Animal Models of Hepatocellular Carcinoma for Local-Regional Intraarterial Therapies

Abstract

Animal models play a crucial role in developing and testing new therapies for hepatocellular carcinoma (HCC), providing preclinical evidence prior to exploring human safety and efficacy outcomes. The interventional radiologist must weigh the advantages and disadvantages of various animal models available when testing a new local-regional therapy. This review highlights the currently available animal models for testing local-regional therapies for HCC and details the importance of considering animal genetics, tumor biology, and molecular mechanisms when ultimately choosing an animal model.

Keywords: Animal Studies, Interventional-Vascular, Molecular Imaging-Clinical Translation, Molecular Imaging-Cancer, Chemoembolization, Liver

© RSNA, 2022

Summary

When determining the most suitable animal model to assess local-regional therapies for hepatocellular carcinoma, careful consideration must be given to the advantages and disadvantages of each model, with emphasis on resemblance to human tumor and microenvironment.

Essentials

• ■ Several types of animal models are available for studying local-regional therapeutics for hepatocellular carcinoma, including mouse, rat, rabbit, woodchuck, and swine models.

• ■ Deciding on a model for testing a particular novel therapeutic agent depends on balancing advantages and disadvantages, with consideration of tumor cell type, mutational status, presence of cirrhosis, tumor immune microenvironment, and ability to use imaging and catheterization.

• ■ Initial testing in smaller animals followed by catheterization protocols in larger animals offers a route for effective translation of preclinical protocols to clinical implementation.

Introduction

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related death worldwide, and the incidence of HCC has been continually rising (1). Although liver transplantation and surgical resection are considered curative treatments for HCC, most patients have a poor prognosis at diagnosis and do not meet the criteria for these treatments (2). Historically, patients with advanced disease were generally offered systemic treatment with the kinase inhibitor sorafenib, but there is now evidence that immunotherapy may be beneficial in patients with advanced disease (2,3). More specifically, the U.S. Food and Drug Administration has approved the use of atezolizumab plus bevacizumab after a phase III randomized controlled trial showed improved overall and progression-free survival compared with sorafenib (4).

However, local-regional therapies, including ablative and transcatheter-directed therapies, are currently the standard of care for patients with HCC. Thermal ablation is an effective treatment for early-stage HCC when tumors are smaller than 3 cm, and such treatment should especially be considered in patients with baseline cirrhosis (5). Transcatheter-directed therapies include transarterial embolization (TAE), conventional transarterial chemoembolization (TACE), drug-eluting bead TACE, and transarterial radioembolization (ie, yttrium 90 embolization). Thermal ablation is generally offered to patients with Barcelona Clinic Liver Cancer (BCLC) stage A HCC, although there is early evidence of its safety and efficacy in the stage C cohort when combined with additional therapies (6). Transcatheter therapies are applied to patients with BCLC stage B or non-BCLC stage B HCC undergoing this treatment as bridge therapy or for downstaging (5,7–11). Although these modalities are currently widely used in HCC treatment, there is little comparative evidence on which transcatheter treatment method is superior or results in less morbidity; thus, operator and institutional preference rather than results from clinical trials often determine the treatment.

While in vitro studies are crucial in the identification of novel treatments and in guiding further modifications of already existent treatment modalities, applying these findings to human clinical trials has been challenging. Animal models play a vital role in understanding the pathophysiology of HCC and disease progression, and they serve as necessary preclinical tests of efficacy prior to introduction into clinical practice. This review focuses on the various animal models currently available for translational research on local-regional therapies for HCC, with some emphasis placed on how these models compare in mimicking the human tumor.

Identifying the Ideal Animal Model

Choosing the suitable animal model is critical in designing translational experiments that will lead to the introduction of a new clinical protocol. There are currently several HCC animal models in use, spanning a range of sizes, including mice, rats, rabbits, woodchucks, and swine. Most importantly, to the extent possible, the animal model should mimic the human pathophysiology of HCC at a molecular, cellular, and tissue level. For example, some animal tumors are not hepatic in origin or involve undefined mutations from DNA damaging agents. Another factor to consider when interpreting the results of treatment on a model generated in the absence of cirrhosis is that human HCC is generally in the background of cirrhosis. Furthermore, the model ideally should have a tumor microenvironment similar to that of human HCC so that the effects of embolization and immunotherapy can be studied. This is especially important when considering that in situ tumor development versus implanted tumors will naturally have a different tumor microenvironment and thus may vary in their response to different therapeutics. To this end, it may be necessary to use immunocompetent animals to successfully study immunotherapy.

The ideal animal model would have anatomy and overall tumor burden that lends itself to treatment with tools used in the clinic with humans, that is, catheter- or ablation-based treatment. It should have features to monitor treatment outcomes, such as radiologic landmarks or biochemical markers that might facilitate a better understanding of how the tumor and surrounding tissues respond to the treatment. Notably, the ideal animal model should be produced at a reasonable cost and within a reasonable time frame to monitor the disease course and the effect of treatment outcomes.

Catheter-directed embolization diminishes arterial blood flow to the tumor, and the presumption is that this results in acute transient ischemic and metabolic changes to the tumor, ultimately leading to tumor cell death. Angiographic stasis of blood flow is the basis for a clinical treatment end point, but the actual level of ischemia caused by embolization is uncertain because it has been shown that a substantial portion of tumor microvasculature not visible at angiography remains perfused (12). This is critical to consider when choosing an animal model, because the translation of embolization-driven hypoxia cannot be assumed. For example, the vascular anatomy of an animal may require a particular embolic formulation to achieve hypoxia levels observed in humans, but the use of the particular embolic formulation may not have a comparable effect in humans.

Mouse Models

Mouse models of HCC have been created by employing exposure to various chemical carcinogens (eg, aflatoxin, carbon tetrachloride exposure, diethylnitrosamine [DEN]), as well as by using xenografts (implanting human tumor cells into immunosuppressed mice), syngenetic implants (implanting mouse HCC cells into other mice), and genetic modification (initiating specific mutation leading to HCC) (13). Mouse models are a popular option for preclinical HCC research owing to the numerous advantages they offer: defined genetics, reproducibility, low cost, established care protocols, ease of handling, and the already existing data on manipulating the model to answer specific clinical questions. Mouse models are smaller and thus more advantageous for certain imaging modalities, including fluorescence imaging, because of the thin depth of tissue. If a cirrhotic model is desired for a particular experiment, profibrogenic agents can be used to induce fibrosis (14). With growing interest in the role of the immune system and in tumor–immune system interactions, there are now so-called humanized rodents for testing novel immunotherapeutics (15). Zhao et al (16) generated a patient-derived xenograft tumor model with functional human T cells, natural killer cells, and monocytes, which allowed them to study the response of specific human immune cells to human HCC tumors in mice.

While the characteristics listed above are advantageous for evaluating various treatments, there are several limitations as well, as shown in the Table. The genomic alterations observed in human HCC are highly heterogeneous, and therefore, mouse HCC models cannot replicate human HCC disease at a molecular level (17). Perhaps the most critical limitation is the size of mice and their visceral vessels that prevents use of catheter-based treatments, which ultimately eliminates mouse models from the evaluation of many potential local-regional therapies. Regardless, multiple studies have been published on ablation of HCC that have provided new insights and opened new avenues of research (18,19).

Advantages and Disadvantages of Most-Used Models for the Study of Local-Regional Therapy for Hepatocellular Carcinoma

Rat Models

The two most commonly used rat HCC models are syngenetic and use either the Novikoff or the Morris rat hepatoma cell lines. The Novikoff N1S1 cell line was derived from exposing Sprague-Dawley rats to 4-dimethylaminoazobenzene (20). These cells can be transplanted back into the livers of Sprague-Dawley rats. The Morris rat hepatoma McA-RH7777 cell line was derived from exposing Buffalo rats to N-2-fluoroenylphthalamic acid (21). These cells can be transplanted into the livers of Buffalo rats or xenogenetically implanted into the Sprague-Dawley rats (22). The most important advantage of these rat models, when compared with the smaller mouse models, is their amenability to transarterial treatments (22). Choi et al (23) demonstrated that McA-RH7777 tumors are more hypervascular than N1S1 tumors and that tumor vascularity can be further enhanced by vascular endothelial growth factor (VEGF) transfection.

One drawback of the Morris and Novikoff models is the lack of concurrent liver cirrhosis. Gade et al (24) addressed this by generating an autochthonous Wistar rat cirrhotic HCC model through oral administration of DEN over 3 months. They successfully performed superselective segmental embolization of these rats and showed the similarities in tumor vascularity between DEN-induced Wistar HCC and human HCC. Figure 1 (reproduced from Gade et al) shows anteriograms obtained during superselective embolization of Wistar rats (25). Importantly, Gade et al (24) noted that the underlying cirrhosis also narrows the therapeutic window for avoiding clinically significant liver injury, similar to HCC in humans. The DEN Wistar model has the added advantage of a local immune microenvironment that is more replicative of human HCC when compared with the syngenetic models, because the implantation process itself causes a local immunologic response.

Figure 1:

Coronal images from superselective hepatic transarterial embolization in Wistar rats. (A) Representative pre-embolization arteriogram shows the selection of a second-order branch of the right hepatic lobe artery, with contrast material extending into the feeding vasculature and the tumor (arrows). (B) Representative postembolization pull-back arteriogram shows vascular pruning and exclusion of the tumor-feeding vessel, as well as the distal tumor within the right hepatic lobe (arrows). (Reprinted, with permission, from reference 25.)

Although rat models offer new opportunities for testing local-regional therapies, the 3-month tumor development time is a limitation. The DEN Wistar HCC model described above is more biologically representative of human HCC compared with the more commonly used rabbit VX2 model described below. However, there are some differences in signaling pathways that should be considered when choosing an appropriate animal model. For example, DEN tumors almost always have the Braf^V637E mutation, which is rare in human HCC (26). Additionally, some of the N1S1 cell lines undergo regression over time, introducing another potential confounding factor when studying the effects of treatment (27).

It should be noted that the small arteries in rats, although larger than those in mice, are very difficult to catheterize for embolization and require considerable operator skill gained through experience. In a study by Nishiofuku et al (28) in which 174 rats underwent TACE via a transcarotid approach, proper hepatic artery catheterization was achieved 74% of the time. They reported success in 92% of tumor-bearing rats compared with 68.3% in non–tumor-bearing rats, as well as a significantly predictive relationship between the rat’s body weight at the time of intervention and successful hepatic artery catheterization (28). While choosing the largest rats available may be a reasonable approach, the equipment required for posttreatment imaging of larger animals may be a limiting factor. Investigators should consider these factors and understand that variability exists in the reported technical success and that there will be a substantial learning curve in laboratories with less-experienced staff. Additionally, procedural-related mortality must be considered because even in an animal with a successful infusion, periprocedural mortality would preclude outcomes study of that animal.

Gade et al (24) used 1-F microcatheters to achieve selective segmental embolization. With arteries that small, the question remains whether target hypoxia levels are attained even if the intended vessel can be successfully catheterized. In humans, embolization is generally terminated when a level of near-stasis is achieved angiographically. However, Johnson et al (12) demonstrated that even when angiographic stasis is reached, considerable tumor microvascular perfusion persists, which may contribute to tumor survival and recurrence. Even if uniform embolization end points are achieved across the cohort, based on gross examination, variability in tumor hypoxia levels likely exists at the cellular level. Performing the embolization on such narrow vessels using small devices further adds to the unpredictability of the actual levels of tumoral hypoxia achieved. In the rat model, Gade et al (24) demonstrated an average tumor necrosis of nearly 50% in response to embolization. It remains speculative if embolized rat tumors experience the same level of hypoxia as embolized human HCC; therefore, it is unclear whether the tumor microenvironment in such a small model accurately reflects that of human HCC.

Rabbit Models

The most commonly used rabbit HCC model is the VX2 tumor, an anaplastic squamous cell carcinoma orthotopically allografted into rabbit hindlimb (29). The rapid growth of the tumor, larger vessel size than in rodents, and overall reliability make this an efficient and popular model of HCC, as shown in the Table. Multiple research groups have studied local-regional arterial and ablative therapies with this model (30–32). Figure 2 demonstrates superselective embolization via the hepatic artery in two VX2 rabbits (30). The larger size of the VX2 model also allows for more cross-sectional imaging compared with rodent models.

Figure 2:

Coronal images from superselective hepatic artery digital subtraction angiography in two representative rabbits. (A, C) Baseline pre–transarterial embolization (TAE) images show peripheral enhancement of VX2 liver tumors (arrowheads). (B, D) Corresponding post-TAE images show complete lack of tumor enhancement (arrowheads) and increased definition of proximal vessels (arrows) owing to reduced flow to the tumor and subsequent reflux of contrast agent into adjacent vessels. (Reprinted, with permission, from reference 30.)

The principal downside of the VX2 tumor model is that it does not mimic the human pathophysiology of HCC at several levels. The VX2 tumor is of squamous cell origin, whereas human HCC is an adenocarcinoma. MicroRNA (miRNA) profiles of rabbit VX2 tumors are quite distinct from human HCC miRNA profiles (33). This may be indicative of various genetic and molecular differences between VX2 and human HCC. Additionally, the recipient immune microenvironment may be altered in this model as a result of the orthotopic allograft. The VX2 tumor is also difficult to grow in vitro and may often require in vivo growth prior to transplantation. This not only makes the initial tumor growth methods more arduous but also limits biochemical studies conducted on the cell cultures. Last, the VX2 model tends to have spontaneous central tumor necrosis, and it may be necessary to use a modified VEGF-overexpressing model to more carefully study tumor response to treatment (34).

Regardless of differences in tumor biology, the VX2 model for HCC remains popular and can be useful when evaluating novel local-regional therapies. The larger vessel size, when compared with rat models, may in theory allow for better reproducibility of embolic delivery and the local tumoral hypoxia. Any investigator who chooses to use this model must be careful when drawing conclusions about the intervention, especially regarding mechanisms.

Woodchuck Models

The woodchuck hepatitis model offers key similarities to human HCC in that it develops in situ subsequent to viral infection. The woodchuck hepatitis virus (WHV) is injected into woodchucks, and HCC develops over several years. During this period, woodchucks undergo routine surveillance to monitor for the appearance of liver tumors (35). While the model is hepatic derived, there is typically no fibrosis or scarring of the woodchuck livers, so the model is not cirrhotic. The anatomy and size of vasculature of the woodchuck are amenable to endovascular treatments (36,37) and ablation (38). Figure 3 demonstrates an example of a woodchuck HCC prior to embolization and follow-up imaging 3 months after treatment.

Figure 3:

(A, D) Axial precontrast T1-weighted, (B, E) arterial phase, and (C, F) portal venous phase images from (A–C) before transarterial embolization (TAE) and (D–F) after TAE after superselective hepatic artery embolization in one woodchuck. (A–C) Pre-TAE images show a larger tumor measuring up to 2.3 cm in the left hepatic lobe (arrow in B). (D–F) Corresponding post-TAE images show decreased size of the tumor after embolization (arrow in E).

The length of time to develop a tumor after infection with WHV is perhaps the most critical disadvantage of this model, because housing the animals for months or even years is expensive. The tumors that ultimately develop are also heterogenous in size and number and may range from benign to malignant. Because the model is not cirrhotic, the doses of particular preclinical treatments may not be harmful to woodchucks with their reserve liver capacity, but the human therapeutic window may be significantly narrower. Last, it may be more challenging to perform TACE in the woodchuck model than in the VX2 rabbit model because the left gastric origin of the left hepatic artery in the woodchuck makes embolizing lesions in the left lobe challenging (37).

Swine Models

Swine HCC models are an especially valuable option with numerous similarities to humans in terms of size, anatomy, and tumor biology. Although small and medium laboratory animal models are suitable for some preclinical and pharmacologic studies, larger animal models are beneficial for studying imaging, surgical, and other interventional procedures. Two swine models that can be considered are the DEN-induced model and the Oncopig model. The DEN-induced model is an autochthonous model generated by intraperitoneal DEN injections over 3 months (40,41). However, among the decided disadvantages of this model are the unknown genetic perturbations in response to DEN, which are a source of variability. Additionally, this model ultimately takes 1–2 years before HCC develops in the background of cirrhosis, an interval that makes this model expensive because it requires large-animal housing for up to 3 years. Such issues severely limit use of this model.

The Oncopig cancer model (OCM) is created by crossbreeding an Oncopig with KRAS and TP53 mutations (after Cre recombinase exposure) with a domestic or miniature pig to develop tumor-bearing pigs (42,43). The OCM has several advantages. First, the tumors will all contain KRAS and TP53 mutations, which are pathways commonly affected in HCC and other human cancers (42). Second, tumors in OCM can be rapidly induced in a much shorter time frame compared with the DEN-induced model. Additionally, the Oncopig is immunocompetent and thus amenable to experiments involving immunotherapies. Recently, Nurili et al (44) induced liver tumors by in situ inoculation in 18 Oncopigs and successfully performed TAE in eight. After selective segmental embolization of tumors, the size of tumors was reduced when compared with untreated tumors, which continued to grow. By demonstrating that the Oncopig liver tumor blood supply is comparable to that of human tumors and that these tumors respond similarly to TAE, Nurili et al (44) opened the door to assessing other arterially directed therapies (45).

The large size of swine models is not only beneficial from an interventional treatment standpoint, but it also allows for easier blood and tissue sampling in longitudinal studies. Drug safety and toxicity measures generated from swine studies are also much more representative of human levels (45). Additionally, the size of swine makes them more amenable for cross-sectional imaging compared with rodents. However, the OCM is a relatively new model and requires unique technical experience. For instance, specific measures must be taken to isolate a specific cell type prior to reimplantation of the cell type. If this is not performed properly, the tumors generated will be composed of a mixture of types of cells of origin (45). A critical disadvantage of this model, like the DEN-induced model, is its overall cost.

Discussion

Currently, no one HCC animal model is ideal for every experimental question. The ideal animal HCC tumor model should mimic human tumors in terms of vascular anatomy, underlying cirrhosis, and tumor biology at the genetic and molecular level. An immunocompetent animal also offers the advantage that local response to a therapy can be reproducibly evaluated over time. The challenge with testing catheter-based therapies is that the tumor immunobiology cannot be considered until the anatomy and vasculature lend themselves to catheterization. Regardless, there are several models in use, including the DEN Wistar rat, rabbit VX2, woodchuck WHV, and Oncopig. Animal models continue to close the translational gap between laboratory research and human clinical trials. When using these models, a greater emphasis is naturally placed on understanding the molecular mechanisms of responses and resistance. When choosing the appropriate model for testing a particular novel therapeutic agent, the pathway, tumor, and cell-specific interactions will ultimately help narrow selection.

Local Hypoxic Effects

One example that highlights the importance of hypoxia in the tumor microenvironment is a study by Duran et al (46) that used evofosfamide, a hypoxia-activated prodrug that has been shown to have anticancer properties in hypoxic environments. VX2 rabbits that underwent TACE and evofosfamide administration had lower tumor volumes and higher necrotic fractions than VX2 rabbits that underwent only TACE. It was demonstrated that the greater tumor necrosis in the evofosfamide group is likely secondary to the expected hypoxia-driven effect. This study is encouraging for future studies of local-regional targeting of tumor hypoxia. When accounting for the unpredictability of the actual postembolization tumoral hypoxia achieved and the differences in squamous cell–derived and true hepatic-derived tumors, there will naturally be differences in the prodrug activation and in the subsequent effect of these prodrugs on tumors. This will need to be considered and accounted for before replicating this study in humans. The same holds true for experiments that involve other hypoxia-induced survival pathways in nonhepatic-derived animal models or animal models with a significant difference in vessel caliber and the extent of hypoxia achieved. For instance, investigations have demonstrated chloroquinoline-induced inhibition of autophagy (25,47). This is where the use of larger animals with true hepatic-derived tumors, such as the woodchuck WHV or Oncopig, will play a role in further bridging the findings in the VX2 model and human HCC.

Other hypoxia-related experiments involve caffeic acid (CA) and the monocarboxylate transporter (MCT) pathway. When human HCC cells are placed into hypoxic conditions, tumor cells increase glycolysis and excretion of lactate, a byproduct of glycolysis. Unless lactate is excreted, lactate feedback inhibits glycolysis at the rate-limiting step of phosphofructokinase. Thus, the removal of lactate becomes vital to cellular survival (48). MCT proteins are upregulated and efflux intracellular lactate (49). CA, a derivative of cinnamic acid with a structure similar to lactate, has been shown to inhibit MCTs and ultimately cause apoptosis of HCC cells (50). When choosing an appropriate animal model for testing CA as a local-regional therapeutic, the animal model must have a biologic response to hypoxia comparable to that in humans and one in which MCTs are upregulated. Although embolization of the rabbit VX2 model with plain TACE has demonstrated the expected decrease in extracellular pH (51), the VX2 model is squamous cell derived and exhibits a different miRNA profile. It is likely that MCT expression in the squamous cell–derived tumor will be different than for HCC; thus, embolization with CA of the VX2 model may have a different efficacy than embolizing an actual hepatic-derived HCC model like the woodchuck WHV.

Role of Immune Microenvironment

The effects of local-regional therapy on the microenvironment of the liver are also very important and must be considered. The immune environment in HCC is suppressed, as T-helper cells and myeloid-derived suppressed cells release immunosuppressive interleukins and downregulate particular pathways (52). After local-regional treatments modulate the local hepatic immune environment, the antitumor immune response may be increased. For example, in an HCC mouse model, Iida et al (53) demonstrated that a single radiofrequency ablation (RFA) treatment inhibited growth of untreated tumors by increasing cytotoxic T-cell response. Similarly in humans, there is an increase in the ratio of CD4/CD8 cells and the number of Th17 cells after TACE, which correlated with improved survival (54). This immunomodulating effect of local-regional therapy has been of much interest recently; as of 2020, there were more than 45 registered clinical trials studying the effects of combining local-regional therapies with immune checkpoint inhibitors (55). Additional research questions will arise as results from these trials are published, and this is expected to be a field of interest moving forward.

From a preclinical standpoint, the differences in the mechanisms and pathways of the immunomodulating effects of local-regional therapies must be considered. In a study by Behm et al (56), VX2 model rabbits were randomized to receive no therapy, RFA monotherapy, CpG monotherapy, or RFA plus CpG oligodeoxynucleotides, which are direct toll-like receptor 9 agonists. They theorized that additional treatment with CpG would increase antitumor T-cell response to RFA by stimulating the innate immune receptors, ultimately increasing presentation of tumor antigens by dendritic cells to cytotoxic T cells. They ultimately found that rabbits treated with a combination treatment had improved mean survival, T-cell activation and cytotoxicity, and decreased formation of new tumor growth. Additionally, cytokine level analysis demonstrated markedly elevated levels of proinflammatory Th1-related cytokines and tumor-specific lymphocytes. The investigators took into consideration the difference in immunogenetics of their syngenetic VX2 model compared with humans or mice and ultimately the benefits of this model, including the overall reliability and amenability for ablation. Conclusions drawn about immunomodulating effects of local-regional therapy may be different, and this must be kept in mind when extrapolating findings to humans.

Role of miRNA Profiles

The miRNAs are small noncoding RNAs that play a critical role in nearly every biochemical pathway by targeting specific mRNA sequences and tagging them for destruction, thereby downregulating protein translation. Multiple recent review articles have highlighted specific miRNAs that have been up- or downregulated in the pathogenesis of HCC (57,58), including development of tumor resistance to sorafenib (59). Numerous studies additionally demonstrated that miRNAs modulate tumor angiogenesis and microenvironment via VEGF and transforming growth factor-β (60,61). This is of particular interest when studying effects of local-regional therapies, because miRNAs may serve as diagnostic and prognostic markers of posttreatment outcome. Liu et al (62) showed that in 136 patients with HCC treated with TACE, higher pretreatment serum miR-200a levels were associated with lower survival. Studies have demonstrated various associations between miR-200a and roles in tumor pathogenesis, but the exact mechanism of the relationship remains to be elucidated. Future translational studies on the role of miRNA profiles and how they change following local-regional therapy will be interesting and will need to account for differences in the animal model used for experimentation, such as differences in the miRNA profile of the rabbit VX2 tumor model compared with humans (33).

Additionally, miRNAs are being studied as therapeutics for targeted treatment of HCC. In a study by Wang et al (63), human HCC (HepG2 cells) xenograft mice models were treated with combinations of doxorubicin and different miRNA encapsulated in biodegradable poly lactic-co-glycolic acid-polyethylene glycol nanoparticles. They found that in mice treated with the nanoparticles, which were administered with a US-guided microbubble-mediated targeted delivery approach, tumor growth was delayed when compared with that in control mice. Similar to diagnostic miRNA studies, studies examining the therapeutic side of miRNAs will also need to consider the unique miRNA profiles of the animal model chosen.

Imaging Considerations

HCC in animal models may be monitored during pre- and posttreatment phases with various imaging methods, the most common of which are US, CT, and MRI. A basic understanding of the appropriate imaging equipment required is crucial when designing a translational experiment. Models based on exogenous exposure and tumor development, such as the DEN Wistar or woodchuck WHV models, often require at least monthly US to screen for tumor development. US also plays a crucial role in the injection of tumor cells into the target tissue in orthotopically allografted models, and it is used for image guidance during ablation procedures. In the posttreatment setting, MRI is usually the preferred longitudinal imaging modality because of the superior soft-tissue contrast and high spatial resolution. MRI allows for accurate demarcation of tumor margins and determination of tumor volume, thus allowing investigators to capture treatment response. Although CT offers high spatial resolution, it is not widely used for monitoring treatment response owing to its lower capacity for soft-tissue contrast and potential carcinogenic effects of high doses of radiation. Cross-sectional imaging in general is easily applied to large animals and is much more challenging to apply to rodents because specialized more expensive MRI scanners with higher magnetic field strengths are required. Freimuth et al (64) developed a standard protocol for imaging HCC in mice using 1.5-T MRI, which has been modified and used by other groups.

In addition to providing temporal and spatial information, advanced imaging modalities provide a noninvasive method to obtain information related to molecular and tumor microenvironment. Bioluminescence is a method in which the luminescence emitted from expression of luciferase is quantified in cancer cells. HCC cells are modified to express this luciferase and are reintroduced into the host, after which luminescence may be tracked longitudinally before and after treatment (65,66). The major benefit of this technique is that it allows for quantitative imaging of tumors that may be difficult to visualize at MRI. Although bioluminescence has provided insights in preclinical research, especially for evaluating drug efficacy, it has not offered substantial translational benefits.

Conclusions

When testing a novel local-regional therapeutic for the treatment of HCC, emphasis should naturally be placed on choosing a model with a tumor and microenvironment that most closely resembles those of humans. These factors requiring consideration, particularly the genetic differences in tumors, participation of the immune system, and variability in the molecular response to embolization, introduce a level of error into animal experiments that help explain the challenge in translation of a local-regional therapy from animal to human. Additionally, experiments should have reliable reproducibility and be performed in a time- and cost-efficient manner. When conducting preclinical testing of a new therapeutic, the interventional radiologist will need to weigh the advantages and disadvantages of each model and consider the molecular mechanism of the therapeutic and barriers to replicating that mechanism in humans. The rat N1S1 model is both reproducible and relatively inexpensive but has drawbacks in the variations in tumor vascularity and hypoxia achieved. The rabbit VX2 model, which currently is perhaps the most popular model, is much more comparable to human vasculature and offers more reproducible embolization end points. This tumor, however, is epithelial in origin and has significant molecular variability, which makes it difficult to predict clinical effects in humans. The larger woodchuck and swine models are much more genetically similar to humans and have vascular anatomy amenable to embolization. However, these larger models require larger housing facilities and much more extended incubation periods, making them an expensive choice. One approach that can be considered is performing initial translational experiments in cheaper and smaller tumor models to establish efficacy and then performing a more focused experiment in larger animals, with tumors more genetically similar to human HCC.