the lung to plasma AUC_0-last ratio of Nc was about 0.97, which was 3.6-fold higher than 0.27 as reported after a PO dose of Nc (40 mg/kg) (Fan et al., 2019), and the dose-normalized plasma AUC_0-∞ of Nc was more than 8-fold higher than that reported after a PO dose of Nc (40 mg/kg) (Fan et al., 2019).

To establish the PBPK model, rather than fully relying on built-in quantitative structure-property relationship data, we obtained critical experimental ADME and physicochemical data. We observed that PDN showed high permeability in artificial lipid membranes in PAMPA measurements. Considering the reported high protein binding of Nc and rat urine pharmacokinetics (Choi et al., 2021; Shamim et al., 2021), Nc is not likely to be excreted via urine in its unchanged form, and metabolism would likely be the major elimination route for both PDN and Nc. Therefore, the metabolism of both compounds was studied in liver microsome and plasma to provide mechanistic knowledge for PBPK model building. Metabolism of Nc has been extensively investigated in several previous in vitro and in vivo studies in mice and humans but has never been investigated in hamsters. In an early phase I study, after a single oral dose of 2000 mg ¹⁴C-labeled Nc in humans, metabolites were excreted in the form of glucuronides as well as unchanged Nc, 2′,5-dichloro-4′-aminosalicylanilide, and 2′,5-dichloro-4′-acetaminosalicylanilide (Andrews et al., 1982). Recent research also demonstrated that CYPs-mediated biotransformation was not important for the disposition of orally administered Nc, whereas glucuronidation mediated by UGT enzymes significantly contributed to the disposition of Nc (Lu et al., 2016; Fan et al., 2019). The UGT-mediated hepatic CL _{in vitro,app} of Nc was reported at 0.18–0.21 mL/min/mg microsomes in mice with apparent K_m value determined at 0.09–1.67 μM (Lu et al., 2016; Fan et al., 2019). The mouse hepatic CL _{in vitro,app} of Nc in liver microsome determined in our study is 0.075 mL/min/mg microsome, lower than the previously reported values. The discrepancy might be from the different microsomal protein concentrations, Nc concentrations, and different incubation periods. The hepatic CL _{in vitro,app} of Nc in human microsome determined in our study is similar to that in mouse microsome, while the CL _{in vitro,app} of Nc in human liver microsome was found about ten-fold higher than that in mouse microsomes in a previous study (Lu et al., 2016). Possible reasons for the discrepancy would be that the microsomal protein concentration used for human microsomes is much lower than that for mouse microsomes in the previous study. For compounds with high protein binding, such as Nc, varied microsomal protein concentrations may alter its unbound concentration in the incubation system, resulting in varied apparent CL _{in vitro,app} (Obach, 1999). Unlike the previous study, our in vitro system used the same microsome protein concentration for different species and could give a better cross-species prediction for the hepatic CL _{in vitro,app} of Nc. We also determined the hamster hepatic CL _{in vitro,app} of Nc, which was found slightly higher than the mouse hepatic CL _{in vitro,app} of Nc. Consistent with the reported stability of Nc towards human and rat plasma carboxylesterases (Bradshaw et al., 2018), no depletion of Nc was found in plasma from all three species tested in our study. Biotransformation from PDN to Nc was also explored. PDN showed fast biotransformation to Nc in the mouse and hamster plasma, but a much lower rate in the human plasma. In mouse and human liver microsomes, only a small proportion of PDN was converted to Nc, while slightly higher proportion of PDN was converted to Nc in hamster liver microsomes. Since a distinct concentration-time curve of Nc was observed in the liver compared to other tissues, the saturable metabolic kinetics of PDN was estimated in liver microsomes and embedded in the PBPK model. Biotransformation of PDN in liver cytosolic fraction was also tested but found neglectable. The plasma protein binding of PDN in rodent plasma could not be experimentally determined due to its instability after 1 h incubation at 37°C and was estimated by an in silico model. Since the Fu_p determined in human plasma experimentally was very close to that predicted by the in silico model (15.2% vs. 15.8%), the predicted rodent Fu_p was adapted for the PBPK modeling.

In vitro-in vivo extrapolation was applied to fit our in vitro data to the PBPK model for the simulation of in vivo pharmacokinetic profiles in mice. It has been widely reported that the liver microsomal incubation system tends to underpredict hepatic in vivo CL_intrinsic, especially for drugs with high plasma protein binding such as Nc or high in vivo CL_intrinsic such as PDN. The underprediction could be attributed to factors such as the in vitro non-specific binding and the lability of enzymes and transporters during preparation and storage processes. Application of the empirical scaling factor (ESF) was reported to successfully correct for such discrepancy (Obach, 1999; Wood et al., 2017; Hallifax and Houston, 2019). The estimated ESF also reflects the contribution of each biotransformation process at the organism level and facilitates more specific and accurate cross-species extrapolation based on the in vitro findings. It is worth noting that, Nc undergoes first-pass metabolism by the intestine and liver. In vitro stability of Nc showed that Nc-glucuronide formation in intestinal microsomes was about 10-fold greater than that in the liver (Fan et al., 2019). A recent report of in vitro assays identified UGT1A1 as the most active human enzyme for the glucuronidation of Nc (Lu et al., 2016), which is highly expressed in both liver and small intestine (Uhlen et al., 2015; Proteinatlas, 2021). Similarly, hydrolysis enzymes such as carboxylesterases and paraoxonase that may contribute to the biotransformation from PDN to Nc in the liver are also expressed in the intestine (Yang et al., 2011). Therefore, in our PBPK model, intestinal clearance of Nc and intestinal biotransformation from PDN to Nc were incorporated. In our study, we did not conduct in vitro metabolism study in intestinal microsomes because hamster intestinal microsomes were not available. In this case, following the common practice to estimate in vivo intestinal metabolic rates, the metabolic kinetic parameters determined in liver microsomes were scaled to the in vivo intestinal biotransformation rates in our PBPK model. In vitro-in vivo extrapolation was applied to estimate the absorption rate of PDN in three species by using the built-in correlation of GastroPlus. The high PAMPA permeability of PDN projected a complete absorption (>99%) into enterocytes in mice, hamsters, and humans.

The established PBPK model was then refined and verified using concentration-time profiles in various tissues and plasma from two preclinical species. The simulated concentration profiles in various tissues and generated pharmacokinetic parameters in mice are within the acceptance criterion (2-fold error range of observed data). The PBPK model optimized for mice was also able to provide an acceptable prediction of plasma and lung concentrations in the hamsters after PO administration of PDN, providing predictability for the simulation results in humans. The simulated absorption of PDN in humans indicated a complete but slightly slower oral absorption compared to hamsters and mice. Low clearance and high volume of distribution leading to a longer t_1/2 of PDN in humans were predicted. The formation of Nc was also predicted to be much slower than that in hamsters and mice. After the oral dose of PDN, the overall Nc to PDN ratio predicted in humans (AUC_0-∞ ratio Nc/PDN of 1.3 after PO 150 mg for a human subject or 2.3 mg/kg dose) was higher than those observed in hamsters (0.44 after PO 30 mg/kg dose) and mice (0.46 after PO 10 mg/kg), possibly due to the slower elimination of Nc in human. Accordingly, after the oral dose of PDN, the t_1/2 predicted for Nc was also longer in humans (6.3 h) than those observed in rodents (3.1 h in mice and 1.9 in hamsters). In our simulations for humans, after the TID 150 mg doses of PDN, the steady-state plasma concentrations of Nc were at 130–210 ng/mL. Such PK profile is comparable to those observed in phase I clinical trial of Nc in prostate cancer patients, where the plasma C_max after TID 500 mg oral Nc dose ranged from 35.7 to 182 ng/mL (Schweizer et al., 2018). At a higher dose of TID 600 mg PDN, the simulated steady-state plasma concentrations of Nc were 0.56–0.85 μg/mL, higher than the observed Nc concentrations from a phase II clinical trial where median plasma C_max of 0.665 μg/mL was achieved after 2 g QD oral Nc dose (Burock et al., 2018). Our model was also able to predict the lung concentrations of Nc in humans. After the TID 600 mg doses of PDN, steady-state lung concentrations of Nc were 1.46–2.40 μg/g tissue, which are 16–120 folds higher than the IC₅₀ of Nc in the in vitro model (20–91 ng/mL) against SARS-CoV-2 (Weiss et al., 2021).

The sensitivity and uncertainty of the model were also systematically assessed. In the human PBPK model, the biotransformation rate from PDN to Nc in plasma is the most important influencer, while the PStc value of PDN to the liver is a second important influencer for Nc plasma and lung exposure. In contrast, the biotransformation rate from PDN to Nc in the liver has little effect on altering the Nc systemic exposure. This could be due to the low membrane permeability of Nc manifested by the significantly prolonged elimination of Nc in mouse liver. The PStc value of Nc in the liver is also a significant influencer for Nc plasma exposure. Based on this analysis, future development of Nc prodrugs should put more emphasis on the plasma biotransformation rate. Fu_p of Nc also has some impact on the plasma and lung concentration profiles of Nc. For high protein binding compounds like Nc, the experimental variation of the unbound fraction value could be very large due to the limited detection sensitivity. Considering such large variations on Fu_p of Nc, Fu_p could be one of the reasons for the uncertainty of the current PBPK model.

A limitation of the current model is the mixed metabolic kinetics. Linear biotransformation rate of PDN was assumed in plasma. In our preclinical studies, CL_p of PDN is consistent, and plasma and tissue AUC_0-∞ of PDN and Nc are proportional to the dose. Linear pharmacokinetics of Nc and PDN was observed when the plasma concentrations are below 623 and 1063 ng/mL for Nc and PDN, and liver concentrations are below 822 and 240 ng/g for Nc and PDN. The current PBPK model may overestimate the elimination of PDN or biotransformation from PDN to Nc in plasma and give unreliable predictions when the dose increases, resulting in higher plasma or liver concentrations. Variability in biotransformation may also come from individual differences in the expression or function of related enzymes. Therefore, further studies on identifying the metabolic enzymes involved are warranted to improve the model prediction accuracy in different situations.

Due to its inherently low bioavailability and low systemic exposure, orally dosed Nc is well-tolerated. The maximum approved oral dose of Nc in its anthelmintic application is 2 g QD (Burock et al., 2018). The acute toxicity test in rats showed its oral LD₅₀ > 5000 mg/kg (Andrews et al., 1982). Dose-limiting toxicities of Nc are mainly gastro-intestinal tract related, such as nausea, vomiting, diarrhea, and colitis (Schweizer et al., 2018). However, the maximum tolerated dose of Nc has never been determined by IV administration. By applying the prodrug strategies, much higher amounts of Nc were delivered into the systemic circulation. Based on our current finding, high exposure levels of Nc were observed in the liver, intestine and kidney. The potential risk of high tissue exposure levels of Nc after the intake of prodrug needs to be determined in future toxicity and safety studies.

In this study, we evaluated in vitro ADME properties and pharmacokinetics of a novel prodrug of Nc (PDN; NCATS-SM4705), which was designed to improve the in vivo exposure of Nc. PDN was efficiently converted to Nc in both liver and blood, resulting in an improved systemic exposure of Nc in mice and hamsters after oral administration. A PBPK model was developed to adequately simulated the pharmacokinetic and tissue distribution profiles of PDN and the formed Nc in mouse and hamster, and the analysis of this model indicated that the plasma and lung concentration of Nc is most sensitive to the permeability of PDN and Nc across the hepatocyte membranes, as well as the biotransformation rate from prodrug to Nc in plasma. Our predicted Nc concentrations in human plasma and lung suggest that a TID dose of 300 mg PDN would provide a lung concentration of Nc 8- to 60-fold higher than the IC₅₀ against SARS-CoV-2 reported from in vitro cell assays.