|Home | About | Journals | Submit | Contact Us | Français|
Hepatic haemangiosarcoma is a deadly malignancy whose aetiology remains poorly understood. Inactivation of the CDKN2A locus, which houses the ARF and p16INK4a tumour suppressor genes, is a common event in haemangiosarcoma patients, but the precise role of ARF in vascular tumourigenesis is unknown. To determine the extent to which ARF suppresses vascular neoplasia, we examined the incidence of hepatic vascular lesions in Arf-deficient mice exposed to the carcinogen urethane (i.p. 1 mg/g). Loss of Arf resulted in elevated morbidity and increased the incidence of both haemangiomas and incipient haemangiosarcomas. Suppression of vascular lesion development by ARF was heavily dependent on both Arf gene-dosage and the genetic strain of the mouse. Trp53-deficient mice also developed hepatic vascular lesions after exposure to urethane, suggesting that ARF signals through a p53-dependent pathway to inhibit the development of hepatic haemangiosarcoma. Our findings provide strong evidence that inactivation of Arf is a causative event in vascular neoplasia and suggest that the ARF pathway may be a novel molecular target for therapeutic intervention in haemangiosarcoma patients.
Haemangiosarcoma is a rare but deadly disease that accounts for approximately one percent of all sarcoma cases and two percent of all primary liver malignancies in people [1–3]. Hepatic haemangiosarcoma development has been associated with occupational exposure to a number of environmental toxins, and presentation of both cutaneous and visceral haemangiosarcomas after radiotherapy is also common [4–8]. Despite the aggressive nature of this disease, there are few treatment options for patients presenting with haemangiosarcoma of the liver or other internal sites. Aggressive surgical resection is the standard form of treatment, but post-operative chemotherapy regimens are limited in both number and efficacy. Consequently, five-year survival rates for haemangiosarcoma patients are generally poor [1, 2]. The limited success of adjuvant therapy reflects a fundamental lack of insight into the aetiology of haemangiosarcoma. An enhanced understanding of the molecular mechanisms underlying this disease will identify novel targets for clinical intervention.
The CDKN2A locus, which houses the ARF (p14ARF in humans, p19Arf in mice) and p16INK4a tumour suppressor genes , is frequently inactivated in hepatic haemangiosarcoma . Epigenetic silencing of ARF occurs in 26% of cases , but a direct, causative role for ARF in haemangiosarcoma formation has yet to be established. ARF is best known for stabilizing p53 activity in oncogene-stressed cells [10–12]. In recent years, additional functions of ARF have also been described , such as controlling the regression of the hyaloid vasculature system in the developing mouse eye [14–18]. However, the extent to which ARF regulates the proliferation or transformation of endothelial cells in adult mice and humans remains poorly studied.
Urethane (ethyl carbamate) is a chemical carcinogen frequently utilized in mouse pulmonary tumourigenesis studies. Occurring naturally in many foods and beverages as a fermentation by-product , urethane is metabolized by cytochrome P450 2E1 to form vinyl carbamate epoxide, which in turn produces potentially mutagenic DNA adducts . Notably, urethane’s metabolites produce the same DNA adducts as vinyl chloride , a chemical known to induce angiosarcoma of the liver in humans . When administered at high doses or with recurrent low dose application, urethane induces long-latency hepatic vascular lesions in wild-type mice [23–26]. One case of urethane-associated hepatic haemangiosarcoma in a human patient has also been reported .
Here we show that Arf-deficient mice exposed to urethane exhibited a significantly increased incidence of hepatic haemangiomas and incipient haemangiosarcomas. Arf loss contributed to multiple stages of vascular tumourigenesis, from early, preneoplastic angiectasis to the ultimate formation of haemangiomas with cellular atypia and haemangiosarcomatous differentiation. We further demonstrate that suppression of vascular neoplasia by ARF is heavily influenced by genetic strain, is Arf gene-dose-dependent, and likely proceeds through a p53-dependent pathway. Our findings present strong evidence that ARF suppresses the formation of vascular tumours in adult mice and, furthermore, provide a model in which to study the aetiology of vascular neoplasia.
All experiments were approved by the FHCRC Institutional Animal Care and Use Committee. Arf-deficient mice  were backcrossed separately onto C57BL/6 (The Jackson Laboratory, Bar Harbor, ME, USA) and NIH/Ola (Harlan Olac, Oxfordshire, UK) strains for 16 and 20 generations, respectively. Arf +/− mice were intercrossed and carcinogenesis studies performed on resultant littermates of each genotype: Arf −/− (n = 69 for C57BL/6; n = 59 for NIH/Ola), Arf +/− (n = 42 C57BL/6; n = 33 NIH/Ola), and Arf +/+ (n = 30 C57BL/6; n = 56 NIH/Ola). Mice deficient for Trp53  were backcrossed onto the C57BL/6 strain for 20 generations, and carcinogenesis studies were performed on male mice of each genotype: Trp53 −/− (n = 11), and Trp53 +/− (n = 22). DNA for genotyping was isolated by digestion of ear tissue with proteinase K in InstaGene Matrix solution (Bio-Rad, Hercules, CA, USA). Each mouse was genotyped as described [28, 29].
Arf- and Trp53-deficient mice were subjected to a single injection of urethane (Sigma, St. Louis, MO, USA) in a PBS vehicle (i.p. 1 mg/g bodyweight) between 21 and 28 days of age. An additional cohort of Trp53 +/− mice (n = 12) was injected with urethane at 14 days of age. A cohort of unexposed C57BL/6 strain Arf −/− (n = 25), Arf +/− (n = 23), and Arf +/+ mice (n = 22) was included for control. Animals were euthanized by CO2 asphyxiation at indicated time points or when moribund. All tissues were examined for the presence of lesions during necropsy. Normal and pathological tissues were frozen in liquid nitrogen and/or fixed for subsequent analysis.
Tissues were fixed in 10% neutral buffered formalin for at least 4 hrs, then processed and embedded in paraffin. 4-µm-thick tissue sections were cut, mounted on glass slides, and stained with H&E. Tissue samples were submitted to the FHCRC institutional histopathology core for von Willebrand Factor staining (#A0082; Dako, Glostrup, Denmark). A small set of tissue sections was also stained for phosphorylated p42/44 MAPK (pERK) (#4376; Cell Signaling Technology, Danvers, MA, USA). Controls included mouse kidney and rabbit IgG isotype control. All H&E slides were analysed by a veterinary pathologist blind to genotype. Diagnosis of angiectasis was based on dilatation of the liver sinusoids in the absence of proliferating endothelial cells. Diagnosis of haemangioma was determined by the presence of abnormal vascular spaces lined by proliferating endothelial cells. A subset of more aggressive haemangiomas presented with cellular atypia, regions of solid growth, and incipient haemangiosarcomatous differentiation . In order to match the original stained slide, global adjustments to white balance, brightness, and/or colour saturation were made to copies of some photomicrographs using Adobe Photoshop®.
Time until development of hepatic vascular lesions causing death or requiring euthanasia was graphically summarized in a Kaplan-Meier plot, and survival differences were analysed for significance using the log-rank test. Animals euthanized for reasons other than the presentation of vascular lesions were considered censored observations in the second log-rank analysis. These animals were included in the analysis of overall mortality for Arf mice. The incidence of lesions by 25 weeks post urethane-injection was compared between strains for each genotype, and between Arf- and Trp53-deficient animals; differences were analysed for significance using the Fisher’s exact test. Mice who died before 25 weeks and did not present with macroscopic hepatic vascular lesions were excluded from the Fisher’s exact test analyses.
To investigate the tumour suppressive role of ARF in vascular neoplasia, we exposed Arf-deficient mice to the carcinogen urethane. Kaplan-Meier analysis demonstrated that C57BL/6 Arf −/− mice injected with urethane suffered significantly shortened lifespans compared to their Arf +/+, Arf +/− and untreated Arf −/− littermates (Fig. 1A). The median life expectancy of urethane-exposed Arf −/− mice was reduced to 262 days from greater than 400 days for all other genotypes. Both Arf −/− and Arf +/− mice also developed lymphomas and sarcomas (data not shown) at a rate similar to that previously reported . Moribund Arf −/− mice frequently presented with lethargy, swollen abdomens, and symptoms of anaemia (e.g. pale ears and digits). During necropsy these animals were found to have grossly diseased livers bearing multifocal, massive thromboses (Fig. 1B). Rupture of these thromboses resulted in abdominal haemorrhaging that caused near-immediate death. In contrast to the marked lethality of urethane to C57BL/6 Arf −/− mice, no Arf +/+ or Arf +/− littermates or unexposed Arf −/− mice succumbed to hepatic vascular lesions by 50 weeks post-injection (Fig. 1C). Lesion incidence and associated lethality were statistically similar between genders (data not shown).
Tumour susceptibility in mice is strongly influenced by genetic background. To investigate the effect of genetic background on urethane-induced vascular lesions, we examined tumour incidence in NIH/Ola strain Arf-deficient mice (Fig. 1D). As in the C57BL/6 Arf +/+ mice, urethane injection did not yield gross hepatic lesions in NIH/Ola Arf +/+ mice. However, in contrast to the 40.9% (9 of 22) of C57BL/6 Arf +/− mice that exhibited macroscopic hepatic vascular lesions by 25 weeks, no NIH/Ola Arf +/− mice (0 of 32) bore lesions at this age. Furthermore, only 66.7% of NIH/Ola Arf −/− mice (12 of 18) developed lesions by 25 weeks, compared to 91.7% of C57BL/6 Arf −/− mice (44 of 48). Hepatic vascular lesions that did develop in NIH/Ola strain mice were typically smaller and less physically deleterious than lesions in C57BL/6 mice. Therefore, the susceptibility of Arf-deficient animals to vascular carcinogenesis is modified by genetic background, with the relative risk of C57BL/6 strain exceeding NIH/Ola.
To determine the histopathological basis of the hepatic lesions, H&E stains for a subset of samples collected at 25 and 50 weeks post-exposure were examined by a veterinary pathologist. The analysis uncovered a tumour progression spectrum of vascular neoplasia in Arf-deficient mice, with lesions ranging from angiectasis to benign haemangioma to haemangioma with cellular atypia (Table 1). Low magnification views of representative wild-type and Arf-deficient liver lobes are shown in Figure 2 (A, B), and characteristic images of the tumour diagnostic criteria are provided (Fig. 2C–E). By 50 weeks post-exposure, 40% of Arf +/+ mice exhibited focal angiectasis but were otherwise unaffected compared to Arf-deficient mice. At 25 weeks, all Arf +/− mice examined presented with angiectasis of the liver. By 50 weeks, 80% of Arf +/− mice exhibited haemangioma lesions, with one mouse displaying multifocal cellular atypia and haemangiosarcomatous differentiation. A nearly identical rate of haemangioma incidence (75%) was observed in Arf −/− mice, but at the earlier time point of 25 weeks post-urethane exposure. 38% and 40% of Arf −/− mice presented with incipient haemangiosarcomas at the 25 and >25 week time points, respectively. One unexposed Arf −/− mouse (n = 3) developed a small haemangioma by 50 weeks, and no vascular lesions were observed in unexposed Arf +/− mice at the same time point (n = 5; data not shown).
The diagnosis of vascular neoplasia was substantiated by immunohistochemical analysis with von Willebrand Factor (vWF) antibody, an established marker of endothelial cells (Fig. 2F–H). While vWF staining was restricted to the endothelial cells lining hepatic blood vessels in unexposed mice (Supplementary Fig. 1A), abundant vWF positive cells were present in urethane-induced tumours (Fig. 2G, H), confirming the endothelial nature of the hepatic lesions. Vascular lesions that developed in urethane-exposed, Arf-deficient animals – including a cardiac haemangioma – were also focally positive for phosphorylated ERK, an indicator of RAS pathway activation (Supplementary Fig. 1A, B). Taken together, these findings suggest that suppression of vascular tumour development by ARF in this model is time- and Arf gene-dose-dependent, in that Arf +/− mice demonstrated both delayed neoplastic formation and malignant progression compared to their Arf −/− littermates.
Many of the tumour suppressive activities previously attributed to ARF are related to its role in stabilizing cellular p53 levels , and Trp53 loss has been implicated in haemangiosarcoma development in mice and humans [32–36]. These facts prompted us to determine the extent to which suppression of vascular neoplasia by ARF is p53-dependent. We analysed the effect of urethane exposure on C57BL/6 Trp53-deficient male mice. All Arf +/+; Trp53 −/− animals (n = 11) succumbed to thymic lymphomas by 15 weeks post-exposure, and none exhibited hepatic vascular lesions at necropsy. However, 9.5% of Arf +/+;Trp53 +/− mice (2 of 21) developed hepatic vascular lesions by 25 weeks (Fig. 3A), and a small cohort of mixed-gender Arf +/+;Trp53 +/− mice exposed to urethane at an earlier time point (14 days) and allowed to age up to 58 weeks post-injection presented with malignant hepatic vascular lesions (3 of 12; 25%; Fig. 3A, B). Although the 25-week lesion incidence rate for Trp53 +/− mice (9.5%) was lower than that observed in Arf +/− mice (3 of 12 male mice; 25%), this difference was not statistically significant (P = 0.3275). The development of vascular tumours in urethane-exposed Trp53-deficient mice supports the argument that tumour suppression by ARF in this model is likely mediated by p53.
The aetiology of hepatic haemangiosarcoma is poorly understood, and, consequently, patients diagnosed with the disease face extremely poor prognoses. Weihrauch and colleagues previously reported that the p14ARF locus is frequently silenced in haemangiosarcomas of the liver . Their work provided correlative but not causative evidence for tumour suppression by ARF in vascular neoplasia. Here we report evidence for a direct, causative role for ARF and its downstream signalling partner p53 in suppressing the formation and malignant progression of primary hepatic vascular tumours. Mice with homozygous deletion of Arf exhibited a significant predisposition for the development of haemangiomas and incipient haemangiosarcomas after carcinogen exposure, demonstrating that loss of Arf is sufficient to promote vascular neoplasia.
Many cancers are known to develop through a step-wise process, from hyperplastic to benign to malignant lesions. Evidence for this type of evolutionary progression, however, has not been established for haemangiosarcoma development . Our study captured a tumour progression spectrum of hepatic vascular lesions. Arf-deficient mice developed lesions ranging from angiectasis to benign haemangiomas and incipient haemangiosarcomas in a time- and Arf gene-dose-dependent manner. In particular, Arf +/− mice displayed a dramatic disease progression between the 25 and 50-week time points. These findings elucidate the evolution of vascular tumours and, moreover, support the potential use of this mouse model in future investigations into the aetiology of haemangiosarcoma.
An interesting finding from our study was that mouse genetic background strongly influenced susceptibility to the development of vascular tumours after urethane exposure. C57BL/6 Arf −/− mice presented with hepatic vascular lesions at a significantly higher frequency than NIH/Ola Arf −/− mice. The strain-specificity of this disease phenotype indicates the presence of genetic modifiers that differentially regulate vascular tumourigenesis between strains. Although a previous study of C57BL/6 Trp53 +/− mice chronically exposed to urethane found a high susceptibility to hepatic vascular carcinogenesis , another study observed protection against hepatic vascular lesions in C57BL/6 strain Vhl-mutant mice compared to A/J and BALB/c strain mutants . Clearly, further investigation is warranted. Our model may lend itself to the future identification of additional genes involved in haemangiosarcoma susceptibility.
Vascular tumours are believed to result from the aberrant proliferation and transformation of endothelial cells . ARF is required for the developmentally controlled regression of the hyaloid vasculature system in the mouse eye , but the question of whether ARF contributes to vasculogenesis or angiogenesis in adults remains largely unanswered. Our finding that loss of Arf promotes the formation of vascular tumours indicates that ARF may, under defined circumstances, control endothelial cell proliferation in adult animals. However, an alternative hypothesis regarding haemangiosarcoma histogenesis posits that the cell of origin is a pluripotent mesenchymal stem cell . This hypothesis holds that although haemangiosarcomas present with an endothelial cell-like appearance, the tumours may in fact not be derived from mature endothelium but rather from a more primitive, progenitor cell.
Arf-deficient mice are prone to spontaneous sarcomagenesis , and there is mounting evidence that ARF suppresses the development of multiple sarcoma types in humans [39–44]. This association between altered ARF/INK4a expression and the development of a wide range of sarcoma subtypes argues that the CDKN2A locus may play a critical role in arresting the growth of a transformed mesenchymal progenitor cell. Indeed, CDKN2A was found to be deleted in hTERT-transduced adult mesenchymal stem cells that exhibited anchorage independence and formed tumours when injected into mice [45, 46]. Taken together with these findings, our study supports the argument that the development of vascular tumours upon Arf loss reflects changes in mesenchymal stem cell differentiation and proliferation rather than a hyperproliferation of endothelial cells. In either case, the absence of vascular lesions in unexposed Arf-deficient mice indicates that for haemangiosarcoma development to occur, Arf loss must be accompanied by a cooperative “event” provided by urethane, presumably an oncogenic mutation.
The canonical function of ARF is to antagonize the MDM2 (HDM2 in humans) ubiquitin ligase in oncogene-stressed cells, thereby stabilizing cellular p53 levels and engaging an anti-proliferative response [12, 47]. Urethane is known to cause oncogenic Ras mutations in rodents [48–50], and mutant RAS has also been identified in human hepatic haemangiosarcoma [51, 52]. In this study, we observed evidence of RAS-ERK pathway activation in urethane-induced vascular lesions. Moreover, we have previously demonstrated that ARF suppresses the malignant progression of RAS-driven squamous cell carcinoma through a p53-dependent pathway , and other groups have shown that stabilization of p53 by ARF is required for cancer protection many weeks after exposure to genotoxins [54, 55]. Our findings here reveal a similar pattern. The ARF-p53 axis acts as an important, delayed barrier to carcinogen-induced haemangiosarcoma development.
Our study provides the first causative link between Arf loss and vascular neoplasia. Nevertheless, the specific contributions of ARF and INK4a, the two genes at the CDKN2A locus, to tumour suppression in humans remain controversial [13, 28, 56–60]. Deletion of the entire chromosomal region is frequent [61–63], but it is clear that selective inactivation of ARF does occur in human cancer . Weihrauch et al have published on the alterations of INK4a and ARF in human hepatic angiosarcoma . Their findings show that loss of INK4a almost certainly contributes to hepatic vascular neoplasia, as 63% of tumours exhibited hypermethylation of INK4a. However, ARF was also found to be hypermethylated in human hepatic angiosarcoma (26%), and in 11% of patients hypermethylation was specific to ARF while INK4a remained unmethylated. Furthermore, the strong evidence that p53 and HDM2 are altered in human vascular neoplasia suggests that their functional partner ARF may also play a role in the disease [33, 34, 36, 64]. In summary, these findings argue that there is selective pressure to inactivate both the ARF-HDM2-p53 pathway and the INK4a-RB pathway in human hepatic vascular neoplasia, as is seen in many other human malignancies [65, 66].
The need for novel therapeutics to treat haemangiosarcoma patients is critical. Current standards of treatment are focused primarily on aggressive surgical resection, and the availability of adjuvant chemotherapies is limited, in part because few studies to date have elucidated the molecular pathways underlying vascular neoplasia [1, 2]. Prognoses for patients diagnosed with haemangiosarcoma will likely not improve without significant advancement in our knowledge of the aetiology of the disease. The identification of ARF as a suppressor of hepatic vascular lesion development illuminates the process of haemangiosarcomagenesis and suggests potential targets for future therapeutic intervention.
Supplementary Figure 1. Hepatic vascular lesions exhibit activation of the RAS-ERK pathway. A: Proliferating endothelial cells, highlighted with vWF (right) were focally positive for phospho-ERK in the hepatic haemangiomas (600X). Arrowheads indicate positive cells. B: A urethane-induced cardiac haemangioma was strongly positive for phosopho-ERK (600X).
We thank Kyung Hoon-Kim for his technical support and the FHCRC Experimental Histopathology core for performing the vWF immunohistochemistry. We are grateful to Sue Knoblaugh and H. Denny Liggitt for their assistance in analysing histological specimens. We also thank Karen Kelly-Spratt and William Grady for their advice on the manuscript. This work was supported by the NCI Mouse Models of Human Cancer Consortium, U01 CA141550, and by NIEHS 5 R01 ES020116. SEB was supported by Public Health Service, National Research Service Award, T32 GM007270 from the National Institute of General Medical Sciences.
Conflict of interest statement: The authors have no conflicts of interest to declare.
Statement of Author ContributionsSEB and CJK designed the study and interpreted data. SEB conducted experiments, analysed data, generated figures, and wrote the manuscript. KEG conducted experiments. RDM conducted experiments and interpreted data. All authors participated in drafting the manuscript and had final approval of the submitted and published versions.