1. Introduction
Non-alcoholic fatty liver disease (NAFLD) affects approximately 1/4 of the global population, posing a significant threat to global health. Non-alcoholic steatohepatitis (NASH), which progresses from NAFLD, is often characterized by inflammation or fibrosis in conjunction with hepatic steatosis. Patients diagnosed with NASH may further develop liver cirrhosis and hepatocellular carcinoma [
1,
2]. Approximately 20% of individuals with NAFLD progress to NASH, with more than 40% of those subsequently developing fibrosis [
3]. Moreover, NASH is often associated with a variety of metabolic disorders and contributes to a 64% increase in cardiovascular risk [
4,
5]. Consequently, NASH is emerging as one of the major diseases afflicting mankind.
The pathogenesis of NASH remains incompletely understood, with the “two-hit hypothesis” or “multiple-hit hypothesis” being widely recognized as the prevailing models [
6]. The “two-hit hypothesis” posits that liver lipid accumulation due to obesity and insulin resistance is the initial insult, followed by a secondary hit involving inflammatory cytokines, adipokines, mitochondrial dysfunction, and oxidative stress. However, this hypothesis fails to fully capture the complexity of human NAFLD. The “multiple-hit hypothesis” suggests that multiple mechanisms, such as fat accumulation and insulin resistance, inflammatory pathways, dietary factors, genetic predisposition, etc., may work synergistically to drive the progression of the disease. To date, no therapeutic drugs for NASH have been approved for marketing in most countries worldwide. Current development in therapeutic targets for NASH mainly focus on metabolic targets, including PPARs, GLP-1R, GCGR, DPP-IV, SGLT2, FXR, and THR-β [
7]. Among which, FXR is an early-developed and well-studied target [
8]. FXR servers as a major intercellular bile acid (BA) receptor activated in the feeding state to regulate metabolism and inflammation [
9,
10,
11]. The interaction between BAs and intracellular FXR not only regulates BA synthesis but also inhibits hepatic lipogenesis and steatosis, reduces hepatic gluconeogenesis, and enhances peripheral insulin sensitivity by upregulating transcription of GLUT4 [
12,
13]. Due to the crucial role of FXR in the progression of NASH, the development of FXR agonists for NASH treatment has become an important research direction. Numerous drugs targeting FXR are currently advancing through preclinical or clinical stages. Obeticholic acid (OCA) is the most extensively studied FXR agonist, which ameliorates inflammation and fibrosis in patients with NASH [
14]. Despite the promising efficacy of FXR agonists in treating NASH, nearly all of them cause side effects such as pruritus and reduced high-density lipoprotein (HDL-C)/LDL-C ratio [
15]. Some studies have suggested that the side effects of FXR agonists may be associated with target selectivity, particularly for TGR5 [
16], prompting increased interest in the development of FXR-selective agonists in recent years [
8]. However, due to insufficient data from approved drugs in the same class and the potential of experiencing pruritus and hepatotoxicity, there have been significant challenges in dose selection of FXR agonists in first-in-human (FIH) clinical trials. Both FDA and NMPA have issued guidelines for FIH dose design [
17], recommending the use of allometric scaling (AS) and other models for dose prediction and selection. This necessitates careful selection of extrapolation species and allometric scaling models to ensure the scientific rigorousness and reliability of the results [
18].
Meanwhile, FXR is primarily expressed in the intestine and liver [
19], with the liver being the main site of action for FXR agonists. Consequently, it is crucial to determine hepatic drug concentration for a quantitative understanding of the pharmacological effects of FXR agonists, as the plasma concentration of FXR ligands did not reliably predict the potency of the FXR agonists [
20]. Meanwhile, existing biomarkers do not effectively reflect the therapeutic effects of drugs for NASH treatment [
21]. However, directly assessing liver drug concentrations in clinical practice is often invasive and challenging. Therefore, it is necessary to predict these concentrations with the help of modeling. Physiologically based pharmacokinetic (PBPK) models are widely used for predicting drug pharmacokinetics (PK) profiles. In addition, the mechanism-based modeling characteristic of PBPK models allows them to predict drug concentrations in multiple tissues, making them an appropriate tool with which to evaluate the concentrations of FXR agonists in liver.
XZP-5610, a novel non-steroidal FXR agonist developed by Xuanzhu Biopharm, has demonstrated safety and efficacy in the preclinical stage. In this study, the physicochemical properties, ADME characteristics, safety, and efficacy data were determined and employed to predict the corresponding parameters in the human and doses in FIH trails using various AS models. The obtained parameters were then used to construct PBPK models of XZP-5610 in rats and healthy Chinese volunteers. After being validated using preclinical and clinical data, the established models were employed to predict the liver concentration of XZP-5610 in humans. The PBPK model established in this research will provide useful information for the clinical design and development of XZP-5610 and other drugs in this class.
3. Discussion
FXR agonists demonstrated remarkable efficacy in the treatment of NASH disease and are considered one of the most promising potential therapeutic drugs. However, obeticholic acid, one of the fastest-progressing drugs, was rejected by the FDA due to severe side effects such as pruritus, gastrointestinal, and hepatic adverse events. Although non-steroidal and highly selective FXR agonists are believed to potentially avoid these adverse events, sufficient clinical evidence is still lacking. Therefore, the rational selection of the dose in a FIH trial is crucial for protecting subjects and expediting the drug development process.
In this study, various methods were used to predict the dose for the FIH trial. Due to insufficient clinical data, a safety factor of 10 was applied in this study. Appropriate selection of extrapolation species and modeling methods is crucial for the FIH dose prediction. Generally, animals with metabolic properties close to humans are preferred as extrapolation species. When utilizing parameters from different species for human dose prediction, two assumptions should be followed to ensure the reliability of the results: (a) there were no species differences in the PK characteristics of XZP-5610 among SD rats, beagle dogs, and humans; in addition, the PK characteristics were linear within the studied dose range, and (b) XZP-5610 produced similar pharmacological activity or toxicity at the same systemic exposure in both animals and humans. For the first assumption, the results of PK experiments in rats and dogs demonstrated linear correlations of both C
max and AUC within the studied dose range. Additionally, metabolite profiles indicate similarities in the type and degree of metabolites across these species. Furthermore, none of these metabolites were substrates for common transporters (predicted). Therefore, it was inferred that XZP-5610 exhibited linear PK characteristics within the studied dose range, without observed species differences. Regarding the second assumption, despite species differences, FXR demonstrated similar distribution characteristics and function [
22,
23], suggesting that it could produce similar activity or toxicity at the equivalent exposure. As a result, a relatively accurate assessment and prediction of exposure and PK parameters of XZP-5610 in humans could be achieved by the established PBPK model. The clinical data further supported the reliability of the methods and results.
To better predict PK parameters in humans, multiple extrapolation models were selected for predicting the clearance and apparent volume of distribution, which integrated features such as single-species, multi-species, and corrections for free fraction and hepatic blood flow. In comparison to the previous literature [
24,
25], the model established in this study reduced random errors by introducing a limited number of parameters, thus improving prediction accuracy and providing a more reliable analytical performance. However, a significant difference in predicted human CL
i.v was observed when using data obtained from two species (SD rats and beagle dogs). Compared to beagle dog, the CL
i.v predicted with rat data was substantially higher (~2 fold). The difference may be attributed to the stronger metabolic capacity of XZP-5610 in rats, as inferred from the results of hepatocyte stability. In comparison to rats, the results from beagle dog and human hepatocytes were more similar; thus, the results predicted with beagle dog were ultimately adopted as the CL
i.v in humans. Similarly, the average bioavailability in beagle dog was used as a typical value for predicting human bioavailability. For V
ss, the predicted values obtained from data in rats and dogs were similar, suggesting the V
ss of XZP-5610 was consistent across different species. Meanwhile, the Øie–Tozer method can be utilized to correct the distribution of drugs in intracellular and intercellular fluids. Finally, the average predicted values derived from different methods and across both species were used as the predicted V
ss for humans. For K
a, as the variation in physiological factors influencing gastrointestinal absorption results in a weak correlation between K
a values in animals and humans, the average value obtained from various species was employed as the representative predicted human K
a, with the highest and lowest values used as the range of human K
a, respectively.
In rat PK and toxicological experiments, sex differences were observed in both i.v. and oral administration, suggesting that these differences may not be related to absorption. It was established that CYP3A served as the primary metabolic enzyme in the metabolism of XZP-5610 (
Figure S1), and the results of metabolites analysis showed that the proportion of parent drug in plasma of female and male rats was 31.33% and 1.75%, respectively. Significant sex differences in some CYP3a subtypes in rats have been reported [
26], and for some compounds, sex differences in PK parameters have been correlated with metabolic enzymes [
27]. Therefore, the sex differences observed in this study may be a result of differences in metabolic enzymes. In humans, the sex differences in CYP2C9 and CYP3A activity were small [
28], which indicates that the possibility of sex differences in human PK profiles is limited. However, statistical analysis of relevant data during clinical trials is still recommended.
Tissue distribution results indicated the liver was the main target organ of XZP-5610, with a liver-to-plasma ratio up to 12.9. Although XZP-5610 was not a substrate for some known liver uptake transporters such as OATP1B1 and OATP1B3, the liver-to-plasma ratio suggested that some other transporters may be involved in the hepatic uptake of XZP-5610. The increased liver concentration was beneficial for its therapeutic effect, but also an important cause of toxicity. In this study, a rat PBPK model was first established and used to predict liver concentrations. A good match between the predicted and measured values in radiolabeled experiment indicated that the model was appropriate for predicting liver concentrations in rats. The model was further extrapolated to humans using in vivo and in vitro experimental results and applied to predict human liver drug concentrations. Combined with the preclinical toxicology results, this model could provide valuable support for the design of clinical trials of XZP-5610. Furthermore, it is worth noting that the tissue distribution studies indicated a Tmax of 2 h for XZP-5610 in the intestines, which was much longer than in other tissues. In fact, further excretion studies have revealed that more than 90% of the administered dose of XZP-5610 was excreted in bile, suggesting that enterohepatic circulation may be the primary reason for the extended Tmax of XZP-5610.
The result suggested that the predicted PK parameters using the PBPK model, established with data from Chinese healthy individuals, aligned well with the clinical data, only with the exception for the 1 mg and 2 mg doses. At these doses, clinical results indicated that the increments in Cmax and AUC were comparatively lower than those of dose escalation. Based on results from radioactive isotope studies in rats, the predominant distribution of XZP-5610 occurred in the gastrointestinal tract and liver, implying a potential contribution of diverse absorption mechanisms in the intestines, including passive diffusion and uptake/efflux transport, for XZP-5610. Evidence for the involvement of transporters could be verified from the Caco-2 experiments, where increased Papp was observed with escalating XZP-5610 concentrations, although XZP-5610 has already been proved not to be a substrate for P-gp and BCRP. To precisely elucidate these mechanisms, further experimental validations are required in subsequent studies.
NOAEL and MABEL doses in animals can assist in calculating the recommended initial dose and effective dose in humans. For the dose selection of the FIH trial, the initial dose of 0.15 mg and maximum dose of 3 mg were used in the SAD trial. According to the extrapolation from animal experiments, the equivalent effective doses in humans ranged from 2.18 mg to 4.80 mg, with most of the predicted MTD at about 3~14 mg. As a result, adverse reactions may occur at a dose of about 3 mg in humans, which is close to the upper limit of the effective dose. As discussed above, XZP-5610 accumulated in the liver, the primary pharmacological and target organ. It was unlikely to obtain a better risk–benefit ratio with a higher dose. Therefore, the maximum dose was set at 3 mg in consideration of safety issues. Fortunately, in preclinical pharmacological experiments, beneficial pathological and blood biochemical changes were observed at doses lower than the MABEL dose. This indicates that setting the efficacy dose to a lower dosage level (~2 mg) is reasonable. The design of the series doses can provide useful information for exploring the effective dose in humans.
Certainly, there are still limitations in our study. Due to the unavailability of clinical samples, the predicted hepatic concentrations by the PBPK model were not validated with clinical data. Additionally, the liver-specific enrichment implies the potential contribution of certain transporters, but unfortunately, XZP-5610 was not a substrate such as OATP1B1/3 in the in vitro test. To address these issues, tissue biopsy samples can be utilized in the subsequent clinical trials to measure the hepatic concentrations of XZP-5610. Meanwhile, more in vitro experiments are needed for the verification of transporters. Nonetheless, our research indicated that the models established in this study can effectively predict the FIH doses of XZP-5610 and its distribution in the plasma and tissues of healthy Chinese adults.