Characterization of inflammatory pseudotumors in a large animal model of liver cancer

Abstract

Background

The development of relevant and robust large animal models of hepatocellular carcinoma is needed to test new therapeutic strategies for this disease. Transgenic approaches hold promise in addressing this complex problem. One such model, the Oncopig, has been reported to develop tumors of up to 4 cm in diameter within 7–14 days at sites of in situ vector inoculation. However, the resulting lesions reportedly contained an extensive inflammatory component that has not been evaluated in detail.

Methods

Herein, we describe our results from multiparametric characterization of the lesions generated using liver biopsy cores incubated in vector solution and replaced in the tissue. The study consisted of 3 animals in 3 cohorts (total of 9 animals) that were evaluated at 14, 21, and 28 days. CT imaging, immunohistochemistry, multiplex immunofluorescence, and comprehensive blood analyses were used to quantify composition of the hepatic masses that developed following AdCre inoculation.

Results

The tumors were hypovascular on CT and predominantly composed of CD45+ cells with a strong lymphohistiocytic component, with no carcinomas identified. Ki‐67 staining showed proliferation of CD45+ immune cells but no neoplastic component. To provide further insight, the results are evaluated in the context of tumor growth kinetics.

Conclusion

While progress has been made in generating targetable lesions, achieving a robust large animal model of liver cancer that faithfully recapitulates the human disease remains a challenging goal.

CT images in arterial and portal venous phases of liver lesions in Oncopig.

1. INTRODUCTION

Because they mimic the human condition more closely than cells grown in culture do, murine models are ubiquitous in cancer research. However, most drugs found to be effective in mice ultimately fail in humans in clinical trials, in part because of the significant gap in physiology between the species. ¹ Thus, scientists have looked to larger animals whose physiology is much closer to that of humans, with greater genetic homology and similarities in immune function.

Model selection for cancer research in large animals presents a dilemma: are spontaneous or induced tumors more useful? Spontaneous tumors generally have the highest relevance, but their availability is unpredictable, making it difficult to execute statistically robust studies. Generally, livestock animals are not suitable sources of spontaneous tumors as they are raised to gain weight quickly and go to market at a comparatively young age when tumors are rare. Companion animals, which can live long enough to develop tumors, might be suitable sources, but species and the types and stages of tumors at presentation vary, as does owners' willingness to subject their pets to study and to pay for experimental treatment. Induced or transplanted tumor models are more readily available than spontaneous tumor models but present their own challenges. Many transplanted models involve either immune‐suppressed or immune‐deficient large animals. Such animals are expensive to acquire, significantly more costly to house than immunocompetent animals, and, despite precautions, inherently more susceptible to complications. ² Furthermore, research performed in the context of an altered immune system may not accurately reflect the disease being studied. Many induced tumor models, in which tumor development can result from toxin, environmental, or viral exposure or treatment with a genetic editing technology such as CRISPR/Cas9, require significant time and expense to develop, and their reliability varies. ³ , ⁴ , ⁵ , ⁶

An ideal large animal inducible tumor model would be immunocompetent, inexpensive, and readily available; enable reliable, reproducible experiments; develop carcinomas in a reasonable time; have specific relevant genetic mutations; and require minimal specific expertise for implementation. ⁷ One recently reported model, the Oncopig, potentially meets several of these criteria. The Oncopig is a transgenic, immunocompetent model that is primed for the activation of the oncogene KRAS (with the KRAS ^G12D mutation) and the loss of the functional tumor suppressor p53 (through the p53 ^R167H mutation) upon exposure to a suitable adenoviral vector encoding Cre recombinase (AdCre) with a cytomegalovirus promotor. Many types of cancer frequently have the KRAS ^G12D and p53 ^R167H mutations, ⁸ and KRAS has been a target for pig research. ⁹ In the Oncopig model, tumors reportedly form within 1–2 weeks after AdCre exposure, in principle at any location in the animal. ¹⁰ However, these tumors were frequently described as undifferentiated and of undetermined origin, not specific to the organ in question. Accurately recapitulating the biology in models of neoplastic disease is essential for advancing the science, correctly interpreting results, and enabling clinical translation. Characterization of the cellular composition of tumors and the behavior of tumors over time are important aspects that profoundly impact the validity of conclusions drawn using the model and hence the utility. Researchers have consistently found that induced lesions in the Oncopig have an extensive inflammatory component, and that tumors are very similar histologically irrespective of location, whether in liver, lung, or pancreas. ¹¹ , ¹² Lymphocytic, plasmacytic, and macrophage infiltrates have been reported along with multinucleated giant cells. Our own interest in the Oncopig pertains to assessing suitability for testing experimental therapies for hepatocellular carcinoma (HCC). Previous studies of HCC with the Oncopig model have focused largely on genetics and, to a much lesser extent, immunohistochemistry. ¹³ However, due to reports of a large inflammatory component, the relative composition of the cell populations of Oncopig liver tumors was of interest. As this has not been described, we report our multiparametric characterization of liver lesions in the Oncopig model and provide additional context with an analysis of tumor growth kinetics.

2. MATERIALS AND METHODS

2.1. General

The study was performed under a protocol approved by MD Anderson's Institutional Animal Care and Use Committee (00002158‐RN00; approved 6/15/2021). The study consisted of 3 cohorts of 3 animals each. The animals underwent baseline imaging and inoculation followed by imaging every 7 days and euthanasia at 14, 21, or 28 days (3 pigs per time point). All procedures, including imaging studies, were done with the animals under general anesthesia. Intramuscular injection of tiletamine hydrochloride and zolazepam 2.2–6.0 mg/kg (Zoetis US, Parsippany, NJ, USA) and buprenorphine 0.01–0.1 mg/kg were used to induce anesthesia; following intubation, anesthesia was maintained with 4% isoflurane. Buprenorphine SR (0.12–0.2 mg/kg) or meloxicam (0.3–0.5 mg/kg) were used postoperatively and subsequently at the discretion of the veterinary staff. Heart rate, respiratory rate, and body weights were recorded at baseline and every 7 days thereafter. Blood was collected for complete blood count (Advia 2120i, Siemens Healthineers, Forchheim, Germany) and comprehensive metabolic panel (Integra 400 Plus, Roche Diagnostics, Rochester NY) before and after inoculation and every 7 days thereafter to assess for systemic responses, metabolic abnormalities, and evidence for liver injury. For red blood cells, these included red blood cell count, hematocrit, hemoglobin, mean cell volume, mean corpuscular hemoglobin, and platelet count. For white blood cells, these included total white blood cell count and differential counts for neutrophils, lymphocytes, monocytes, basophils, and eosinophils. Serum chemistries included sodium, potassium, calcium, chloride, phosphorus, glucose, total protein, albumin, globulin, bilirubin, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, and lactate dehydrogenase. An overdose of phenytoin and pentobarbital was used for euthanasia.

2.2. Animals

Nine male transgenic Oncopigs bearing the CAG promoter and LSL‐KRAS ^G12D and p53 ^R167H mutations (only males provided, University of Illinois Urbana‐Champaign) were inoculated with the AdCre vector (10⁹ pfu of Ad5CMVCre‐eGFP per site; University of Iowa Viral Vector Core, Iowa City, IA) according to published methods. ¹⁰ Briefly, a 17/18G coaxial biopsy system was percutaneously inserted into the liver under CT guidance. A core biopsy specimen was obtained, incubated with the vector for 20 min, combined with Gelfoam (Pfizer, New York, NY), and replaced at the original biopsy site, after which the needle was removed and the animal allowed to recover. The procedure was repeated in right lateral, right medial, left medial, and left lateral lobes of the liver for a total of four inoculations per animal. Success was defined as CT imaging on at least two weekly scans showing new masses in liver consistent with areas of prior inoculation.

2.3. Imaging

Computed tomography (CT) imaging studies were performed using a Dual‐Energy Definition Edge system (Siemens Healthineers, Forchheim, Germany). CT was performed using a liver protocol (three phase scanning, including non‐contrast, arterial phase, and portal venous phase). Images were reconstructed at slice thicknesses of 1 mm or 3 mm. Images were further analyzed for contrast enhancement and volumes using OsiriX MD v14.0.1 (Pixmeo, Geneva, Switzerland). Regions of interest (ROIs) for analysis were chosen within the central lesions, at the margins, and nearby but unaffected liver tissue. Lesion volumes were calculated both in OsiriX and using 3D Slicer software version 5.2 (https://www.slicer.org/ accessed 6/8/2025) after manual segmentation of portal venous phase images.

2.4. Pathology

After euthanasia, each animal's liver was removed en bloc, photographed (Nikon D750, Nikon USA, Melville, NY), and the entire organ was serially sectioned at approximately 1 cm slice thickness, and lesion diameters were measured. Gross and microscopic specimens were examined by a board‐certified veterinary pathologist (NWF) for assessing morphology, describing and interpreting the findings, determining markers to use for further study, validating regions of interest, and interpreting results for immunohistochemistry and immunofluorescence. Representative sections of gross specimens were photographed prior to further processing. Tissue sections of lesions and surrounding tissues were placed in 10% neutral buffered formalin for fixation and subsequent processing. Hematoxylin and Eosin (H&E) staining and chromogenic immunohistochemistry were performed with Hematoxylin as the counterstain. Formalin fixed paraffin embedded (FFPE) tissue was sectioned at 4 μm with a microtome. Staining was performed using a Leica Bond RX (Leica Biosystems, Deer Park, IL, USA) using the Bond Refine DAB Detection Kit DS9800. Antibodies were diluted in Bond Diluent AR9352. Antigen retrieval reagents were Bond ER1 (Citrate Buffer) AR9961 and Bond ER2 (EDTA) AR9640. For fluorescence staining, an Opal 7‐color kit (Akoya Biosciences, Marlborough, MA, USA) was used with the Akoya Antibody Diluent/Block ARD1001EA for Buffer and PKI Block. The Akoya Opal Polymer HRP Ms. + Rb ARH1001EA was used for the Opal Polymer and the H₂O₂ block used was Fisher 30% H₂O₂ H325‐500. Antibodies were diluted with the Akoya Antibody Diluent Block. The Opal reagents were as follows: Opal 480 FP1500001KT, Opal 520 FP1487001KT, Opal 570 FP1488001KT, Opal 620 FP1495001KT, Opal 690 FP1497001KT, and Opal 780. These were diluted with Akoya Amplification Diluent FP1609 at 1:100 except Opal 780, which was diluted with Akoya Antibody Diluent/Block at 1:25. Spectral DAPI FP1490 was diluted with PBS at 3 drops per 1 mL. Order of antibody application was as follows: CD31, SMA, CD45, vimentin, Ki‐67, and pan‐CK. Whole slide digital imaging was performed using an Aperio AT2 for brightfield and a Leica Versa 8 for fluorescence (20× objective, pixel dimensions 39 839 × 35 926). Cell types were assessed using a panel comprising antibodies against Iba‐1 (a marker of macrophages; Abcam, Cambridge, UK, cat. no. 178847), CD3 (T cells; 1 Abcam, cat. no. 6669), CD20 (B cells; Thermo, cat. no. PA5‐16701), MPO (neutrophils; Abcam, cat. no. 9535), pan‐cytokeratin (pan‐CK; an epithelial marker used here for biliary hyperplasia and neoplastic cells with epithelial differentiation; Bioss, Woburn, MA, USA, cat. no. 1712R), arginase‐1 (hepatocytes, M2 macrophages), and vimentin (fibrosis; Cell Signaling Technologies, Danvers, MA, USA, cat. no. 5741) and CD45 (leukocytes; Abcam, cat. no. 10558). For quantitative analyses using multiplex immunofluorescence, we created an 18 × 12 element tissue microarray with 1 mm cores. Two cores were obtained from each representative central and peripheral area of liver lesions excluding areas of necrosis. Peripheral samples were taken at the interface of normal tissue and inflammatory lesions. Cell composition in the TMA was evaluated with entire cores taken as a region of interest using the quantitative image analysis program HALO (v3.6, High‐plex algorithm, Indica Labs, Albuquerque, NM, USA) for overall percentage and cell density. GraphPad Prism (GraphPad Software, Boston, MA, USA) was used to analyze for changes over time using one‐way ANOVA (p < 0.05) and for generation of pie charts to demonstrate changes in composition. Slides were imaged using a Versa 8 whole‐slide fluorescent scanner (Leica Biosystems). Control tissues included pig tonsil, spleen, pancreas, and lung. Samples were evaluated using the nuclear stain DAPI in conjunction with CD31 (endothelium; Abcam, cat. no. 28364), smooth muscle actin (SMA; Abcam, cat. no. 5694), CD45 (leukocytes; Abcam, cat. no. 10558), vimentin (Cell Signaling Technologies, cat. no. 5741), Ki‐67 (proliferation; Abcam, cat. no. 16667), and pan‐CK (biliary cells, neoplastic cells with epithelial differentiation, Bioss, cat. no. 1712R) to determine total cell counts and the relative abundances of each marker in the samples.

3. RESULTS

Animals tolerated imaging and AdCre inoculation well. No complications from any of the procedures were noted. Vitals remained within normal range over the course of the study. Pigs continued to grow and gain weight as anticipated.

3.1. Laboratory values

No clinically significant deviations from normal ranges were observed. ¹⁴ , ¹⁵ , ¹⁶ Data for complete blood counts and serum chemistries over the duration of the study for all subjects are presented in Figure S1 in Supporting Information.

3.2. CT findings

Axial CT images of the liver with contrast in arterial phase and portal venous phases with a representative tumor are shown in Figure 1. Overall success rate for induction of lesions was 78% across the entire group, with a maximum volume of 14 mL. Attenuation in each ROI was as follows, presented as arterial/portal venous phases: peripheral lesion, 117 HU/99 HU, central lesion, 52 HU/54 HU, normal liver parenchyma, 80 HU/102 HU. In the arterial phase (Figure 1A), the margin enhancement against background liver was relatively low; this was the case even in the 28‐day cohort where more vessel growth might be anticipated. In addition, the main corpus of each lesion did not show enhancement with contrast at any point. Finally, the enhancement of normal liver parenchyma in the portal venous phase (Figure 1B) equaled or exceeded that of the margin in the arterial phase. Regression of lesions over time was also noted, with complete disappearance in some cases. This was most notable across the interval from 21 to 28 days where all five remaining tumors in this cohort at this time point (of 12 theoretically possible) were markedly reduced in size.

FIGURE 1.

Axial CT images demonstrate the variation and dynamics of attenuation at each region of interest (ROI) in a representative tumor at 14 days. (A) Arterial phase showing peripheral tumor (green), central tumor (orange), and adjacent normal liver parenchyma (blue). (B) Portal venous phase showing peripheral tumor (blue), central tumor (orange), and adjacent normal liver parenchyma (yellow). Hounsfield units and other parameters for each region are shown.

3.3. Gross and histopathology

A representative specimen of a lesion with typical appearance from the 14‐day cohort is shown in Figure 2A. Lesions ranged in color from pale tan to red, were well demarcated, and were initially firm but had variably necrotic to caseous centers (reminiscent of granuloma) over time. The overall success rate for lesion induction across all inoculations in all animals was 78% (28 of 36 possible tumors). Tumor size spanned a wide range with a maximum diameter of 3 cm noted. Samples from all tumors had the same overall appearance both macroscopically and microscopically. H&E of a representative specimen shows the intense, mixed inflammatory response, including lymphocytes, plasma cells, neutrophils, and histiocytes (Figure 2B). Langhans type and foreign body type multinucleated giant cells were also frequently observed (Figure 2C). Focal, small areas of central necrosis (approximately 10%–30%) were observed not infrequently, with the relative amount increasing as tumor size increased.

FIGURE 2.

Pathology and histopathology of tumor. (A) Gross liver specimen at 14 days. The tan nodule in the center of the specimen can be readily identified. (B) Representative H&E section of tumor at 14 days showing extensive mixed inflammatory infiltrate including lymphocytes, plasma cells, neutrophils, and histiocytes. Scale bar: 300 μm (C) High‐power view showing frequent Langhans type and foreign body type multinucleated giant cells interspersed within the tumor. Scale bar: 100 μm

3.4. Chromogenic immunohistochemistry

Individual immunohistochemical stains for profiling the lesion cell composition for each cohort are shown in Figure 3 and were assessed qualitatively. Areas of AdCre injection were infiltrated predominantly by large numbers of CD45+ cells (leukocytes), which appeared to decline slightly with time. lba‐1+ cells (macrophages) and CD3+ cells (T cells) predominated, with fewer CD20+ cells (B cells) and MPO+ cells (neutrophils). Vimentin+ cells (fibrotic or other mesenchymal cells) were also present and remained relatively consistent over time. There were scattered, small areas of cells expressing pan‐CK (a marker for bile ducts, biliary hyperplasia, and neoplastic cells with epithelial differentiation); morphology of these cells was typical for bile ducts and biliary hyperplasia, not carcinoma. Moreover, these cells did not differ across samples from animals at different time points. Two populations of arginase‐1+ cells were identified based on cell and tissue morphology consistent with M2 macrophages and hepatocytes. Iba‐1 positivity colocalized with the arginase‐1+ M2 macrophage population, which peaked at 21 days before declining slightly at 28 days.

FIGURE 3.

Chromogenic immunohistochemistry for the indicated markers at 14, 21, and 28 days. Scale bar: 250 μm

3.5. Multiplex immunofluorescence

Multiplex immunofluorescence composite images and corresponding H&E‐stained sections with photomicrographs of peripheral and central lesions with all structural markers are shown in Figure 4. A strong inflammatory response persisted for the duration of the study. Enlarged views are provided in Figure S2 in Supporting Information.

FIGURE 4.

Composite multiplex immunofluorescence images demonstrate peripheral and central tumor composition at 14, 21, and 28 days. The multiplex immunofluorescence images appear below images of the corresponding H&E‐stained sections. Sections were stained for DAPI (blue), CD31 (teal), SMA (green), CD45 (gold), vimentin (red), Ki‐67 (white), and pan‐CK (magenta). All images were taken at 20× magnification. See Figure S2 in Supporting Information for enlarged views. Scale bar: 100 μm

Multiplex immunofluorescence images of staining for individual markers and DAPI in peripheral and central lesions over time are shown in Figure 5. There were significant increases in vimentin and CD45+ leukocytes over time. Vimentin and SMA showed co‐expression in some areas, consistent with myofibroblasts. At 28 days, SMA was absent or nearly absent in peripheral lesion but abundant in the immediately adjacent normal liver parenchyma. Similarly, CD31 staining was low and increased minimally over time in the lesions but was strong in adjacent normal liver at all time points. By 28 days, lesions consisted predominantly of CD45+ leukocytes and vimentin+ cells, presumably myofibroblasts. Finally, pan‐CK, a marker of epithelium, potentially reactive biliary hyperplasia, and neoplastic cells with epithelial differentiation, was present in only small, isolated areas; as a result, changes in the marker could not be accurately assessed over time. Larger images for each panel in Figure 5 are provided in Figure S2 in Supporting Information.

FIGURE 5.

Multiplex immunofluorescence images demonstrate staining for the indicated markers in peripheral and central tumor at 14, 21, and 28 days. DAPI (blue) was used as a nuclear stain. Separate images for each biomarker, as well as composite images of all biomarkers, were obtained from the same specimens of peripheral and central tumor, which allowed for correlation with H&E images. Sections were stained for CD31 (teal), SMA (green), CD45 (gold), vimentin (red), Ki‐67 (white), and pan‐CK (magenta). All images were taken at 20x magnification. Scale bar: 100 μm See Figure S2 in Supporting Information for larger images of individual images.

CD45 and Ki‐67 images show that CD45 staining was intense and strongly overlapped with Ki‐67 staining (>90% co‐expression) as shown in Figure 6. This is noteworthy when viewed in light of the paucity of staining by the pan‐CK marker in the preceding figure.

FIGURE 6.

Multiplex immunofluorescence images. The images demonstrate CD45 (gold) and Ki‐67 (white) staining in peripheral tumor at 28 days. DAPI (blue) was used as a nuclear stain. The co‐localization of Ki‐67+ and CD45+ cells (>90%) demonstrates the marked proliferation of leukocytes. Scale bar: 100 μm

3.6. Quantification of cellular composition using immunofluorescence

The relative contributions of various marker populations to the total population over time were determined using immunofluorescence of the tissue microarray. The number of CD31+ cells did not change substantially over time as shown in Figure 7. SMA+ cells, CD45+ cells (leukocytes), vimentin+ cells, and pan‐CK+ cells showed varying degrees of increase over the course of the study, most notably with the preponderance of CD45+ cells in the 28‐day cohort. Bar graphs for each individual population over time with statistical indicators are provided in Figure S3 in Supporting Information.

FIGURE 7.

Individual cell population contributions to the overall composition of peripheral and central tumor at 14, 21, and 28 days from the TMA.The relative contribution of CD45+ cells increased over time. The “Other” category comprises primarily hepatocytes.

4. DISCUSSION

The vascularity of the target tumor is a fundamental requirement for the delivery of transarterial therapy. In the present study, CT demonstrated modest contrast enhancement around the lesions but a lack of enhancement within the lesions, which was also noted in a previous study of liver cancer using the Oncopig model ¹⁰ ; thus, the lesions could not be characterized as hypervascular. The paucity of neovascularity likely limits the penetration of embolic and therapeutic agents into lesions.

Immunohistochemistry and multiplex immunofluorescence showed that the lesions had a heterogeneous inflammatory cell composition. No previous studies using the Oncopig reported an assessment of CD45 expression, yet this pan‐leukocyte marker is important for interpreting the histologic findings. ¹⁷ In our study, histologic lesions were characterized by mixed inflammation with a strong lymphohistiocytic component. Strong colocalization of CD45 staining with Ki‐67 staining (over 90% of Ki‐67+ cells co‐expressed CD45) was consistent with proliferating leukocytes and an inflammatory response; neoplastic features were not observed. We also observed fibroblasts, and myofibroblasts consistent with wound healing and an inflammatory lesion.

The results were similar to prior reports in a broad sense, with Iba‐1, CD3, and vimentin positive staining. Likewise, consistent with findings from other investigators, Langhan's type multinucleated giant cells were frequently observed; these cells were Iba‐1+ and this histologic feature is typical of a granulomatous response. The relatively low amount of CD31+ cells is consistent with the CT findings, particularly in the central lesion, where no contrast enhancement was observed. Cancer‐associated fibroblasts (CAF), which differ significantly from other fibroblasts, impact the tumor microenvironment through immunosuppression. A key biomarker of CAFs is SMA, ¹⁸ whose expression remained low throughout the present study. Overall, the lesions' proliferating leukocytes, fibroblasts, and myofibroblasts are reminiscent of a group of benign entities called inflammatory pseudotumors. These resemble carcinomas clinically and on diagnostic imaging but are not malignant.

The in‐situ induction method employed in the present study was reported to have a success rate equivalent to that of the more complex cell line method, with no histologic difference in the resulting tumors. ¹⁰ Our overall success rate of 78% for generating tumors was in keeping with this and other prior work. Of note, the cell line method for inducing liver tumors was reported to be unsuccessful unless the animals also had cirrhosis induced by transarterial ethanol treatment. ¹⁹ Such sensitivity calls the utility of the model into question. In the present study, we found no evidence of carcinomas, a finding that was also noted in 4 of 7 tumors in a recent study of lung tumors in the Oncopig model. ¹² One previously proposed explanation for the lack of carcinomas is that a successful immune response controlled the neoplastic growth. This begs the question of what is actually being treated in a given experiment. The immune response in the model also lacks a corresponding clinical scenario of an overwhelming immune response to a solid tumor that consists entirely, or nearly entirely, of immune cells. In addition, we found that lesion regression was prevalent, as noted in previous studies. ¹² , ¹³ , ²⁰ , ²¹

To our knowledge, no previous studies using the Oncopig model analyze lesion growth kinetics in depth. Growth kinetics data provide essential context for interpreting findings in the Oncopig model given the disparity between long clinical experience and what is found in the model. Cell population dynamics in cancer have been under scrutiny since Collins first proposed the idea of exponential growth in 1956. ²² , ²³ Historically, aggressive cancers such as some Ewing sarcomas and non‐Hodgkin lymphomas, have doubling times under 30 days. ²⁴ This time frame serves as a benchmark, as the mean doubling time for less aggressive cancers is considerably longer. For example, the mean doubling time is 99 days for head and neck cancer, 167 days for breast cancer, 204 days for HCC, and 302 days for cervical cancer. ²⁵ , ²⁶ , ²⁷ , ²⁸ Based on our own experience ²¹ and previous reports of Oncopig lesion growth, ¹⁰ we calculated growth using the volume of the core biopsy sample as an idealized starting point. In approximately 30 days the inoculum volume would be expected to double from 12.7 μL to 25 μL. The reported lesion volumes of 4.8–33 mL at 7 days ¹⁰ require that the doubling times in this model must be 14–42 h (see Figure S4 in Supporting Information for calculations). In the example of a 4 cm tumor at 7 days, this represents a 1300‐fold discrepancy, contrasting starkly with clinical experience. In addition, it is known that aggressive tumors with short doubling times (<30 days) also have a high frequency of mitotic figures; for example, Ewing sarcoma can have up to 42 mitotic figures per 10 high‐power fields, ²⁹ and non‐Hodgkin lymphoma can have up to 49 mitotic figures per 10 high‐power fields. ³⁰ Thus, an abundance of mitotic figures in the induced tumors would be predicted. However, despite doubling times that are in the range of 30‐fold shorter in this model than in aggressive tumors, no studies of Oncopig liver tumors have reported or discussed mitotic figure data. Histopathology and advanced staining and imaging showed mixed inflammation and no typical criteria of malignancy.

The present study had several potential limitations, including the small number of animals and the lack of a control group of wild type animals treated with the AdCre vector. However, in a study of pancreatic cancer using the Oncopig, a vector‐treated wild type control group showed no evidence of pancreatitis or phlegmon, ¹¹ which suggests that viral infection per se does not cause inflammation. From a technical standpoint, including a control group treated with the empty vector (i.e., AdCre alone, without the transposition payload) or only Gelfoam (without the vector), which would have no general response to AdCre exposure, could increase the study's rigor. Finally, the study was conducted in male pigs. Although no a priori difference would be predicted, no conclusions can be drawn in the present work as to the potential impact of sex on the outcomes.

5. CONCLUSIONS

The development of a realistic and robust large animal model for testing new HCC treatments remains a challenging goal. In the present study, hepatic tumors were consistently produced in the Oncopig model but were dominated by an inflammatory component and characterized by rapid growth followed by regression and resorption on CT. Infiltrates were predominantly composed of macrophages, lymphocytes and myofibroblasts, with relatively fewer neutrophils and B cells; there was no clear evidence of carcinoma in any section examined. Mutation in KRAS is not common in HCC ⁸ ; thus, the relevance of the Oncopig model for mechanistic studies combining local interventions and systemic therapies for HCC will likely be called into question. Furthermore, the tumor kinetics of rapid growth and short doubling time we observed, irrespective of the resorption we observed at 28 days, were far outside the range of clinical experience for HCC and even the most aggressive cancers. These findings suggest that, in its current iteration, the Oncopig model may not be suitable for cancer biology studies investigating a specific signaling pathway, other than inflammation, in HCC; rather, the model may be better suited for liver cancer research that requires the presence of an in vivo mass, such as robotic surgery studies, imaging software development, and training programs. In addition, our findings regarding the transient nature of the lesions have implications for selecting endpoints in studies using the Oncopig. Existing data, particularly the resorption, indicate that the model may not be suitable for longitudinal studies with endpoints such as overall survival. Studies whose key endpoints are changes in volume or contrast enhancement as a consequence of intervention may be especially vulnerable to confounding due to resorption. Researchers must consider these factors if they are contemplating the use of the Oncopig model to evaluate the efficacy of new embolic materials or ablative devices. Additional work is needed to clarify the nature and origin of cells in the lesions, including whether they are resident or recruited. Experiments in which tumor cells are subjected to flow cytometry or peripheral immune cell populations are labeled prior to inoculation will elucidate the nature of the prominent inflammatory response in the model.

AUTHOR CONTRIBUTIONS

Erik N. K. Cressman: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; supervision; validation; visualization; writing – original draft; writing – review and editing. Samantha Hicks: Data curation; formal analysis; investigation; methodology; writing – original draft. Natalie W. Fowlkes: Data curation; formal analysis; methodology; visualization. Danielle L. Stolley: Data curation; investigation; methodology; project administration; visualization. Maria Sophia Stenkamp: Formal analysis; investigation; methodology; writing – original draft.

FUNDING INFORMATION

This study was supported by Institutional Research Grant, MD Anderson Cancer Center. UPWARDS Training Program (Undergraduate Students Working Towards Research in Science), grant/award number: 1R25CA240137‐01A1; the CPRIT Research Training Award CPRIT Training Program, grant/award number: RP210028.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to disclose.

ETHICS STATEMENT

Ethics Statement Ethics approval was granted by the MD Anderson Institutional Animal Care and Use Committee. (Protocol number: 0002158)

Supporting information

Data S1:

Data S2:

Data S3:

Data S4:

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.