Fenretinide, Troglitazone, and Elmiron Add to Weight of Evidence Support for Hemangiosarcoma Mode-of-Action From Studies in Mice
ABSTRACT
Pharmaceuticals and chemicals can induce hemangiosarcomas (HS) in mice, often through mechanisms that involve cell proliferation but do not directly damage DNA. A proposed mode-of-action (MOA) for hemangiosarcoma, developed at an international workshop, outlines five key elements: hypoxia, macrophage activation, increased angiogenic growth factors, dysregulated angiogenesis/erythropoiesis, and endothelial cell proliferation. The current study aimed to strengthen the evidence supporting this MOA by examining these key elements in mice treated with three compounds known to induce HS with varying potency: fenretinide (high potency), troglitazone (intermediate potency), and elmiron (low potency). Multiple assessments were conducted in B6C3F1 mice after 2, 4, and 13 weeks of treatment, including evaluations of hypoxia (using hypoxyprobe and transcriptomics), endothelial cell proliferation, and clinical and anatomic pathology. All three compounds showed strong evidence of dysregulated erythropoiesis, indicated by a decrease in red blood cells and a failure to increase reticulocytes, as well as macrophage activation, with increases ranging from 4- to 11-fold. This pattern of hematological changes in mice could potentially serve as an early indicator for evaluating endothelial cell proliferation in organs suspected of developing HS. Fenretinide demonstrated all five key elements of the proposed MOA, while troglitazone showed four elements, and elmiron demonstrated three. Transcriptomic analysis supported the five elements of the MOA but was not more sensitive than hypoxyprobe immunohistochemistry in detecting hypoxia. The overall transcriptional evidence for the key elements of the proposed MOA also aligned with the known potency of these compounds for HS induction. These findings, combined with previous research on 2-butoxyethanol and pregabalin, enhance the weight-of-evidence supporting the proposed MOA for HS formation.
INTRODUCTION
Hemangiosarcomas (HS) are tumors originating from endothelial cells (ECs) and are characterized by poorly differentiated, proliferating ECs. These tumors are induced by a variety of chemicals in rodent bioassays, particularly in mice, and have become an increasing concern for regulatory agencies, sometimes leading to delays in pharmaceutical approval. HS can occur spontaneously and in response to numerous different compounds in mice, but are not observed in rats and are rare in humans. To the best of our knowledge, there are only two instances where a chemical stimulus has been shown to induce hemangiosarcoma in both humans and rodents (mice or rats). Both of these examples involve genotoxic carcinogens (vinyl halides and Thorotrast) and primarily result in liver hemangiosarcoma. In contrast, commercial pharmaceutical products and chemicals that induce HS solely in mice appear to operate through non-genotoxic mechanisms involving cell proliferation. The pharmacology of these compounds is diverse, encompassing calcium channel blockers, antipsychotics, phosphodiesterase-5 inhibitors, dipeptidyl peptidase-4 inhibitors, antiarrhythmics, gonadotropin receptor antagonists, antisense compounds, nitric oxide releasers, hemolytic compounds, and vascular endothelial growth factor (VEGF) inducers.
In 2009, a mode-of-action (MOA) for hemangiosarcoma was proposed based on information presented at an international workshop. Emerging data suggested that these diverse compounds share a common point of intersection and interaction with growth regulatory pathways. Specifically, agents that induce hemangiosarcoma appear to have initiating events that lead to local tissue hypoxia and macrophage activation. These two changes subsequently increase angiogenic growth factors. While analogous to physiologic angiogenesis, it was hypothesized that agents inducing hemangiosarcoma produce dysregulated angiogenesis and/or erythropoiesis. Dysregulated angiogenesis implies that the hypoxia cannot be compensated by the formation of new blood vessels or that the stimulated ECs do not form functional blood vessels, such as the blind vessels commonly observed in hemangiosarcoma in the mouse liver. Dysregulated erythropoiesis in the bone marrow and spleen, accompanied by an increase in macrophages, suggests a failure to produce functional erythrocytes in these organs, contributing to sustained tissue hypoxia.
The components of the proposed MOA for hemangiosarcoma have been extensively investigated using 2-butoxyethanol (2-BE) and pregabalin, both of which induce hemangiosarcoma in mice but not rats. 2-BE primarily induces hemangiosarcoma in the liver, while pregabalin induces it in the bone marrow, liver, and spleen. Both compounds induce local tissue hypoxia through different mechanisms: 2-BE via the induction of hemolysis, and pregabalin presumably through its effects on systemic pH mediated by an increase in bicarbonate. Pregabalin and 2-BE have also been shown to activate macrophages, which may play a contributing role in this MOA by increasing angiogenic cytokines such as interleukin (IL)-6 and increasing reactive oxygen species, which can also contribute to deoxyribonucleic acid (DNA) damage in ECs. Thus, divergent initiating events can converge on hypoxia and macrophage activation.
Immunohistochemical staining of cell markers has been employed to understand the pathogenesis of hemangiosarcoma and to gain insight into the human relevance of these tumors. For instance, ECs in hemangiosarcoma may arise from the transformation of differentiated ECs in the target organ or from the recruitment of transformed bone marrow-derived stem cells or endothelial progenitor cells (EPCs). In mice, both spontaneous and chemically induced hemangiomas and HS appear to originate from an EPC lineage based on their expression of CD34, VEGFR2, and CD31, but not factor VIII antigens. In humans, hemangiomas appear to be derived from late EPCs or differentiated ECs, while HS are derived from bone marrow-derived hematopoietic stem cells or early EPCs based on their expression of CD117, CD34, and CD45 antigens. Unlike in mice, hemangiomas and HS in humans appear to have distinct pathogenesis, supporting the conclusion that human hemangiomas do not progress to HS. Therefore, cell marker analysis has not provided definitive evidence to conclude whether this tumor type is relevant to humans. However, several lines of evidence suggest why mice are more susceptible than rats or humans to the induction of hemangiosarcoma by nongenotoxic agents, including a higher spontaneous incidence in mice, a correlation between spontaneous EC proliferation rates and spontaneous incidence across species, the ability of agents like troglitazone to increase EC proliferation in mouse but not human cells, human disease states suggesting greater resistance in humans, and lower antioxidant levels in mice compared to rats and humans, potentially contributing to enhanced susceptibility.
To further test the components of this MOA framework, three compounds with varying potency and mechanisms of action were selected to increase the weight-of-evidence: elmiron (sodium pentosan polysulfate, a bladder pain analgesic), troglitazone (a peroxisome proliferator-activated receptor gamma [PPARγ] agonist), and fenretinide (a retinoic acid receptor [RAR] agonist). These compounds exhibit low to high potency for inducing HS in mice, but not rats, respectively. Troglitazone and fenretinide, but not elmiron, have been shown to produce hemangiosarcoma in multiple organs with multiple tumors within a single animal. Elmiron has the lowest potency, producing hemangiosarcoma only in the liver without tumor multiplicity. Troglitazone has intermediate potency, inducing hemangiosarcoma in the bone marrow, brown adipose/subcutis fat, liver, and spleen. Fenretinide is the most potent agent, inducing hemangiosarcoma within one year of treatment and an almost 100% incidence by two years. In male mice, fenretinide induces hemangiosarcoma in the hematolymphoreticular system (bone marrow, liver, lymph node, spleen), epididymis, epididymal white fat pads, and subcutis fat. Based on the tumor responses observed with these three compounds, male mice were chosen for a time-course study of these initiating events, as they appeared to exhibit a higher response rate than females. A battery of endpoints was assessed after 2, 4, and 13 weeks of treatment, including hypoxia (using hypoxyprobe and transcriptomics), EC proliferation, and clinical and anatomic pathology. Additionally, the methods previously applied in studying 2-BE and pregabalin were evaluated for their utility when applied to weaker agents such as troglitazone and elmiron.
MATERIALS AND METHODS
Animals
Male B6C3F1 mice were obtained from Charles River Laboratories (Kingston, New York). All mice used in this study were housed in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International and were used in accordance with an approved animal care and use protocol, the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals (National Research Council, 1996), and all applicable regulations.
Study Design and Dose Rationale
This work was conducted as three separate studies: 10GR145 (fenretinide), 10GR202 (troglitazone), and 10GR221 (elmiron). Prior to the 13-week studies, one-month pilot studies were performed with each compound using the same dose levels as reported, assessing body weight, food consumption, clinical pathology, organ weights (liver, brown adipose tissue, white adipose tissue), and histopathology (data not reported). The pilot studies showed no adverse findings and confirmed adequate target organ responses (hypoxia, cell proliferation, and transcriptomics). Fenretinide (N-[4-Hydroxyphenyl] retinamide) is a retinoic acid receptor agonist. Carcinogenicity studies in B6C3F1 mice have demonstrated that it induces hemangiosarcoma in the subcutis and abdominal soft tissue with an incidence approaching 100% after two years at doses of 50 and 250 mg/kg (Table 1). Mechanistically, fenretinide is thought to bind to retinoic acid receptors in vitro, regulating cell growth and angiogenesis possibly through retinoic acid receptor-independent pathways. It has also been shown to induce endothelial cell apoptosis via increased production of ceramide through unknown mechanisms. Based on the one-month pilot study showing no adverse findings, a dose of 500 mg/kg fenretinide administered via diet for up to 13 weeks was chosen to investigate its effects on endothelial cell proliferation and hypoxia status in male B6C3F1 mice. In the one-month pilot study, the 500 mg/kg fenretinide dose resulted in blood levels of 2846 and 2188 ng/ml on study days 15 and 29, respectively.
Troglitazone, a thiazolidinedione and a known peroxisome proliferator-activated receptor gamma agonist, has been shown to increase the incidence of hemangiosarcoma in B6C3F1 mice at doses of 400 and 800 mg/kg (Table 1). It has also been shown to increase endothelial cell proliferation in numerous tissues, including the heart, brown fat, and white fat. A dose of 1200 mg/kg was chosen to investigate the effects of troglitazone on endothelial cell proliferation and hypoxia status in male B6C3F1 mice when administered orally for up to 13 weeks, based on the pilot study and previous work that also used this dose. A previous four-week study in B6C3F1 mice reported area under the curve values of 5.3 and 31 mg.h/ml at doses of 1000 and 5000 mg/kg, respectively, and confirmed no significant toxicity. In humans, an 800 mg troglitazone dose results in an area under the curve value of 5.6 mg.h/ml and a maximum concentration value of 3.24 mg/ml.
Elmiron (xylan hydrogen polysulfate, sodium pentosan polysulfate) is a highly sulfated, semi-synthetic pentose polysaccharide derived from beechwood shavings. It shares properties with heparin and exhibits antilipidemic, marked anticoagulant, fibrinolytic, antiflogistic, and mild antihypertensive effects, as well as inhibiting vascular smooth muscle cell proliferation. In the USA, elmiron is used to alleviate pain from interstitial cystitis, with its mechanism of action attributed to its preferential localization in the lining of the pelvis, ureters, and bladder of the urinary tract. Elmiron has also been shown to increase hemangiosarcoma in the liver of B6C3F1 mice treated with 168 and 504 mg/kg (Table 1). A dose of 1000 mg/kg was chosen to investigate the effects of elmiron on endothelial cell proliferation and hypoxia status in male B6C3F1 mice when administered orally for up to 13 weeks, based on previous work that used this dose without significant toxicity for 13 weeks and confirmed by our one-month pilot study. In humans, a 400 mg elmiron dose results in an area under the curve value of 1178 ng.h/ml and a maximum concentration value of 54 ng/ml; toxicokinetic data from mice is unavailable.
For all three compounds, a time course study of 2 (transcriptomics only), 15, 29, and 92 days was designed to determine when proliferative and/or hypoxic changes occur in the tissues where hemangiosarcomas develop. A roughly two-fold higher dose than the high doses used in carcinogenicity studies was employed to maximize the potential to measure these changes. Since the one-month pilot studies did not result in any significant findings (body weight, food consumption, clinical signs, or clinical pathology) considered to be at a maximum tolerated dose, higher doses than those used in carcinogenicity studies were chosen to maximize the ability to detect the measured changes, consistent with the concentration-time relationship where no observable effect levels generally decrease as the duration of exposure increases.
Hematology
Blood samples containing EDTA as an anticoagulant were collected by cardiac puncture from 10 untreated control mice or 10 mice receiving fenretinide, troglitazone, or elmiron on days 15, 29, and 92. The following hematology parameters were evaluated using a Siemens Advia 120 automated hematology analyzer (Tarrytown, New York): white cell count, red cell count, hematocrit, hemoglobin, mean corpuscular volume, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, red cell distribution width, platelet count, mean platelet volume, white blood cell differential, and reticulocytes. In addition, large unstained cells were measured on this analyzer. Only parameters that demonstrated statistical significance at least once per compound tested are presented.
Bone Marrow Collection
Bone marrow samples were collected from all mice on days 15, 29, and 92 by flushing a single femur with bovine serum albumin. Total nucleated cell count was determined in all samples using a Siemens Advia automated hematology analyzer (Tarrytown, New York). The opposing femur, as well as the spleen and liver, were harvested from each animal for histopathology. Total nucleated cell count provides a quantitative measurement of the total number of cells per femur. The following procedures were also performed on bone marrow samples as previously described: preparation of cytocentrifuge slides for microscopic determination of bone marrow differentials, enumeration of total myeloid and total erythroid numbers, enumeration of bone marrow macrophages, and calculation of the myeloid to erythroid ratio. A total of 500 cells were counted using a microscope to calculate myeloid and erythroid cell numbers. The opposing femur, spleen, and liver were harvested from each animal for histopathology.
Histology
Mice were not fasted prior to necropsy. Animals and groups were randomized using a computer program before necropsy, which was initiated at approximately 8 AM for each time point. Mice were euthanized by one of two procedures: isoflurane gas anesthesia followed by exsanguination for endothelial cell proliferation endpoints, or by cervical dislocation under ketamine/acepromazine anesthesia (75/2.5 mg/kg, respectively) for hypoxia endpoints. All procedures used in these studies were reviewed and approved by the Institutional Animal Care and Use Committee. At necropsy, the following tissues were collected, weighed, and fixed in 10% neutral buffered formalin: liver, spleen, brown scapular adipose tissue, and white epididymal adipose tissue. Additionally, femurs were collected, fixed in 10% neutral buffered formalin, and then decalcified by immersion in ImmunoCal (Decal Corporation, Tallman, New York). All tissues were processed to paraffin blocks according to routine histological procedures. The slides were reviewed by a pathologist in an unblinded manner.
Hematoxylin and Eosin Staining
Five-micron sections were cut and mounted on glass slides. All Hematoxylin and Eosin staining was completed using the Ventana Symphony (Ventana Medical Systems, Tucson, Arizona) slide stainer using the system’s reagents according to the manufacturer’s directions.
Immunohistochemistry
Five-micron sections were cut and mounted onto charged slides and used for the following immunohistochemical procedures:
Hypoxyprobe Immunohistochemistry. To expose the antigen recognized by Hypoxyprobe (HPI, Burlington, Massachusetts), sections were pretreated with a Citrate Buffer (Chemicon, Temecula, California) using a Decloaking Chamber from Biocare Medical, Walnut Grove, California. All Hypoxyprobe immunohistochemical detections were performed on an automated Dako Autostainer (Dako, Carpinteria, California). Endogenous peroxidase was blocked by incubation in 3.0% hydrogen peroxide (aq) for 10 minutes. Dako Protein block was applied for 20 minutes to block nonspecific reactions. The Hypoxyprobe reagent was applied at a dilution of 1/500 for 60 minutes at room temperature and detected using Dako Envision+ Rabbit HRP Polymer for 30 minutes, visualized with Dako Liquid diaminobenzidine (DAB+). All slides were then counterstained for nuclear detail using Mayer’s hematoxylin, dehydrated, cleared, and mounted with a permanent mounting media. Negative controls included a rabbit isotype control antibody (Vector, Burlingame, California) incubated for 60 minutes using the same detection system. Additionally, tissues from mice exposed to hypoxia, as well as those from mice maintained in a normal oxygen environment, were run in parallel as further positive and negative tissue controls for detecting local tissue hypoxia.
Endothelial Cell Proliferation Immunohistochemistry. To expose antigenic sites, sections were pretreated with Borg Retrieval Solution (Biocare Medical, Walnut Grove, California) using a Decloaking Chamber from Biocare Medical. All CD31 and bromodeoxyuridine (BrdU) immunohistochemical detections were performed on an automated Dako Autostainer. Endogenous peroxidase was blocked by incubation in 3.0% hydrogen peroxide (aq) for 10 minutes. Dako Protein block was applied for 20 minutes to block nonspecific reactions. Goat antiCD31 antibody (Santa Cruz Biotechnology, Santa Cruz, California) was applied at a dilution of 1/1000 for 60 minutes at room temperature and detected using a peroxidase-conjugated donkey anti-goat IgG secondary antibody at 1/250 for 45 minutes (Jackson Immunoresearch, West Grove, Pennsylvania), followed by a 10-minute incubation in a 1/50 dilution of Cy5-TSA (Perkin Elmer, Waltham, Massachusetts). Next, a sheep antiBrdU antibody (Novus Biologicals, Littleton, Colorado) was applied at a dilution of 1/300 for 60 minutes at room temperature and detected using an Alexa fluor 488 donkey anti-sheep IgG secondary antibody at 1/300 for 45 minutes (Invitrogen, Carlsbad, CA). All slides were then counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen, Carlsbad, California) for 10 minutes, washed in PBS followed by distilled water, and coverslipped with Prolong Gold Antifade mounting media (Invitrogen, Carlsbad, California). Negative controls included goat and sheep isotype control antibodies run at the same concentrations as CD31 and BrdU, respectively (Jackson Immunoresearch).
Immunohistochemistry for Oxidative Stress and Mitochondrial Injury. To expose antigenic sites, sections were pretreated with Borg Retrieval Solution or EDTA (Invitrogen, Carlsbad, California) for manganese superoxide dismutase (mnSOD) and voltage-dependent anion channel 1 (VDAC-1 or porin), respectively, using a Decloaking Chamber. All mnSOD and porin immunohistochemical detections were performed on an automated Dako Autostainer. Endogenous peroxidase was blocked by incubation with 3.0% hydrogen peroxide (aq) for 10 minutes, followed by a 20-minute incubation with Dako Protein Block to block nonspecific binding. Rabbit anti-mnSOD antibody (Thermo Scientific, Kalamazoo, Michigan) and rabbit anti-porin antibody (Epitomics, Burlingame, California) were applied at dilutions of 1/100 and 1/750 for 1 hour at room temperature. Both antibodies were detected using Dako Envision+ Rabbit HRP Polymer for 30 minutes and visualized using Dako Liquid DAB+. All slides were then counterstained for nuclear detail using Mayer’s hematoxylin, dehydrated, cleared, and mounted with a permanent mounting media. Negative controls included a rabbit isotype control antibody (Vector, Burlingame, California) incubated for 60 minutes using the same detection system.
Cell Proliferation
For the elmiron, fenretinide, and troglitazone studies, paraffin sections were cut from the liver, subscapular brown fat, and epididymal white fat. Tissues were stained as described earlier. Proliferating endothelial cells were measured using the iCyte LSC. The approach was conducted as previously described. A mosaic scan of the entire tissue section was created at 20× objective and 20 micrometer step size to create a low resolution tissue map. From this map ten 760 × 568 micrometer scan areas were randomly distributed within the tissue, which were then subjected to high-resolution image capture using a 40× objective and 0.25 micrometer step size. The total area sampled at high resolution for each liver tissue section was 4.32 square millimeters. A 405 nanometer laser was used for 4′,6-diamidino-2-phenylindole excitation, while Alexa Fluor 488 and Cy5 were excited with 488 and 633 nanometer lasers, respectively. Contour thresholds were automatically optimized by the software for each fluorophore based on intensity. Similar to flow cytometry, scattergrams of contoured nuclei, proliferating nuclei, and CD31 positive areas were used to identify density and location of these events within the liver section. Nuclei within CD31 positive regions were characterized as endothelial cell nuclei, and those nuclei positive for bromodeoxyuridine labeling were measured as proliferating cells. Co-localized endothelial cells and proliferating cells were characterized as proliferating endothelial cells. From these associations, the following measurements are reported: total nuclear count, relative number of CD31 positive cells, absolute number of CD31 positive cells, relative number of bromodeoxyuridine positive cells, absolute number of bromodeoxyuridine positive cells, relative number of CD31 positive and bromodeoxyuridine positive cells, and absolute number of CD31 positive and bromodeoxyuridine positive cells. Cell counts from each animal were obtained from the same total tissue area of 4.32 square millimeters as previously described. Briefly, proliferating cells were identified by nuclear colocalization of bromodeoxyuridine and 4′,6-diamidino-2-phenylindole signal, while endothelial cells were identified as those 4′,6-diamidino-2-phenylindole events within CD31 positive regions. Those CD31 positive nuclei that were also bromodeoxyuridine positive were categorized as proliferating endothelial cells. Raw counts of these events were obtained from the LSC and were validated by viewing images of the events falling within the above groups. When necessary, data points representing staining artifacts were removed through careful deselection of events using strict intensity, shape, and size parameters. Raw data was exported into a Microsoft Excel spreadsheet where percent values, means, and standard deviations were tabulated. Statistical analyses were performed separately for each tissue, day, and endpoint. The measurements for tissue, day, and endpoint were transformed to normal scores, and the transformed data were then analyzed using a 2 sample t-test. Results were reported at the 0.05 or 0.01 levels of significance.
Assessment of Transcriptomic Effects
Ribonucleic acid isolation, transcriptional profiling, and quantification of expression changes were conducted as described previously. Briefly, total ribonucleic acid was extracted from the liver, subscapular brown fat, and epididymal white fat, and complimentary deoxyribonucleic acid was synthesized. Complimentary ribonucleic acid was transcribed from complimentary deoxyribonucleic acid using the GeneChip One-Cycle Labeling Kit from Affymetrix, hybridized to Mouse M430 v2.0 GeneChip microarrays, and analyzed using the “affy” and “limma” packages of the Bioconductor suite of microarray analysis tools available for the R statistical environment. An analysis of variance model was used for the statistical analysis. Probe sets were considered to have changed qualitatively in a specific comparison if a false discovery rate-adjusted p-value of 0.2 was obtained, their average expression intensity was above 100 in either treatment group, and they had an absolute fold change greater than 1.3. Genes represented by multiple probe sets were considered to have changed if at least 1 probe set was observed to change. Gene expression changes that met these criteria were called “statistically significant differentially expressed genes” and have the directional qualities of “up” or “down”, meaning they can be upregulated or downregulated in response to treatment.
Causal Reasoning and Text-Mining Analysis
The Causal Reasoning Engine is a computational platform to identify hypothetical upstream controllers from differentially expressed genes observed in a specific experiment. The method relies on a large knowledgebase of biological concepts and entities and their causal relationships. Relationships are manually curated from peer-reviewed scientific literature and public and proprietary databases. Inferred upstream controllers are called “hypotheses” as they are statistically significant potential explanations of the gene expression changes. For this study, hypotheses related to hypoxia, specifically response to hypoxia and the transcriptional activities of hypoxia-inducible factor 1 alpha and endothelial PAS domain protein 1, inflammation, and cell proliferation were primarily evaluated. These hypotheses were defined in previous work with reference compounds/treatments. The Causal Reasoning Engine provides enrichment p-values to assess deviations from a null model of random assignment of differentially expressed genes to the knowledgebase. We considered p-values less than 1e-4 as significant and inspected hypotheses with p-values less than 0.05 in cases of biological interest. In addition, we conducted a literature search for molecular entities related to “hemangiosarcoma” using the Pfizer-internal LitMS textmining platform. LitMS regularly scans all available PubMed abstracts for co-occurrence of relevant terms, specifically in our use case for synonyms of “hemangiosarcoma” and all known gene and protein names. Internally, LitMS uses the Apache Lucene information retrieval library to perform dictionary look-ups and extract contextual and positional features from the text, which are then passed to one of several scoring algorithms to assess relevance to specific topics of interest. Finally, we utilized Medical Subject Heading clouds for interpreting Causal Reasoning Engine results as described previously. Briefly, each relationship in the knowledgebase is linked to one or more scientific articles. If articles have been manually annotated with Medical Subject Heading terms in the PubMed database, it becomes possible to assess Medical Subject Heading terms, such as “leucocytes”, “DNA damage”, or “inflammation”, for their enrichment in relationships supporting a specific hypothesis.
RESULTS
Organ Weights
Body weight changes were only observed on day 29; fenretinide-treated mice had a slight decrease in body weight (4.58%) whereas troglitazone-treated mice had a slight increase in body weight (1.07-fold) (Table 2). Both weight changes were transient and returned to normal values by day 92. Elmiron did not significantly affect body weight. Increases in absolute and relative liver weight were observed for all 3 compounds variably across the time points. Fenretinide-treated mice had statistically significant increased liver weight for all time points (1.16- to 1.37-fold absolute, 1.2- to 1.39-fold relative), whereas troglitazone-treated mice had increased liver weight at day 15 (1.11-fold absolute, 1.0-fold relative) and day 29 (1.11-fold absolute), but there was also a trend for increased liver weight at day 92 (1.04-fold absolute). Elmiron-treated mice had statistically significant increased liver weight at day 15 (1.11-fold absolute, 1.06-fold relative) and day 92 (1.22-fold absolute, 1.22-fold relative) with a trend for increased liver weight at day 29 (1.07-fold absolute). Increases in absolute and relative spleen weight were observed in fenretinide and elmiron-treated mice. Fenretinide-treated mice had statistically significant increased spleen weight at all time points (1.12- to 1.39-fold absolute, 1.14- to 1.42-fold relative), whereas elmiron-treated mice had increased spleen weight only at day 92 (1.16-fold absolute, 1.15-fold relative) with a trend for increased spleen weight at day 29 (1.13-fold absolute). In contrast, troglitazone-treated mice had statistically significant decreased spleen weight at day 92 (19.74% absolute, 25.2% relative). Statistically significant decreases in absolute and relative white adipose tissue weight were observed for fenretinide-treated mice at all time points (60.26%–74.86% absolute, 59.54%–73.99% relative), but decreases in brown adipose tissue weight were observed on day 15 (16.74% absolute, 14.59% relative) and day 29 (21.9% absolute) with a trend for decreased brown adipose tissue weight at day 92 (12.51% absolute). Troglitazone and elmiron-treated mice did not have biologically significant organ weight changes in white or brown adipose tissue.
Hematology and Bone Marrow—Effects of Fenretinide
Fenretinide induced multiple effects on red blood cell parameters that were present within 15 days of dosing and increased in intensity during the 92 days of the study (Table 3). The red blood cell count in fenretinide-treated mice were 5.7%, 9.5%, and 9.8% lower than controls on days 15, 29, and 92, respectively. Decreases in hemoglobin concentrations were even more pronounced, reaching 17% lower than controls by day 92. Reticulocyte counts were 17% lower in treated mice compared with controls on day 15 and failed to show a significant increase on either days 29 or 92 despite decreases in red blood cell count and hemoglobin. Red cell distribution width showed a progressive increase over time, indicating increased variability in red cell size.
Fenretinide treatment also induced effects on multiple white blood cell and platelet parameters that were present by 15 days of dosing and persisted through the 3 months of the study (Table 4). Fenretinide produced 2- to 3-fold increases in white blood cell count with elevations in peripheral blood neutrophils, lymphocytes, and monocytes at all 3 time points. Large unstained cells are a unique cell type identified only with the Advia hematology analyzer. Large unstained cells indicate cells that are peroxidase negative but larger than normal lymphocytes. Increases in large unstained cells are associated with the presence of blasts, immature cells, or cells that possess both monocyte and lymphocyte features. Fenretinide induced approximately 4-, 5-, and 6-fold increases in large unstained cells on days 15, 29, and 92. Platelet counts were significantly decreased from 16% to 39% over the 3 assessment days.
Fenretinide decreased bone marrow total nucleated cell count by 16% on day 29 and 31% on day 92 (Table 5). The myeloid to erythroid ratio was increased at all 3 time points due to a decrease in total erythroid cells. Decreases in total erythroid cells were 23% lower than untreated controls on day 15 and reached 55% lower than controls by day 92. Bone marrow macrophage numbers were increased by approximately 6-fold on days 15 and 29 and nearly 11-fold by day 92 following fenretinide treatment.
Hematology and Bone Marrow—Effects of Troglitazone
Troglitazone produced small, but statistically different, decreases in red blood cell count and hemoglobin that remained less than 4% compared with the controls but was not different from controls by day 92. Reticulocyte counts were decreased by 14% and 29% in troglitazone-treated mice on days 29 and 92, respectively (Table 3). Troglitazone decreased white blood cell count by 50% by day 92, largely due to decreases in lymphocytes (Table 4). Troglitazone produced no effects on large unstained cells or platelet count (Table 4).
Troglitazone produced statistically different decreases on bone marrow myeloid to erythroid ratio on days 29 and 92. However, the effects were small (approximately 25%). The only persistent change in bone marrow was an increase in macrophage number of 4-, 7-, and 4-fold with troglitazone treatment on days 15, 29, and 92, respectively (Table 5).
Hematology and Bone Marrow—Effects of Elmiron
Elmiron had very limited effects on red blood cell parameters, with the most notable change being a decrease in reticulocyte counts of 16%, 9%, and 21% on days 15, 29, and 92, respectively (Table 3). Neutrophil, monocyte, and platelet counts showed minimal changes in elmiron-treated mice, but they were not persistent over time. Elmiron did produce a significant 3-fold increase in large unstained cells by day 92 (Table 4).
The most persistent changes induced by elmiron were decreases in bone marrow myeloid to erythroid ratio produced by increases in total erythroid cells. Total macrophage numbers in bone marrow were also increased by approximately 4-, 7-, and 8-fold on days 15, 29, and 92, respectively (Table 5).
Histopathology
Hepatocyte hypertrophy was observed at all time points for fenretinide and troglitazone-treated mice, which would correlate with increased organ weight findings (Table 6). The hepatocyte hypertrophy was primarily centrilobular for both compounds. The severity of the hepatocyte hypertrophy increased over time in fenretinide-treated mice, and single-cell hepatocyte necrosis was also observed at days 29 and 92. Elmiron-treated mice had hepatocyte vacuolation and mixed cell infiltrates observed at day 92; the vacuolation may have contributed to the increased organ weight finding at that time point. Extramedullary hematopoiesis was observed in the spleen of fenretinide and elmiron-treated mice at days 29 and 92, which correlates with increased spleen weight. The background incidence of extramedullary hematopoiesis in B6C3F1 mice in 3-month studies is not routinely collected by our laboratory or Contract Research Organizations but has been reported to be 17.6% (6.1–28 range) in males and 28.8% (12–52 range) in females from NTP feeding studies. Elmiron-treated mice also had hyperplasia of the splenic marginal zone and accumulation of foamy macrophages in the splenic sinusoids at day 92, which may have also contributed to the increased spleen weight at that time point. In contrast, lymphoid depletion with germinal center apoptosis was observed at day 92 in troglitazone-treated mice, which correlates with the decreased spleen weight at that time point. Decreased cytoplasmic fat content, as well as transdifferentiation into a brown adipose phenotype, was observed in the white adipose tissue of fenretinide-treated mice at all time points and may have contributed to the decreased weight seen at those time points. Decreased cytoplasmic fat content in white adipose tissue was also observed in elmiron-treated mice at day 29, which was transient and not present at day 92. In addition, elmiron-treated mice had mixed cell infiltrates present in the white adipose tissue at days 29 and 92. Troglitazone-treated mice did not display any white adipose alterations but did exhibit transdifferentiation of brown adipose tissue into a white adipose phenotype at all time points. Fenretinide-treated mice displayed hypereosinophilic adipocyte cytoplasm in brown adipose tissue at day 29, which was transient and not present at day 92. No findings were observed in the brown adipose tissue of elmiron-treated mice.
Cell Proliferation Analysis
Administration of fenretinide to male B6C3F1 mice for 15 and 29 days resulted in expected changes in overall proliferation and endothelial cell proliferation in brown adipose tissue and white adipose tissue (Table 8). Specifically, no changes were observed within the subscapular brown adipose tissue of animals treated with fenretinide at either time point. However, statistically significant increases in absolute and relative total proliferating cells and proliferating endothelial cells were observed at 92 days. For white adipose tissue, a statistically significant increase in absolute and relative proliferation was present at all time points.
Fenretinide effects on the liver were more complex, and results of total cell number, endothelial cell number, total proliferating cells, and proliferating endothelial cell values are reported in Table 8 and Supplementary Table 1. Following 15 days of fenretinide treatment, significant increases in relative endothelial cells, absolute and relative proliferating cells, as well as absolute and relative proliferating endothelial cells were observed compared with untreated controls. However, no significant differences were observed in absolute and relative proliferating endothelial cells between livers of treated versus untreated animals at day 29. Interestingly, this change was not observed at day 92, and statistically significant changes in total nuclear count as well as absolute and relative proliferating cells and proliferating endothelial cells were again seen.
Administration of troglitazone to male mice resulted in no proliferative effect changes of epididymal white adipose tissue, as was reported previously. In contrast, statistically significant increases in absolute and relative total proliferating cells and proliferating endothelial cells were observed in brown adipose tissue at days 15 and 92 of troglitazone administration (Table 8 and Supplementary Table 2). Additionally, absolute and relative total endothelial cells were significantly decreased at 92 days but not at any other time points. No changes were observed at 29 days for any of the endpoints in brown adipose tissue.
Measurements of total cell number, endothelial cell number, total proliferating cells, and proliferating endothelial cells were also made on the livers of study animals, which are reported in Table 8. Following 15 days of troglitazone treatment, significant increases in absolute endothelial cells, absolute proliferating cells, and absolute and relative proliferating endothelial cells were observed compared with untreated controls. However, no significant changes were observed in livers of animals receiving troglitazone at days 29 and 92 when compared with animals receiving vehicle only.
Administration of elmiron to male mice for all 3 time points resulted in no changes in overall proliferation and endothelial cell proliferation in brown adipose tissue (Table 8 and Supplementary Table 3). However, statistically significant increases were found in relative endothelial cell numbers at day 29 and in total nuclei count as well as absolute numbers of endothelial cells at 92 days. No changes were observed within the white adipose tissue of animals treated with Elmiron at all 3 time points.
Effects of Elmiron administration in the liver were complex (see Supplementary Table 3). Following 15 days of Elmiron treatment, no significant changes in the measured endpoints between treated and untreated controls were observed. However, statistically significant decreases were observed at day 29 in total nuclei count, as well as relative and absolute endothelial cell numbers. At day 92, this change was no longer observed, but statistically significant increases were present in relative and absolute proliferating cells as well as relative and absolute proliferating endothelial cells.
Porin and mnSOD Immunohistochemistry
For all 3 compounds, liver tissues from day 92 were evaluated for porin, a mitochondrial injury marker, and manganese superoxide dismutase, an oxidative stress marker, immunostaining. Both mitochondrial markers showed increased immunostaining in the treated mice compared with their concurrent controls. Porin immunostaining (Figure 2) demonstrated a more robust staining difference than manganese superoxide dismutase immunostaining (data not shown), and fenretinide-treated mice had greater porin immunostaining intensity than the troglitazone- or elmiron-treated mice.
Transcriptomics
Tissue samples from the days 15 to 29 necropsies were profiled on Affymetrix GeneChips to measure global changes in gene expression. Samples from day 92 were not used because we wanted to profile the earlier molecular events involved in initiating hemangiosarcoma. A potential limitation of the transcriptomic analysis was that whole tissue was used for the isolation of the ribonucleic acid rather than endothelial cells isolated from the tissue. However, whole tissue analysis was conducted since hepatocytes, Kupffer cells, and infiltrating leukocytes will all affect sinusoidal endothelial cells, which is why whole tissue analysis was ultimately selected. Differentially expressed genes were evaluated using the Causal Reasoning Engine as described in the “Materials and Methods” Section. Hypotheses related to elements of the generalized mode of action framework for hemangiosarcoma were specifically queried. Similar approaches were reported earlier with 2-butoxyethanol and pregabalin. As shown in Table 9, fenretinide had the largest number of differentially expressed genes, followed by troglitazone then elmiron. All differentially expressed genes are listed in Supplementary Table 4.
In the work described here, the focus was on hypotheses related to hypoxia, inflammation, and hemangiosarcoma/angiogenesis/cell proliferation. The primary goal was to assess whether transcriptomics detected evidence of hypoxia, macrophage activation using inflammation pathways as the biomarker, angiogenic growth factor upregulation, and hemangiosarcoma pathway activation in the target organ for hemangiosarcoma. The second goal was to assess whether transcriptomics was more sensitive than hypoxyprobe in detecting hypoxia. For macrophage activation, it is tissue-specific activation that is most critical. For hypoxia, it can either be tissue-specific, as in the case of troglitazone, or global hypoxia, as seen with 2-butoxyethanol and pregabalin. The results of the Causal Reasoning Engine provide inferred upstream controllers that explain the differentially expressed gene response and are referred to as hypotheses; the direction of these hypotheses may not be the same as the directional changes of individual genes. The hypotheses supporting hypoxia were initially identified previously and included hypoxia-inducible factor 1 alpha, endothelial PAS domain-containing protein 1, and “response to hypoxia.” Inflammation hypotheses also came from the same paper, including toll-like receptor 4, nuclear factor kappa B, interleukin 1 beta, and interleukin 6, but were supplemented by others associated with “well-known” inflammatory mediators such as lipopolysaccharide, tumor necrosis factor alpha, and interferon gamma. Finally, we identified hypotheses related to “hemangiosarcoma” using the LitMS tool that is part of the Causal Reasoning Engine. Some of the hemangiosarcoma hypotheses overlapped with those reported for endothelial cell proliferation previously, such as phosphatase and tensin homolog, and numerous processes involved in cell cycle progression such as myc oncogene and cyclin D1. Additional hypotheses were related more specifically to angiogenesis, for example, vascular endothelial growth factor, transforming growth factor beta, mesenchymal-epithelial transition, telomerase reverse transcriptase, and tumor protein p53; importantly, these growth factors also stimulate endothelial cell proliferation. A tabular summary of these hypotheses is shown in Table 9, and a complete list of all significant hypotheses is shown in Supplementary Table 5.
Effects of fenretinide. The original goal of this study was to profile both white adipose tissue and epididymis, but only white adipose tissue was profiled because the transcriptomic data from the epididymis did not yield any differentially expressed genes, likely due to the inability to reproducibly dissect uncontaminated regions of epididymal tissue. As noted earlier, treatment with fenretinide produced the strongest transcriptional response of all 3 compounds, with over three thousand differentially expressed genes at days 15 and 29. Support for hypoxia was seen at both time points, with an increase in endothelial PAS domain-containing protein 1 at days 15 and 29, and an increase in hypoxia-inducible factor 1 alpha and “response to hypoxia” also seen at day 29 only. Response to hypoxia was among the most significant hypotheses of the 354 that were identified in the day 29 treatment group. Hypotheses supporting a mixed inflammatory response were seen at day 15, while these same hypotheses were strongly increased at day 29. In fact, the lipopolysaccharide hypothesis was the top-ranked of all at day 29, with tumor necrosis factor alpha nearly equally as strong. When clustered, the top 10 hypotheses from the day 29 group form a network of inflammatory changes as shown by the associated Medical Subject Heading cloud (Figure 3). At day 15, there was evidence of vascular endothelial growth factor A increase, and by day 29 there were increases in several angiogenic growth factor hypotheses and a concomitant decrease in transforming growth factor beta. With the hemangiosarcoma-related hypotheses, there were overlapping hypotheses at both time points (peroxisome proliferator-activated receptor gamma, telomerase reverse transcriptase, myc oncogene, Harvey rat sarcoma viral oncogene, tumor protein p53, vascular endothelial growth factor A), but in some cases, the directions were opposite. The hemangiosarcoma hypotheses for both time points cluster into well-connected networks, and their associated Medical Subject Heading clouds include terms consistent with relevant pathological processes, such as cell transformation and neovascularization (Figure 4).
Effects of troglitazone. Both brown adipose tissue and liver from mice treated with troglitazone for 15 or 29 days were profiled. There were more differentially expressed genes in the liver at both time points than in the brown adipose tissue, but not as many differentially expressed genes compared with white adipose tissue from fenretinide-treated mice (Table 9). The strongest hypotheses in the liver and brown adipose tissue were associated with the primary pharmacology of the drug, including peroxisome proliferator-activated receptor gamma (data not shown). The only changes associated with hypoxia were decreased hypoxia-inducible factor 1 alpha and decreased response to hypoxia at day 29 in the liver. For inflammation, there were changes at both time points in the brown adipose tissue and only a weak response (increased lipopolysaccharide) at day 29 in the liver. Growth factor changes were not seen in either tissue at day 15. In the brown adipose tissue at day 29, changes in vascular endothelial growth factor A and transforming growth factor beta were seen, but in opposite directions as seen with fenretinide. In the brown adipose tissue, several hypotheses associated with the LitMS tag “hemangiosarcoma” were changing in brown adipose tissue at days 15 and 29, with only weak changes at both time points in the liver.
Effects of elmiron. The transcriptional response in livers of mice treated with elmiron was the weakest of all 3 compounds. On day 15, the most statistically significant change was in 1 hypothesis related to inflammation (toll-like receptor 7), and weak changes in the growth factor erythropoietin and growth factor receptor MET. On day 29, there were no significant changes related to hypoxia, inflammation, growth factors, or hemangiosarcoma.
DISCUSSION
Although hemangiosarcomas are a rare type of tumor in humans, their occurrence is more frequent in rodents, particularly mice, and this has presented challenges for drug approval. A common mechanism of action appears to underlie the nongenotoxic induction of hemangiosarcoma in mice, as suggested by the similar responses observed in mice treated with 2-BE, pregabalin, and troglitazone. While the specific initiating events might differ depending on the compound, evidence indicates that the convergence of these multiple initiating events leads to a disruption of angiogenesis and/or erythropoiesis resulting from hypoxia or macrophage activation. Prior research has demonstrated that pregabalin and 2-BE exhibit key components of this proposed model for hemangiosarcoma induction, including dysregulated erythropoiesis/angiogenesis, the generation of hypoxia, increases and activation of macrophages, and endothelial cell proliferation. In a previous review, troglitazone data had not utilized hypoxyprobe or the CD-31/BrdU dual cell proliferation methods that were employed with 2-BE and pregabalin. Therefore, troglitazone was included in the current study along with two other biologically diverse drugs exhibiting varying potency for hemangiosarcoma induction: fenretinide (high), troglitazone (intermediate), and elmiron (low). The objective was to compare the key elements in the proposed model for nongenotoxic hemangiosarcoma induction in mice with the findings from previous work on 2-BE and pregabalin. The five key elements of this mechanism of action include hypoxia (assessed by Hypoxyprobe and transcriptomics), macrophage activation (assessed by hematology and transcriptomics), growth factors (with extramedullary hematopoiesis as a surrogate, also assessed by transcriptomics), dysregulated angiogenesis/erythropoiesis (assessed by hematology and transcriptomics), and endothelial cell proliferation (assessed by CD31/BrdU dual staining). For transcriptomics, the focus was on identifying evidence of hypoxia, macrophage activation (using inflammation pathways as a biomarker), angiogenic growth factor upregulation, and hemangiosarcoma pathway activation in the target organ for hemangiosarcoma. The two primary goals of this report were to evaluate the proposed mechanism of action by including three additional compounds to strengthen the weight of evidence and to assess the robustness of the experimental endpoints, such as comparing hypoxyprobe with transcriptomics.
Fenretinide is a synthetic retinoid that has been investigated for its effects on obesity and type-2 diabetes through its apoptotic actions on adipocytes. In a two-year carcinogenicity study in B6C3F1 mice, fenretinide was found to be a highly potent inducer of hemangiosarcoma, producing nearly 100% incidence of hemangiosarcoma in male and female mice within the hemolymphoreticular system, including the bone marrow, liver, lymph node, and spleen. In male mice, hemangiosarcomas were also reported in the epididymis and epididymal white adipose tissue; consequently, these tissues were included in the study design. Dysregulated erythropoiesis was evident in fenretinide-treated mice, characterized by a decrease in reticulocyte response despite decreased red blood cell counts and hemoglobin levels, coupled with a significant increase in bone marrow macrophage numbers (six- to eleven-fold at days 15, 29, and 92). Consistent with dysregulated erythropoiesis, spleen weights were increased at all time points, and extramedullary hyperplasia and extramedullary hematopoiesis were observed at days 29 and 92. These changes were accompanied by splenic hypoxia at days 29 and 92, as detected by Hypoxyprobe. Liver endothelial cell proliferation was statistically significant at days 15, 29, and 92, although liver hypoxia was only observed at day 29. On day 92, increased immunostaining for mitochondrial markers porin and mnSOD in the liver indicated a response to mitochondrial injury and oxidative stress, respectively. Notably, porin immunostaining intensity was greater with fenretinide treatment compared to troglitazone or elmiron, suggesting that fenretinide results in a more pronounced mitochondrial injury response than the other two compounds. MnSOD staining intensity was similarly elevated for all three treatments, indicating that these compounds led to increased oxidative stress in the liver, a response consistent with that reported for 2-BE. Fenretinide-treated mice exhibited a decrease in white adipose tissue weight, consistent with the transdifferentiation of white adipose tissue to a brown adipose phenotype, as well as an increase in total cell proliferation across all time points. Consistent with increased cell proliferation in white adipose tissue at day 29, gene expression data analysis revealed numerous hypotheses indicative of angiogenic growth factors such as ANGPT1, fibroblast growth factor 2, VEGFa, EPO, and TGFb. TGFb is known to inhibit endothelial cell proliferation, so the predicted decrease in its expression would be expected to be permissive for proliferation. Evidence for hypoxia in white adipose tissue using Hypoxyprobe was observed only at day 29, and this was strongly supported by transcriptomics evidence of increased hypoxia. Although there was no histological evidence of inflammation in the white adipose tissue, the transcriptomics data provided strong support for an increased inflammatory response at day 29, including a specific reference to “macrophage activation.” Therefore, fenretinide demonstrated all five key elements of the proposed mechanism of action for hemangiosarcoma: hypoxia (in the liver, spleen, and white adipose tissue) with associated mitochondrial injury and oxidative stress (in the liver), macrophage activation (in the bone marrow and white adipose tissue based on transcriptomics), growth factors (extramedullary hematopoiesis and transcriptomics in white adipose tissue), dysregulated erythropoiesis (hematology), and endothelial/total cell proliferation (in the liver and white adipose tissue). The correlation between hypoxia and endothelial cell proliferation was strongest for the liver compared to white adipose tissue. Furthermore, the high number of significant hypotheses related to “hemangiosarcoma” identified at both days 15 and 29 of fenretinide treatment further supports the proposed mechanism of action.
Troglitazone is a thiazolidinedione that binds to and activates PPARc, resulting in the inhibition of hepatic gluconeogenesis and the enhancement of insulin activity. In a two-year mouse carcinogenicity study, troglitazone produced a 2.7-fold increase in the total incidence of hemangiosarcoma in male and female mice compared to the concurrent control groups (27% incidence for all organs). Thus, the incidence of hemangiosarcoma in troglitazone-treated mice was 3.7-fold lower than in fenretinide-treated mice. In male mice treated with troglitazone, hemangiosarcomas were elevated in the liver, bone marrow, skin/subcutis, and spleen. Dysregulated erythropoiesis was apparent in troglitazone-treated mice based on the decrease in reticulocyte counts (14%–29% on days 29 and 92, respectively) despite decreased red blood cell counts (days 15 and 29) and hemoglobin levels, coupled with an increase in bone marrow macrophages (four- to seven-fold on days 15, 29, and 92). In contrast to fenretinide, spleen weights were decreased at day 92, secondary to lymphoid depletion, consistent with previous reports of troglitazone’s anti-inflammatory actions. There was no evidence of splenic hypoxia at any time point via hypoxyprobe. Troglitazone has been reported to cause hyperplasia of brown adipose tissue, and this increase in adipocytes could have been a potential source of angiogenic growth factors. In the current study, we also observed transdifferentiation of brown adipose tissue into a white adipose phenotype in troglitazone-treated mice, consistent with PPARc activation. An increase in brown adipose tissue endothelial cell proliferation was observed at days 15 and 92, but without any evidence of hypoxia via hypoxyprobe or transcriptomics. Hepatocellular hypertrophy, consistent with enzyme induction, was observed at all time points, while hepatic endothelial cell proliferation was only increased at day 15, and hypoxia was observed at days 15 and 92. In the liver, at day 29, there was transcriptional evidence for a decrease in hypoxia, a finding also observed in bone marrow following seven days of treatment in mice with 2-BE, another compound that induces hemangiosarcoma. Previous research reported a 2.95-fold increase in liver endothelial cell proliferation in female mice administered troglitazone. Increased immunostaining for porin and mnSOD was also observed in the liver at day 92, supporting mitochondrial injury and oxidative stress responses in this tissue, the latter being consistent with that seen with 2-BE. Other research reported a four- and 1.52-fold increase in brown adipose tissue endothelial cell proliferation after seven and fourteen days, respectively, in female mice treated with a higher dose of troglitazone. We observed a 2.4-fold increase in brown adipose tissue endothelial cell proliferation at day 15, similar to the other two groups. In summary, troglitazone demonstrated four of the five key elements of the proposed mechanism of action for hemangiosarcoma: hypoxia (in the liver and brown adipose tissue) with associated mitochondrial injury and oxidative stress (in the liver), macrophage activation (in the bone marrow), dysregulated erythropoiesis (hematology), and endothelial/total cell proliferation (in the liver and brown adipose tissue). The correlation between hypoxia and endothelial cell proliferation was less strong than that observed with fenretinide. Because single cell necrosis was observed in the liver, a confounding transcriptional response for the liver alone cannot be ruled out. This limitation was minimized because the transcriptional analysis focused on confirming previously described changes associated with the mechanism of action observed with other compounds that induce hemangiosarcoma.
Elmiron is used for the chronic treatment of interstitial cystitis, a disorder characterized by bladder pain and urinary urgency in the absence of infection. In sub-chronic studies, chronic inflammation and an increased incidence of vacuolated histiocytes in multiple tissues, including the rectum, lymph nodes, liver, and spleen, were observed. In a two-year carcinogenicity study in B6C3F1 mice, elmiron increased hepatic hemangiosarcoma incidence in male and female mice compared with the concurrent controls. Hemangiosarcomas were evident only in the mouse liver, unlike the findings from the fenretinide and troglitazone carcinogenicity studies, which showed multiple organs with hemangiosarcoma. The incidence of hemangiosarcoma in elmiron-treated mice was 5.6-fold lower than in fenretinide-treated mice. Dysregulated erythropoiesis was apparent in elmiron-treated mice, characterized by a decrease in reticulocyte response despite decreased red blood cell counts and hemoglobin levels, coupled with an increase in bone marrow macrophage numbers (four-, seven-, and eight-fold on days 15, 29, and 92, respectively). The decreases in red blood cells and hemoglobin were similar in magnitude and consistent with previous reports. Consistent with dysregulated erythropoiesis, spleen weight was increased at day 92, extramedullary hematopoiesis was observed at days 29 and 92, and foamy macrophages were observed at day 92. The foamy macrophages have been attributed to an accumulation of neutral and acidic mucins, lipids, and possibly elmiron itself within the lysosomes of macrophages, similar to that seen with lysosomal storage disorders. This change is hypothesized to lead to the sustained local production of growth factors, such as VEGF, which may in turn result in increased endothelial cell proliferation. These changes were accompanied by splenic hypoxia at day 92, as detected by Hypoxyprobe. The extramedullary hyperplasia observed in the spleens of elmiron-treated mice is most likely a secondary adaptive response to the erythrocyte decreases and the absence of a regenerative bone marrow response seen with these compounds. As expected, there were no effects on brown adipose tissue and white adipose tissue, except for a mixed cell infiltrate in the white adipose tissue. In the liver, a statistically significant increase was found in both relative and absolute proliferating endothelial cells at day 92, without any evidence of hypoxia at any time point. The mitochondrial markers, porin and mnSOD, had increased immunostaining in the liver at day 92, although to a lesser degree than that seen for the other two compounds. In summary, elmiron demonstrated three of the five key elements of the proposed mechanism of action for hemangiosarcoma: macrophage activation (in the bone marrow and spleen), dysregulated erythropoiesis (hematology), and endothelial cell proliferation (in the liver). The correlation and magnitude of changes between hypoxia and endothelial cell proliferation were less strong than that observed with fenretinide and troglitazone. It is noteworthy that hypoxia in the liver was not detected, as this is the organ where hemangiosarcoma was observed. Finally, transcriptional evidence for any of the key steps in the proposed mechanism of action was weak for elmiron.
Of the five elements examined, there was strong evidence for dysregulated erythropoiesis and macrophage activation across all three compounds. Specifically, a decrease in red blood cells and a failure to increase reticulocytes, coupled with four- to eleven-fold increases in bone marrow macrophages, were consistently observed. This pattern of hematological changes in mice might serve as an early biomarker to evaluate endothelial cell proliferation in suspected target organs for potential hemangiosarcoma formation. Although the mechanism for increases in bone marrow macrophages is unclear, an apoptotic effect on bone marrow adipocytes could potentially stimulate macrophages to remove the extruded nuclei of bone marrow nucleated erythrocytes. The correlation between tissue hypoxia and endothelial cell proliferation in the organs where hemangiosarcoma occurred was strongest with fenretinide (liver, white adipose tissue—evidence of both hypoxia and cell proliferation), intermediate with troglitazone (liver—evidence of both hypoxia and cell proliferation, brown adipose tissue—evidence for cell proliferation but not hypoxia), and weakest for elmiron (liver—evidence for cell proliferation with no evidence of hypoxia). The hypoxyprobe immunohistochemistry probe suffers from a low dynamic response, and it was hoped that transcriptomics would improve upon this endpoint. Unfortunately, this was not the case. Although there was evidence of hypoxia using both approaches in white adipose tissue at day 29 in fenretinide-treated mice, in livers from troglitazone-treated mice, the increase in hypoxyprobe signal at day 15 was not accompanied by changes in transcriptomics. Conversely, at day 29 in the troglitazone-treated mice, there was transcriptional evidence of hypoxia in the absence of a signal with hypoxyprobe. The evidence for increased endothelial cell growth factors was indirect, inferred from the spleen extramedullary hematopoiesis that was only observed with fenretinide and elmiron. Transcriptomics did demonstrate a strong angiogenic growth factor signal with fenretinide (day 29, white adipose tissue) and weaker changes with troglitazone and elmiron. Regarding dysregulated angiogenesis, transcriptomic data demonstrated a strong signal with fenretinide at day 29 in the white adipose tissue, and weaker signals at day 29 in the liver and brown adipose tissue with troglitazone.
In summary, this study was conducted with three biologically diverse compounds exhibiting varying potency for hemangiosarcoma induction: fenretinide (high), troglitazone (intermediate), and elmiron (low). The goal was to compare the key elements in the proposed mechanism of action and increase the weight of evidence for hemangiosarcoma formation by using these three drugs with divergent pharmacological properties. The five key elements of the mechanism of action are hypoxia in target organs, macrophage activation in the target organ and/or bone marrow, increased angiogenic growth factors, dysregulated angiogenesis/erythropoiesis, and endothelial cell proliferation in the organs where hemangiosarcoma is observed. Fenretinide demonstrated all five key elements of the proposed mechanism of action for hemangiosarcoma, troglitazone demonstrated four of the five elements, and elmiron demonstrated three of the five elements. The overall transcriptional evidence for the key elements of the proposed mechanism of action was also consistent with the potency of hemangiosarcoma induction: fenretinide greater than troglitazone greater than elmiron. These data, coupled with previous work involving 2-BE and pregabalin, strengthen the weight of evidence for this mechanism of action for hemangiosarcoma. Importantly, this time-course study demonstrates that several of the changes in the proposed mechanism of action for hemangiosarcoma induction can be detected in studies lasting up to three months, which are routinely used in safety assessment.