Humanized Liver Mouse Models for Predicting Human Drug Metabolism and DDI

Drug development requires preclinical evaluation to determine the safety profile and potential efficacy, which includes understanding how a compound is absorbed, distributed, metabolized, and excreted (ADME). However, even with preclinical studies, around 80~90% of drug candidates will fail to make it to the market.^{1, 2} The top reasons for failure include: 1. toxicity (30%), 2. ineffectiveness (40~50%), and 3. reduced ADME (10~15%).² Animal models are used to predict human ADME and safety profiles however, these studies do not always correlate to clinical findings and often overestimate the potential of drug candidates.^{3, 4} Given this low success rate, it is important to understand the driving factors that lead to drug failure. Here, we will discuss how species-specific differences in metabolism might contribute to the failed clinical success of drug candidates. Then, we will introduce the humanized liver mouse model and other alternative preclinical models that can better predict clinical success.

Species-specific differences in drug metabolism

The vast majority of drugs are primarily metabolized by one of the cytochrome P450 (CYP) enzymes.⁵ It is well known that CYP expression is species-specific, with differences in both the number of isoforms for each subfamily and their relative abundance (Table 1).^6-16 Furthermore, slight differences in the amino acid sequence can affect the catalytic site which controls the metabolism rate of CYPs, with reported differences being >3-fold between mouse and human CYPs.¹⁷ Additionally, nuclear receptors, including pregnane x receptor (PXR), constitutive androstane receptor (CAR), and aryl hydrocarbon receptor (AHR), have species-specific differences in their binding sites.^{18, 19} These nuclear receptors are primarily responsible for CYP induction and can therefore greatly impact drug metabolism through their regulation of drug-metabolizing enzymes.

Table 1: CYP isoform expression comparison between species (Data from Martignoni M 2006, Yamazaki K 2025, Sultana N 2025, Kikuchi R 2010, Cowart LA 2002, Graham RA 2006, Uno Y 2011, and Nebert DW 2013,)^{6, 9, 10,12-16}

To add to the complexity of drug metabolism predictions in pre-clinical models, drug clearance is affected by hepatic transporter expression and liver zonation. Transporters, such as organic anion transporting polypeptides (OATPs) and organic cation transporter (OCT) have species-specific differences in substrate specificity, expression profiles, and homology, which impact human-metabolism predictions.^{20, 21} In regards to the liver anatomy there are clear differences, for example rats lack a gallbladder whereas it is present in humans, mice, monkeys, pigs, and other preclinical models. Additionally, many preclinical models have multiple liver lobes whereas humans do not. It is known that liver zonation plays a role in drug metabolism, since CYP expression is primarily located in the pericentral (zone 3) region for humans.²² Unlike humans, the zonation expression profile for CYP1A2 in rats and pigs have restricted or panlobular expression profiles, respectively. Whereas mice have similar zonation patterns as humans for CYP1A2 and CYP3A4, primarily located in the pericentral region.²² It is important to consider these species-specific differences when selecting preclinical models for ADME studies. Next, we will introduce another preclinical model that has been exceling in predicting human-specific drug metabolism and drug-drug interactions (DDI).

Predict human-specific drug clearance and drug-drug interactions with the PXB-mouse

The PXB-mouse is a humanized liver mouse model that has ~97% human hepatocyte engraftment and expresses similar levels of CYP enzymes, hepatic transporters, and has a similar zonation for CYP2E1 as humans.^{11, 23} It is important to note that while this model excels for the majority of compounds, it might lack the ability to predict immune-mediated drug-induced liver injury given that the model lacks an adaptive immune system. Additionally, the PXB-mouse liver has residual mouse hepatocytes which could impact drug metabolism. Using a SCID mouse control or, for sensitive compounds, selective knock-down of mouse CYPs will help rule out the involvement of murine hepatocytes.^{9, 10} Importantly and despite this, the PXB-mouse has accurately predicted human-specific drug-metabolism for a wide-variety of compounds including those that are cleared by non-CYP-mediated metabolism.

In regards to CYP-mediated drug metabolism the PXB-mouse has been able to accurately predict CYP-mediated clearance and DDI for multiple CYP isoforms. For example, Nakayama et al., demonstrated that the PXB-mouse could be used to predict CYP3A4 DDI.²⁴ Repeat dosing with efavirenz effectively induces CYP3A4 induction in PXB-mice similar to clinical findings (Figure 1).²⁴ Rifampicin also induces CYP3A4, albeit at slightly higher levels than humans (Relative fold-change of CYP3A is 9.3 in PXB-mice compared to 7.8 in human liver).²⁴ 

PXB-mice accurately predicted the human-clearance of several compounds (Figure 2B), including midazolam, alprazolam, and theophylline, unlike the in vitro predictions with primary human hepatocytes (Figure 2A) and SCID mice (Figure 2C).²⁵  Furthermore, PXB-mice can be used to determine which CYP isoforms play a role in the drug metabolism for compounds since they express human-specific CYP isoforms.^{26, 27} 

Figure 1: Induction of CYP3A4 protein by rifampicin (left) and efavirenz (right) in PXB-mice after repeated administration. (Adapted from Nakayama K, et al., 2025)²⁴

Figure 2: Comparison of human clearance (fraction metabolized by CYP3A) in vitro versus in vivo. The dotted line represents y=x. (A) Correlation of the percent change in clearance between humans and PXB-mice (B) or SCID mice (C) was analyzed. White circles represent CL_total (total plasma clearance) and black circles CL_h (hepatic plasma clearance). The linear regression best fit for CL_h is denoted by the blue line. The black dotted line represents y=x and the red dotted lines represent y=x+15. (Modified from Miyake T, et al., 2023)²⁵

There are compounds that are metabolized by non-CYP enzymes such as uridine diphosphate-glucuronosyltransferase (UGT) and alcohol dehydrogenase (ADH)/aldehyde dehydrogenase (ALDH)-mediated metabolism. Importantly, the PXB-mouse has been shown to accurately predict clearance for these compounds as well. In a recent published study, PXB-mice were administered with TMG-123, an alcohol containing compound, to predict which drug-metabolizing enzymes are involved in humans.²⁷ This paper demonstrated that while CYP2B6, CYP2C8, CYP2E1, and CYP3A4/5 contributed to TMG-123 metabolism, ADH/ALDH was the primary metabolizing pathway.²⁷ For compounds metabolized by UGT, PXB-mice accurately predicted human clearance in 100% of the compounds tested (Figure 3B).²⁸

Figure 3: PXB-mice more accurately predict human clearance (within 2-fold error) than in vitro hepatocyte and SCID mice. Correlation of human clearance between hepatocytes (A), PXB-mice (B), and SCID mice (C). Solid line is the nonlinear regression line. The dashed line represents 2-fold and the dotted lines represent 3-fold. (Modified from Miyake T, et al., 2025)²⁸

In fact, the PXB-mouse model outperformed in vitro human hepatocytes (Figure 3A), SCID (Figure 3C), and monkey (the 3-fold margin of error (MOE) was 83% and the 2-fold MOE was 58%) for UGT-, and CYP-mediated clearance of compounds as well as for low clearance compounds (89% of 9 compounds tested were within 2-fold range for the PXB-mouse).^{28, 29, 30} Additionally, this model can accurately predict a drugs half-life. Of 14 compounds tested, ten compounds half-life predictions were within 2-fold error demonstrating the versatility of the PXB-mouse model (Figure 4).³¹

Figure 4: Correlation analysis between half-life predictions and observed half-life in human datasets using PXB-mice (left) and SCID control mice (right). The solid black line is the best-fit line, the area between the dotted lines is within two-fold error and the area between the long-dashed lines is within 3-fold error. (Miyamoto M, et al., 2019)³¹

Taken together, the PXB-mouse is a versatile model capable of accurately predicting human-specific PK and DDI for a wide variety of compounds. Contact our scientific experts to discuss how the PXB-mouse model can enhance your PK/DDI research.

PXB-cells: long-term primary human hepatocytes for in vitro ADME/DMPK

There is a push to greatly reduce animal testing globally, increasing the need for new approach methodologies (NAM) and computational models.^{32, 33} Our fresh and cryopreserved PXB-cells (isolated human hepatocytes from our humanized liver mouse model (PXB-mice)) can be maintained in culture for >30 days and have high expression of drug metabolizing enzymes, transporters, as well as high liver functionality. We recently covered how the PXB-cells can be used in NAMs in more depth, see our blog post PHH drive advances in NAMs. Here we will focus on the versatility of these cells for DMPK.

Fresh PXB-cells express similar levels of CYP enzymes as cryopreserved and primary human hepatocytes (PHH) (Figure 5)³⁴, and generally have a higher expression of Phase I and II enzymes compared to Huh7 and HepG2 cell lines (Figure 6).³⁵ In fact, confluent cultures of PXB-cells express the drug metabolizing enzymes long-term (after 21 days).³⁶ Having high expression levels of CYP enzymes and being able to culture the PXB-cells long-term makes this model ideal for in vitro ADME/DMPK studies.

Figure 5: Comparison of gene expression of CYP isoforms for PXB-cells, hepatocyte-like cells derived from human induced pluripotent stem cells available from three vendors (Vendor A, B, and C), and HepG2 cells. The average expression of 22 primary hepatocyte donor lots is shown as a green line. (Modified from Horiuchi S, et al., 2022)³⁴

Figure 6: PXB-cells express high levels of phase I and II enzymes when cultured with DMSO. Phase I and II drug metabolizing enzyme expression of PXB-cells (HH) cultured in media containing either DMSO, DMSO2 or PBS as well as HepG2, and Huh7 cell lines were assessed at 14 days. (Modified from Sugahara G, et al., 2023)³⁵

PXB-cells are useful for CYP activity assays and can be used in co-culture systems to further enhance the CYP activity of PXB-cells. ^{37, 38} Furthermore they can be used for CYP-mediated toxicity studies. In a recent published study, PXB-cells were treated with a CYP inhibitor, aminobenzotriazole (ABT), prior to aflatoxin B (AFB1) exposure to determine whether CYP activation was required for AFB1 hepatoxicity. AFB1 toxicity was reduced when pan-CYP activities were inhibited (Figure 7).³⁹ Using siRNA knockdown of CYP3A4 demonstrated that this enzyme is involved in AFB1 toxicity, as cell viability was improved with the loss of CYP3A4 compared to control siRNA.³⁹ Taken together, PXB-cells express high levels of CYP long-term and are susceptible to CYP induction making them an ideal in vitro model for predicting drug metabolism and toxicity.

Figure 7: CYP inhibition significantly improves cell viability in aflatoxin-treated PXB-cells. CYP3A activity was measured in PXB-cells 5 days after treatment with aminobenzotriazole (ABT) or negative control (A). Cell viability was assessed for PXB-cells that were exposed to aflatoxin B (AFB1) with or without ABT treatment for 8 days. (B) Statistical significance is denoted as *p<0.05, **p<0.01, n.s. not significant. (Ishida Y, et al., 2020)³⁹

Conclusion

Taken together, the PXB-mouse and PXB-cell models are valuable tools for in vivo and in vitro ADME/DMPK studies. Interested in learning more about our models? We have compiled a list of frequently asked questions below and encourage you to reach out to our scientific team here to discuss these models further!