The field of RNA therapeutics, with its potential for treating a wide range of diseases, continues to experience rapid growth and attracts significant investment. According to the American Society of Gene & Cell Therapy (ASGCT), as of Q1 2025, 35 RNA therapies have been approved globally and another 1,298 are currently in development (between preclinical and pre-registration stages) [1].
RNA therapeutics encapsulate several therapeutic modalities, including small interfering RNA (siRNA), messenger RNA (mRNA), and antisense oligonucleotides (ASO) (Table 1). IQVIA Pipeline Intelligence [2] noted that these three types dominate the development landscape, collectively representing 80% of the RNA therapeutics pipeline. Among these, siRNA is the most prevalent category from preclinical to pre-registration stages accounting for 37% of the pipeline. The range of therapeutic indications is very broad, with rare diseases topping the list of targets (Figure 1).
Despite the promising number of therapies in development, the path from laboratory to clinic is fraught with challenges, particularly in translating preclinical findings to human outcomes. Traditional animal models, while valuable, are often inadequate in predicting human responses due to species-specific differences in physiology, metabolism, and disease manifestation. This fundamental disconnect frequently leads to late-stage failures in clinical trials, leading to significant time and resource expenditures. Clinical failures can be attributed to a combination of reasons including a lack of clinical efficacy or toxicity (70-80%), poor drug properties (10-15%) or commercial reasons (10%) [3]. For RNA therapeutics in particular, delivery of the drug to target organs/tissues and toxicity due to off-target binding are some of the most significant challenges [4].
Humanized liver chimeric models have been developed to address the translational challenges. Such models provide a more human-relevant environment for the preclinical testing of RNA therapeutics. The PXB-mouse is created using transgenic mice that allow for the ablation of endogenous mouse hepatocytes and accept the engraftment of xenotransplanted human hepatocytes [5,6,7]. This process results in a mouse with up to 95% of its liver repopulated with functional human hepatocytes. The outcome is a chimeric model that combines the benefits of a small animal model with the biological relevance of human liver tissue.
Features of the PXB-mouse include:
There are many reasons for the clinical failure of RNA therapeutics, including off-target effects, delivery challenges, target engagement challenges and low therapeutic efficacy. Despite these obstacles, humanized liver mouse models have consistently proven effective in recapitulating human outcomes in RNA therapeutic development. Here we examine four examples from a wide range of published literature that showcase the value of humanized liver mouse models in this field.
Chronic hepatitis B virus (cHBV) infection remains a significant global health challenge; the WHO estimated 254 million people were living with the condition in 2022 plus 1.2 million new infections each year [18]. RNA interference (RNAi) therapeutics have shown promise in targeting cHBV, but safety concerns have hindered their development.
HBV is associated with the expression of various proteins including hepatitis B surface antigen (HBsAg). It is hypothesized that large quantities of HBsAg contribute to T- and B-cell dysfunction, impairing the host’s ability to eradicate the HBV infection. A potential treatment involves reducing HBsAg using RNA interference via siRNA. Since there are overlapping templates within the X region of the HBV genome, a single siRNA could selectively and effectively target all HBV transcripts [19].
In a recent study, researchers used the PXB-mouse as a preclinical model to accurately predict the safety and tolerability of investigational RNAi therapeutics in healthy volunteers [19]. The study compared two siRNAs that target all major HBV mRNA transcripts: ALN-HBV and VIR-2218. These siRNAs have the same sequences, except that VIR-2218 has been chemically modified via enhanced stabilization chemistry plus (ESC+) resulting in a single substitution of a glycol nucleic acid modification within the seed region. This modification was an attempt to minimize off-target effects.
PXB-mice (12-18 weeks of age) received subcutaneous injections of ALN-HBV or VIR-2218 at doses of 12, 36, or 100 mg per kg of animal weight. Blood analysis over seven weeks revealed markedly lower alanine aminotransferase (ALT) levels following administration of VIR-2218 compared to ALN-HBV, indicating reduced liver damage (Figure 2A).
Importantly, these preclinical findings were supported by subsequent clinical studies in healthy volunteers as well as those with chronic HBV (Figure 2B-D). Here, the PXB-mouse model accurately predicted the improved safety profile of VIR-2218, demonstrating its value in assessing potential off-target effects and safety concerns early in the development process.
Effective delivery of RNAi therapeutics is as crucial as their sequence design (Figure 2). Multiple options are available, but lipid nanoparticles (LNPs) have emerged as a popular choice for siRNA delivery due to their ability to target various tissues while protecting the siRNA from degradation. Most recently, the spotlight was on LNPs as a key component of the COVID-19 vaccines [20,21]. Despite these advancements, issues in ensuring RNAi uptake by the correct tissue and achieving cross-species compatibility can complicate preclinical testing. This is where humanized liver mouse models offer a distinct advantage, helping to bridge these gaps by closely mimicking human responses to RNA therapeutics.
In a study by Okada et al, an LNP encapsulated siRNA was effectively used to target Dock11, a host factor regulating covalently closed circular DNA (cccDNA) formation by HBV in PXB-mice [22]. cccDNA forms during HBV replication and acts as a viral reservoir in cells. The persistence of cccDNA and the inability to effectively target it with therapeutics is a key reason that a cure for HBV remains elusive.
In PXB-mice, the LNP-encapsulated siRNA targeting DOCK11 showed highly effective knockdown of human DOCK11 in PXB-mice and, importantly, a clear reduction in cccDNA levels (Figure 4). The study employed the same LNP formulation as that used in the FDA-approved drug Onpattro®, highlighting the cross-compatibility of human-tested LNPs in the PXB-mouse.
Another example of the effective use of LNPs in PXB-mice to predict human outcomes comes from a hepatitis delta virus (HDV) study [17]. HDV infects an estimated 10−20 million people globally and is associated with severe fulminant hepatitis, which often leads to cirrhosis and an increased risk of hepatocellular carcinoma. Despite the severity of the disease, there is an unmet clinical need for effective treatments.
HDV infection requires the presence of HBsAg; HDV can either establish itself as a superinfection in individuals already carrying HBV or through a simultaneous coinfection when a person is exposed to both HBV and HDV at the same time.
In this study, researchers used humanized mice dually infected with both HBV and HDV to evaluate the effectiveness of HBV-targeting siRNA therapy in controlling HDV infection, comparing it to a direct anti-HDV siRNA approach.
The results revealed that in vivo treatment with an anti-HBV RNAi agent successfully reduced both HBV and HDV viremia, showing the potential of this approach in managing HDV infection.
Specifically, treatment with ARB-1740, delivered via LNP technology, resulted in a 2.3 log10 reduction in HBV viremia and a 2.6 log10 decrease in serum HBsAg levels, which led to a subsequent 1.6 log10 reduction in HDV viremia (Figure 5 A, B, C, D). In contrast, HDV-targeting siRNA effectively inhibited HDV in both the blood and liver compartments without impacting HBV (Figure 5D, E). Additionally, PEGylated interferon-alpha reduced HBV viremia by 2.0 log10 but did not affect HDV viremia under the conditions of this study (Figure 5). These findings demonstrate the inhibitory effect of ARB-1740 on HDV, supporting its potential as a therapeutic option.
Note that, as anticipated by the investigators, the human chimeric mouse model showed no overt signs of liver damage (including cirrhosis or hepatocellular carcinoma) in either of the coinfection or superinfection studies described, mostly likely due to the lack of an adaptive immune response in this system.
Overall, these studies emphasize the value of the humanized liver mouse model in generating translationally relevant results, which can guide the selection of effective siRNA delivery methods.
Metabolic dysfunction-associated steatohepatitis (MASH), formerly known as non-alcoholic steatohepatitis (NASH), is becoming the most common cause of liver disease. To date, therapies that have shown promise in mouse MASH models have not translated well to humans, underscoring the need for more predictive preclinical models [23,24].
It has been shown that MASH can be established in PXB-mice by feeding them a high fat diet [25]. PXB-mice fed these diets recapitulate the key features of human metabolic dysfunction-associated fatty liver disease (MAFLD)/MASH including hepatocyte ballooning, inflammation, and importantly, fibrosis.
Researchers used MASH diet-fed PXB-mice to test GalNAc-siTaz, an siRNA targeting the gene for TAZ, a transcriptional regulator [26]. The researchers chose this model reasoning that "a mouse NASH model whose livers are populated with human hepatocytes would be particularly valuable in testing hepatocyte-targeted siRNA therapies." The PXB-mice were fed a high-fat, choline-deficient, L-amino acid-defined diet for six weeks to induce MASH. This was followed by six weekly injections of GalNAc-siTAZ or a control siRNA (GalNAc-control) while maintaining the MASH diet. The results were promising. GalNAc-siTAZ lowered human hepatic TAZ and IHH (Indian hedgehog) a TAZ target that promotes MASH fibrosis (Figure 6B). In addition, treatment with GalNAc-siTAZ decreased liver inflammation, hepatocellular injury, hepatic fibrosis, and profibrogenic mediator expression compared to the control (Figure 6 C-F). These effects indicated that GalNAc-siTAZ decreased the progression of MASH in mice reconstituted with human hepatocytes.
This study demonstrates the value of humanized liver models in assessing both target engagement and therapeutic efficacy for complex metabolic liver diseases like MASH.
Hypertriglyceridemia, characterized by elevated triglyceride levels, is associated with an increased risk of cardiovascular diseases. In cases of severe hypertriglyceridemia (sHTG), where triglyceride levels exceed 1,000 mg/dL, the risk of developing acute pancreatitis is 5-10 times greater than in the general population.
STP125G, an siRNA therapeutic targeting apolipoprotein C3 (ApoC3), developed by Sirnanomics, was tested in PXB-mice to demonstrate its efficacy in reducing triglyceride levels [27]. ApoC3 is a key player in triglyceride metabolism and has recently been recognized as a factor influencing cardiovascular, metabolic, and neurological disease risk.
The study in PXB-mice showed a single dose of STP125G resulted in high-efficiency, durable knockdown of ApoC3, and significant reductions in both mRNA and protein levels of ApoC3 were observed up to six weeks post-treatment (Figure 7A). Corresponding reductions in triglycerides and cholesterol were observed, returning to control levels by week 8 (Figure 7B).
The results in the humanized liver model provided strong support for ApoC3-targetting siRNA as a therapeutic approach in hypertriglyceridemia management, paving the way for clinical development.
The development of RNA therapeutics, particularly siRNAs, represents a promising frontier in modern medicine. Yet, the path from preclinical studies to successful clinical outcomes is fraught with challenges, many of which stem from the limitations of traditional animal models.
Humanized liver chimeric models, such as the PXB-mouse, offer a powerful solution to these challenges. By providing a more human-relevant environment for preclinical testing, these models enable direct targeting of human genes in vivo and more accurate assessment of drug metabolism and pharmacokinetics. As a result, you obtain a better prediction of efficacy and safety in humans and can evaluate therapeutics against human-specific pathogens.