Ikk4a/Arf Inactivation with Activation of the NF-κB/IL-6 Pathway Is Sufficient to Drive the Development and Growth of Angiosarcoma

Angiosarcoma is a malignant vascular tumor of endothelial origin. There has been an increase in angiosarcomas over the last 30 years. This type of sarcoma is known to have an extremely high mortality rate (∼79% to 83%) due to rapid growth and metastasis (1–4). Lesions are thought to arise from the transformation of endothelial cells that are resident within tissues or from circulating stem cells recruited from bone marrow or sites of extramedullary hematopoiesis (5, 6). Though the exact etiology of angiosarcoma is unknown, many diverse environmental factors are linked to its development.

Genetic mutations or gene amplifications for VEGF, MDM2, p53, RAS, MYC have been investigated in a large number of clinical angiosarcoma tissues (7–9), but the genetic changes identified are complex (10). In human primary angiosarcoma, the INK4a/ARF locus on human chromosome 9p21 encoding cycle-regulatory proteins, p16^INKa and p14^ARF, is frequently disrupted by either deletion of the locus or methylation of the promoter (11, 12). Overexpression of p16^INK4a leads to senescence, whereas loss of p16^INK4a results in immortalization in human cells (13, 14). The ARF tumor suppressor (p14^ARF in human, p19^ARF in mouse) activates p53 (15, 16). Of interest, p53 is frequently mutated in human angiosarcoma (4). Loss of p53 links to enhanced activation of NF-κB just as methylation of ARF results in failure to induce p53 and resultant enhanced NF-κB mediated transcription.

Aberrant NF-κB is associated with tumorigenesis, angiogenesis and metastasis (17). Moreover, apart from effects of the NF-κB pathway intrinsic to the developing tumor cells, it is increasingly clear that the tumor microenvironment and its leukocyte infiltrate can have antitumor or protumor effects. One major target gene transcriptionally regulated by NF-κB is IL-6, a cytokine produced and released by a variety of cell types including macrophages, fibroblasts, endothelial cells, T cells, and B cells, which are frequently associated with enhanced tumor growth. IL-6 secretion by angiosarcoma lesions has been associated clinically with locally aggressive behavior and metastasis (18). Moreover, phosphorylation of STAT3 is common in cutaneous angiosarcoma lesions (19) and tumors with constitutive NF-κB activation often exhibit constitutive phosphorylation of STAT3, potentially through an IL-6 stimulated event (20). Despite these findings, the underlying molecular links between genetic alteration and malignancy have not been defined.

In the study described herein, we observed strong activation of the NF-κB/IL-6/STAT3 pathway in human hemangioma and angiosarcoma lesions. Using Ink4a/Arf null FVB mice that recapitulate genetic and molecular events of angiosarcoma tumorigenesis, we asked how regulating between NF-κB and STAT3 signal pathways and targeting of the NF-κB pathway, specifically in angiosarcoma tumor cells versus myeloid cells, would affect growth of angiosarcoma lesions as compared with systemic inhibition of NF-κB using a small molecule Ikkβ inhibitor.

Detailed breeding and genotyping are included in Supplemental Methods. Animal care and use procedures were followed according to the protocols approved by the Institutional Animal Care and Use Committee, Department of Animal Care, Vanderbilt University, Nashville, TN.

To characterize the distribution of immune cells in the tumor microenvironment, a single cell suspension was obtained after enzymatic digestion of the primary angiosarcoma tissues of Ikkβ^wt or Ikkβ^Δ/Δ mice. Cells were stained and analyzed by flow cytometry by a BD LSRII FACScan flow cytometry and analyzed using the BD FACSDiva program (for details see Supplemental Methods).

A single cell suspension of angiosarcoma cells (1 × 10⁷cell/mouse) killed by heating to 55°C for 30 minutes or by fixation in 2% paraformaldehyde for 10 minutes was injected intraperitoneally into Ikkβ^Δ/Δ mice, or into control Ikkβ^wt mice. Eighteen hours after injection, the leukocytes infiltrating into the peritoneum were collected, counted and stained with fluorochrome-conjugated antibodies, followed by flow cytometry analysis as described above.

Total RNA was extracted using the RNeasy Mini Kit (Qiagen), and cDNA synthesis was carried out using SuperScript III First-Strand Synthesis System (Invitrogen) according to the manufacturer's instructions. Real-time quantitative PCR was carried out with the BioRad CFX-qPCR instrument using the SsoFast EvaGreen Supermix assay (BioRad), with β-actin serving as a control gene (see Supplemental Methods for detailed protocols). Two-tailed Student t tests were conducted to calculate the statistical significance of the ΔCt between Ikkβ wild type and Ikkβ deleted samples. Primers used in this study are listed in the Supplemental Methods.

For in vitro neutralization, 5 × 10⁴ angiosarcoma cells/mL/well (12-well plate) were cultured in DMEM/F-12 medium containing 1% FBS and with increasing concentrations (0 to 1000 ng/mL) of IL-6 antibody (Clone number MP5-20F3, R&D Systems). Subsequently 200 μL of protein A/G Plus-agarose (Santa Cruz Biotechnology, Inc.) was added and samples were rotated overnight at 4°C to remove antibody-bound IL-6. IL-6 protein remaining was detected by ELISA (R&D System). The antibody concentration that totally removed the IL-6 in the conditioned medium was used as the “neutralizing antibody concentration”. For in vivo neutralization, FVB strain mice with subcutaneous inoculated angiosarcoma cells (1 × 10⁷) were administrated 500 μg of rat antibody to mouse IL-6 by intraperitoneal injection twice weekly for 3 weeks. Matched control mice (n = 5) received the same concentration of rat IgG1.

Cytokine array using RayBiotech mouse cytokine antibody array G series 2 Kit (32 cytokines, RayBiotech Inc.) and ELISA assay for IL-6 (R&D Systems) were carried out as previously described (ref. 21; see Supplemental Methods for details).

The percentage of early stage apoptotic cells was evaluated using the Total Cytotoxicity Kit (Immunochemistry Technologies, LLC) according to the manufacturer's directions. The early apoptotic cells are defined as those that are SR-FLICA positive, but 7-AAD negative.

The nitric oxide (NO) content in peripheral blood and angiosarcoma containing blood was determined using the QuantiChrom Nitric Oxide Assay Kit (DINO-250, Bioassay Systems) following the manufacture's protocol.

Immunoblotting analyses of cytoplasmic extracts from cultured primary angiosarcoma cells were carried out as previously described (21).

NF-κB transcription factor binding sites in the promoters of gp130 and Jak2 were identified using the program at http://www.cbrc.jp/research/db/TFSEARCH.html. Three copies of the gene specific NF-κB promoter sequence from the Jak2 or gp130 promoters were cloned into a Gaussian luciferase reporter and activity assays were carried out after transient transfection using protocols detailed in Supplemental Methods

Data are expressed as mean ± SEM with sample size per experimental group specified. Prism software (GraphPad), Statistical Analysis System (SAS), and R software were used for statistical analyses and graphics. Unless specified otherwise, the 2-sample Student t test with Satterthwaite's approximation was used to determine the statistical difference between two independent experimental groups. The cumulative incidence of spontaneous angiosarcoma between groups was calculated using the method of Kaplan and Meier and compared using the log-rank statistic. A 2-sided P value ≤ 0.05 was considered to indicate a significant difference between groups.

Human vascular tissue microarray from US Biomax and MD Anderson Cancer Center (University of Texas, Houston, Texas) were examined for RelA/p65, IL-6, CD31, and pSTAT3 protein expression in samples of 10 normal blood vessels, 55 benign hemangiomas, and 55 angiosarcomas. Tissues were obtained with informed consent according to the IRB guidelines of the MD Anderson Cancer Center. Immunofluorescence was quantitated in relation to nuclear or cytoplasmic staining with DAPI visualization of nuclei. Strong expression of RelA/p65 was seen in 10% of normal blood vessels, which increased to 49% in benign hemangiomas (P = 0.034), and 58% in angiosarcoma (P = 0.006). Neither nuclear pSTAT3 staining nor strong IL-6 was expressed in normal blood vessels, whereas 33% of the benign hemangiomas exhibited strong nuclear pSTAT3 staining (P = 0.05) and 30% of angiosarcomas (P = 0.097) exhibited strong IL-6 expression (Fig. 1). Thus, RelA/p65 expression is present in normal vascular tissue and increases with neoplastic and malignant transformation, which is associated with the increased IL-6 expression and nuclear STAT3 translocation in the transforming or transformed lesion. However, our observation that high levels of IL-6 and pSTAT3 were not simultaneously expressed in the same type of lesion indicates the complexity of signal transduction pathway activation during angiosarcoma development.

Ink4a/Arf deficiency in mice has been shown to be associated with the development of fibrosarcoma, lymphoma and angiosarcoma lesions (22, 23). Sharpless and colleagues showed that angiosarcoma is a common tumor for animal germline deficient for both Ink4a and p53 (24). In prior studies we observed that loss of Ink4a/Arf in C57Bl/6 mice resulted in a high incidence of spontaneous lymphoma and fibrosarcoma lesions (25). In the study described herein we observed that the FVB strain of Ink4a/Arf^−/− mice exhibits a high incidence of angiosarcoma lesions, with an approximate incidence of 30% within 100 days. Once the angiosarcoma lesions are visible, they grow rapidly, whereas tumors did not develop in the wild type Ink4a/Arf FVB mice. The angiosarcomas in Ink4a/Arf^−/− mice occur in association with the vasculature of muscle, liver, lung, chest wall, spleen, intestine, and skin (Fig. 2A, a–g, respectively). Splenic angiosarcoma (Fig. 2A, e) was associated with greatly enlarged spleen. The splenic angiosarcomas frequently undergo metastasis to liver (Fig. 2A, h). Histological H&E staining of angiosarcoma primary lesions arising in subcutaneous muscle tissue, spleen, liver, and lung (Fig. 2B, a–d, respectively) shows lesions are comprised of large islands of blood cells, in sharp contrast to the normal tissues of liver (Fig. 2B, e) and lung (Fig. 2B, f). Angiosarcoma lesions exhibit vasculature disorder but stain positive for Von Willebrand Factor (VWF; Fig. 2C, a) and CD31 (Fig. 2C, c), consistent with poorly differentiated angiosarcoma in patients (4, 26, 27). Lesional staining for VWF is contrasted to the staining of intact blood vessels in lung (Fig. 2C, b).

The formation of angiosarcoma lesions has been associated with reduced expression of inhibitors of angiogenesis and enhanced expression of angiogenic factors such as IL-8, IL-6, VEGFA, and bFGF (5). Serum cytokines in angiosarcoma bearing versus nontumor bearing mice were examined with a G Series 2 32 cytokine array. Serum IL-6 was 11 ± 2.5-fold higher in tumor bearing compared with tumor-free mice (n = 6 Fig. 2D) and this was the sole cytokine present at significantly different levels in tumor bearing versus nontumor bearing mice. Analysis of the IL-6 levels in angiosarcoma tumor tissue, as compared with adjacent nontumor muscle tissue, spleen, bone marrow and thymus, revealed very high levels of IL-6 in angiosarcoma lesions compared with normal adjacent tissues (869 ± 165 pg/g protein vs. 10 ± 2.6 pg/g protein, n = 6; Fig. 2E). IL-6 has been shown to activate endothelial progenitor cells and play a crucial role in angiogenesis and vascular remodeling in vitro and in vivo (28, 29). Thus it is plausible that IL-6 contributes to transformation of endothelium into malignant angiosarcoma when Ink4a/Arf is deleted, recapitulating genetic traits of human angiosarcoma.

Angiogenesis and vasculogenesis are supported by cytokine mediated cross-talk between tumor cells and endothelial cells in the tumor microenvironment. To evaluate the role of tumor secretion of IL-6 on tumor cell viability, IL-6 antibody or control isotype IgG was added to cultured angiosarcoma cells in increasing concentrations to neutralize secreted IL-6 (Fig. 3A). The treated cells were then collected, stained with SR-FLICA and subjected to FACS analysis to determine the percentage of apoptotic cells. Treatment with 100 and 1000 ng/mL IL-6 antibody (P < 0.01) markedly increased the percentage of apoptotic cells compared with IgG controls (Fig. 3B) and induced cell death, based 7-AAD staining (Supplementary Fig. S1A; P < 0.01).

To investigate whether anti-IL-6 induced apoptosis in vivo and affected tumor progression, angiosarcoma cells were subcutaneously inoculated into FVB mice. When tumor size reached around 180 mm³, mice received anti-IL-6 antibody or the same amount of isotype IgG antibody as a control (30, 31), After 6 days treatment, the anti-IL-6 treatment induced apoptosis as indicated by the cleaved PARP, but not in IgG control treatment (Fig. 3C), reduced cell proliferation (Supplementary Fig. S2C) and reduced tumor volume 1.9 (P < 0.01, n = 6, Supplementary Fig. S1B). In another series of experiments, mice were treated with antibody immediately following tumor transplant. After 3 weeks of treatment, the tumor volume in IL-6 Ab treated animals was 4-fold reduced in comparison tumor grafts of IgG control treated animals (188 ± 69 mm³ vs. 792 ± 151 mm³, P < 0.01; Supplementary Fig. S1C and S1D). Antibody neutralization of IL-6 in the blood of tumor-bearing mice was confirmed by ELISA (26.4 ± 25.6 pg/mL in IL-6 antibody treated vs. 676 ± 118 pg/mL in IgG treated mice (Supplementary Fig. S1E). Anti-IL-6 not only induced apoptosis, but also reduced NF-κB transcriptional activity based upon luciferase reporter signaling in angiosarcoma cells (P < 0.01, n = 6, Supplementary Fig. S1F).

To learn the impact of anti-IL-6 on the cytokine profile, serum samples from mice described in Supplementary Fig. S1B were subjected to cytokine array (Fig. 3D). After 6 days treatment, the serum cytokine profile was analyzed using a cytokine array (32 cytokines) approach. The value of cytokines was compared with the serum sample from a tumor-free mouse and expressed as fold-change. Results showed the active form of IL-12, p40, was markedly elevated in tumor bearing mice compared with nontumor bearing mice (24-fold for IgG1 vs. 49-fold anti-IL-6), and this elevation was 2-fold higher in mice treated with anti-IL-6 as compared with those treated with control IgG1. Other serum cytokines were elevated in tumor bearing mice IgG1 treated mice compared with nontumor bearing mice, but anti-IL-6 antibody reduced the elevation: RANTES (8.9-fold vs. 2.8-fold), IL-6 (7.9-fold vs. 0.1-fold), and MCP-5 (6.6-fold vs. 0.8-fold), as indicated in Fig. 3D. The IL-6 in the serum of tumor-bearing mice was biologically active based on induction of Stat3 phosphorylation in splenic cells (Supplementary Fig. S2A). Moreover, the levels of phosphorylated Stat3 declined when IL-6 was neutralized by IL-6 antibody (Supplementary Fig. S2B). Not surprisingly, neutralization of IL-6 statistically impacted the tumor infiltrating immune cells by decreasing the macrophage population (P = 0.044), increasing B cells (P = 0.012), and increasing CD4+ T cells (P = 0.028), in comparison with the IgG1 isotype control group (n = 6, Student t test). There were no significant changes on the neutrophils (P = 0.139) or CD8+T cells (P = 0.187) between the treated and control animal tumors (Supplementary Fig. S2D). Data here revealed that tumor-secreted IL-6 is not only a major effector on angiosarcoma growth through inhibitory apoptosis, but also is a regulator in tumor immunity. As IL-6 is transcriptionally regulated in part by NF-κB, we asked whether inhibition of NF-κB may effectively inhibit IL-6 expression and angiosarcoma growth.

Disruption of the Ink4a/Arf gene may be causally linked to angiosarcoma through an NF-κB dependent mechanism because p16^INK4a and p19^ARF are negative regulators of NF-κB, positive regulators of p53 (32), and act as inhibitors of cyclin-dependent kinases (33). To determine how depletion of Ikkβ would affect angiosarcoma cells, knockdown of Ikkβ was carried out by stable expression of a lentiviral short hairpin RNA (shRNA) targeting Ikkβ. With this approach, 95% of the Ikkβ protein was knocked down whereas Ikkγ was intact (Fig. 4A). When Ikkβ was knocked down in the angiosarcoma cells there was a decline in p65 phosphorylation on Ser-536, Stat3 phosphorylation on Tyr-705, and membrane gp130 expression, whereas p53 protein levels increased (Fig. 4A).

Silencing Ikkβ in cultured angiosarcoma cells abrogated IL-6 secretion into the medium (4.7 ± 4.6 pg/mL, P < 0.01, n = 3), in contrast to IL-6 (2233 ± 235 pg/mL) secreted from the same number of angiosarcoma cells with intact Ikkβ protein (Fig. 4B). Of interest, silencing of Ikkα did not affect IL-6 secretion (2755 ± 369 pg/mL vs. 2233 ± 235 pg/mL, P = 0.17, n = 3) by angiosarcoma cells. Thus, Ikkβ is an important mediator in the NF-κB pathway that is also essential for IL-6 production by angiosarcoma.

As Ikkβ interference reduced phosphorylation of Stat3 in angiosarcoma cells (Fig. 4A), we asked whether Ikkβ/NF-κB signaling impacted the intrinsic expression of gp130, Jak2 or the sIL-6R. Towards this end, cell surface gp130 was examined by FACS. Silencing Ikkβ in angiosarcoma cells resulted in 92.8% reduction of cell membrane gp130 (Fig. 4C) that was accompanied by a 91% reduction gp130 protein based upon Western blotting (Fig. 4D), a marked suppression of gp130 mRNA based upon qRT-PCR (15.6-fold; Fig. 4E), and a dramatic inhibition of gp130 NF-κB-mediated promoter activity, P < 0.05; Fig. 4F). Similarly, silencing Ikkβ in angiosarcoma cells resulted in an 89% reduction in JAK2 protein (Fig. 4G), a 26.8-fold reduction in Jak2 mRNA (Fig. 4H), and dramatic reduction in the JAK2 NF-κB-regulated promoter activity, P < 0.01 (Fig. 4I) Thus, constitutive activity of the Stat3 pathway in angiosarcoma cells is tightly controlled by Ikkβ/NF-κB signaling through intrinsic transcription regulation of gp130/Jak2.

As there is no detectable IL-6Rα expressed by the angiosarcoma cells (data not shown), we evaluated expression of the soluble IL-6 receptor (sIL-6Rα), which like the IL-6Rα can interact with gp130 to initiate signaling in response to IL-6. We determined that the major source of the sIL-6Rα is from mouse blood (848 ±198 pg/mL, Fig. 4J) whereas angiosarcoma cells secrete a more modest quantity of the sIL-6Rα (126 ± 8.5 pg/mL, Fig. 4K). There was no detectable sIL-6Rα in the serum-containing medium not exposed to the angiosarcoma cells. These data show that disruption of Ikkβ reduced expression of the sIL-6Rα in vitro and in vivo.

Human tissue microarray shows that RelA/p65 is highly expressed in benign and malignant human vascular tumors (Fig. 1). Angiosarcoma cells depleted of Ikkβ through expression of the Ikkβ shRNA grew ∼4.5 times slower in vitro than cells expressing the nonsilencing shRNA (Fig. 5A), indicating that Ikkβ is crucial for the growth of Ink4a/Arf^−/− angiosarcoma cells. However, the reduction in cell growth in response to knockdown of Ikkβ led to the slow cell growing in vitro was not associated with enhanced apoptosis (P = 0.53, Fig. 5B). To determine whether similar effects in vitro would be observed in vivo, FVB mice were subcutaneously injected with Ink4a/Arf deficient angiosarcoma cells stably expressing lentiviral shRNA Ikkβ or shRNA empty vector as a control. Three weeks after injection of cells, angiosarcomas lesions were observed in all control mice, but no tumors were observed in mice injected with Ikkβ-depleted angiosarcoma cells (Fig. 5C), indicating that Ikkβ is crucial for angiosarcoma tumorigenesis resulting from deficiency of p16^Ink4a/p19^Arf. Thus, there is potential for treating angiosarcoma lesions by inhibiting IKKβ. However, because IKKβ also affects the antitumor or protumor capacity of tumor associated myeloid cells (34), it was important to access how loss of IKKβ would affect the myeloid cells in angiosarcoma tumor bearing mice.

Bone marrow myeloid cells were isolated from wild type FVB mice and evaluated for chemotactic response to conditioned medium from IKKβ expressing or Ikkβ depleted angiosarcoma cells. Flow cytometry data indicate that the population of purified bone marrow cells consisted of 38% B cells, 21% PMN and immature myeloid cells, 8% macrophages, 4% dendritic cells, and 29% other types of cells, such as stroma cells, stem cells, and endothelial cells. These cells showed robust chemotaxis in response to medium conditioned by Ikkβ expressing and IKKβ deficient angiosarcoma cells, but medium from Ikkβ expressing angiosarcoma cells produced a much stronger chemotactic response (Supplementary Fig. S3A). There was no obvious difference in cell migration in response to increasing concentrations of murine IL-6, suggesting chemotaxis in response to conditioned medium is due to secretion of NF-κB regulated factors other than IL-6 (Supplementary Fig. S3B).

Cells of the myeloid lineage, especially mononuclear phagocytes, are crucial for establishment of innate immunity and cytokine regulation of acquired immune responses (35). Moreover, it has been reported that NF-κB regulates the shift of immune cells from an antitumor to a protumor phenotype (34). To determine the role of NF-κB in myeloid cells on the immune response to the tumor we disrupted Ikkβ only in myeloid cells by breeding mice carrying loxP-flanked Ikkβ allele with LysMCre mice in which the Cre-recombinase is expressed under the control of the murine lysozyme-M gene regulatory region (Fig. 6A). Upon breeding of FVB-LysMCre mice to FVB mice harboring loxP-flanked target genes, these animals efficiently undergo Cre-loxP-mediated recombination in macrophages and neutrophils, but not in B and T cells or in the majority of dendritic cells (36). To examine the efficiency of LysMCre mediated deletion of the loxP-flanked Ikkβ allele in tumor tissue, Ikkβ^wt (LysMCre::Ink4a/Arf^−/−) mice were bred with mT/mG (membrane-Tomato/membrane-Green) mice to obtain an additional genomic background of Gt(Rosa)26Sortm4(ACTB-tdTomato-EGFP) in which the loxP-flanked tdTomato following the EGFP (mG) cassette was inserted into the Gt(ROSA)26Sor locus. These mT/mG mice served as a Cre-reporter strain (37). Myeloid cells from mice carrying this insertion shift fluorescence from red to green after Cre-mediated recombination to yield green Ikkβ^−/− myeloid cells. The GFP-positive infiltrated cells were sorted from angiosarcomas lesions arising in 4 mice with a background of homogenous mT/mG, positive for LysMCre, and deficient for Ink4a/Arf genes, using specific cell marker antibodies noted in Materials and Methods. The percentage of GFP positive cells infiltrating the tumor reflects the specificity and efficiency of the LysMCre mediated Ikkβ deletion in the cells (Supplementary Fig. S3C).

To examine the impact of myeloid Ikkβ on angiosarcoma development in Ink4a/Arf null mice, tumor incidence was recorded in 63 Ikkβ^Δ/Δ (Ikkβ^F/F::LysMCre::Ink4a/Arf^−/−) mice and 73 Ikkβ^wt (LysMCre::Ink4a/Arf^−/−) mice over a 4-month period. There was a significant increase in the cumulative incidence of angiosarcoma (48%) in Ikkβ^Δ/Δ mice when compared with that (27%, P < 0.01) in the control Ikkβ^wt animals (Fig. 6B). Moreover, the latency of angiosarcoma arising in Ikkβ^Δ/Δ mice (89 ± 17 days, n = 31 of 63) was shorter than in Ikkβ^wt mice (103 ± 17 days, n = 20 of 73; Supplementary Table S1). Thus, myeloid Ikkβ positively contributes to antitumor immunity in this mouse tumor model.

The percentage of tumor-infiltrating neutrophils (CD45+CD11b+Ly6G+) was significantly lower in angiosarcoma tissues of Ikkβ^Δ/Δ mice than in Ikkβ^wt mice (P < 0.01, n = 5; Fig. 5C). However, there was no statistical difference between tumor-bearing Ikkβ^wt and Ikkβ^Δ/Δ mice for tumor-infiltrating macrophages (CD45+CD11b+Ly6G-), dendritic cells (CD45+CD11C+), B cells (CD45+B220+), and T cells (CD45+CD4+, CD45+CD8a+; Fig. 6C and D). Tumor infiltrating immunocytes can produce toxic products such as NO to kill tumor cells. To determine whether NO release by infiltrating myeloid cells is important in this mouse angiosarcoma model, NO levels were determined in the peripheral blood and tumor blood of mice. Measurements of NO in both peripheral and intratumoral blood revealed lower levels of NO in Ikkβ^Δ/Δ mice as compared with Ikkβ^wt counterparts (Fig. 6E), implying that neutrophil-derived NO contributions to host antitumor response were blunted when myeloid cells lacked IKKβ. To investigate the impact of myeloid Ikkβ on tumor cell viability in vivo, angiosarcoma cells derived from primary angiosarcoma of FVB mouse deficient for Ink4a/Arf were genetically engineered to stably express nonsecreted Gaussian luciferase (18–285aa). These cells were injected intraperitoneally into Ikkβ^wt or LysM-Cre Ikkβ^Δ/Δ mice. Eighteen hours after injection, peritoneal cells were collected, lysed and the luciferase activity was determined. Enhanced tumor cell survival, as reflected by an increase in luciferase activity, indicates a reduction in antitumor activity of recipient host with deletion of myeloid Ikkβ (Fig. 6F). Thus neutrophil-derived NO contributions to host antitumor response were blunted when myeloid cells lacked IKKβ. These data indicate that infiltrated cells play an important role in tumor immunity through a mechanism involving NO.

To investigate the pro-tumor mechanism of CD11b+ cells null for Ikkβ, the intratumoral neutrophils and macrophages were sorted by flow cytometry and the intracellular cytokine profile was examined. Deletion of myeloid Ikkβ resulted in elevated IL-4 and reduced IL-12 and IFNγ in both neutrophils and macrophages (P < 0.01, n = 3; Supplementary Fig. S4A and B). Simultaneously, cellular arginase I increased 3.8-fold in neutrophils (P = 0.024, n = 5, by Student t test) and 6.3-fold in macrophages (P = 0.011, n = 4, by Student t test), compared with Ikkβ wild type cells (Supplementary Fig. S4C). Together, these data indicate that intact Ikkβ in myeloid cells is necessary for the antitumor phenotype of myeloid cells in this angiosarcoma model. Moreover, data suggest that loss of IKKβ in myeloid cells results in a shift to the N2/M2 protumorigenic type of leukocytes, with enhanced tumor growth.

A need for intrinsic NF-κB for angiosarcoma formation suggests that IKKβ is a promising therapeutic target. However, the antitumorigenic effects of IKKβ in myeloid cells raised the issue of which role for IKKβ would predominate if small molecule IKKβ inhibitors were used. To address this question, we systemically delivered BMS-345541, a highly selective inhibitor of Ikkβ, to mice bearing angiosarcoma xenografts. When the tumor size reached ∼250 mm³ (∼2 weeks postimplantation), mice were treated with 75 mg/kg BMS-345541 twice per day and the control tumor bearing mice received a similar volume of vehicle only. After 1 week of drug treatment, tumor growth of the BMS treated mice was significantly reduced compared with vehicle treated controls (307 ± 80 mm³ vs. 548 ± 119 mm³, respectively, P < 0.01, n = 10). A similar result was obtained after 2 weeks treatment, where tumor volume was 392 ± 88 mm³ versus 1098 ± 200 mm³ in treated versus control, respectively (mean ± SEM, P < 0.01, Student t test; Fig. 7A). Upon systemic targeting of Ikkβ with BMS345541, the angiosarcoma tumor tissue showed obvious nuclear apoptotic features, in sharp contrast to control tumors treated with vehicle alone (Fig. 7B). Moreover, IL-6 levels markedly declined with systemic BMS345541 treatment (Fig. 7C) and results in reduced response to exogenous IL-6 through reduced expression of gp130 and Jak2 (Fig. 7E). In addition, inhibition of IKKβ by BMS-345541 resulted in reduction of Stat3 phosphorylation in immune cells (Supplementary Fig. S5A). Knockdown of Stat3 in angiosarcoma cells (Supplementary Fig. S5B) was associated with the reduced IL-6 production (Supplementary Fig. S5C). Systemic inhibition of Stat3 with 10 mg/kg S31–201 achieved antitumor activity, as evidenced by 71% inhibition of tumor growth (P < 0.01, n = 6, Supplementary Fig. S5D). Thus, Ikkβ appears to be a potential target for treatment of angiosarcoma lesions especially because systemic inhibition of IKKβ overcomes the tumor promoting phenotype observed when Ikkβ is deleted only in myeloid cells (Fig. 7D).

In work we described here, a genetically modified mouse model was used to recapitulate genetic and epigenetic events that characterize human malignant angiosarcoma and to provide insight for new therapeutic strategies for treatment of malignant angiosarcoma. Immunofluorescent analysis of tissue microarrays of human blood vessels, benign hemangioma, and malignant angiosarcoma lesions reveal a trend toward higher levels of expression of RelA/p65 and IL-6 in metastatic angiosarcoma. High expression of RelA/p65 and IL-6 are accompanied by nuclear localization of phospho-STAT3. In this study, disruption of Ink4a/Arf genes in FVB mice is associated with spontaneous angiosarcoma formation in multiple organs and metastasis with activation of the Ikkβ/NF-κB/IL-6/Stat3 pathway, thus recapitulating major traits seen in human angiosarcoma lesions. Knocking down Ikkβ or antagonizing IL-6 in cells from spontaneously arising angiosarcoma resulted in a dramatic antitumor effect. Moreover, systemic inhibition of the NF-κB pathway allowed suppression of the NF-κB/IL-6/STAT3 signaling pathway and inhibition of tumor growth. Surprisingly, when Ikkβ was deleted only in myeloid cells, angiosarcoma growth was enhanced and the myeloid cells that infiltrated the tumor exhibited a pro-tumor or M2-like phenotype. This provides the first clear demonstration that activation of the NF-κB pathway in myeloid cells is important for the antitumor M1-like phenotype. However, because systemic inhibition of Ikkβ effectively inhibited tumor growth, our data indicate that NF-κB activity in other cells is required for the tumor promoting effect of Ikkβ^−/− myeloid cells. Our data offer hope for new therapies for angiosarcoma patients and new insights into the role of NF-κB in regulation of immune response to tumor.

Mouse angiosarcoma cells produce a large amount of IL-6 and mice with angiosarcoma exhibit elevated serum IL-6, mimicking the high serum IL-6 in angiosarcoma patients, which was speculated to be as associated with clinical malignancy, angiogenesis, vascular remodeling, and metastasis (18). The human IL-6 gene (5kb) is located on chromosome 7p20 and the IL-6 promoter contains a NF-κB binding site between nucleotides 75–63 upstream of the IL-6 mRNA (38). Inflammatory factors can activate NF-κB to drive IL-6 production (39). Constitutive activation of NF-κB has been observed frequently in a variety of malignant cancers, including melanoma, breast cancer, colon cancer, leukemia, lymphoma, myeloma, and HNSCC (40). The aberrant NF-κB activity is often associated with the hyper-expression of IL-6 (41, 42). We observed that Ikkβ, but not Ikkα, is responsible for the IL-6 production in the angiosarcoma cells, based upon our finding that stable knockdown of Ikkβ in the angiosarcoma cells blocks IL-6 production and release, whereas knockdown of Ikkα enhances IL-6 secretion by angiosarcoma cells in vitro. Similarly, inhibition of NF-κB by the super-repressor of IκB (S32/36A) abrogated the IL-6 paracrine activation of STAT3 in HNSCC cells (41). Moreover, when Ikkβ was knocked-down in angiosarcoma cells in vitro, there was a reduction in expression of gp130 and the sIL-6R by angiosarcoma cells and exogenous IL-6 failed to induce phosphorylation of RelA/p65 at Ser 536 and phosphorylation of Stat3. These data are in agreement with prior work showing a strong link between IKKβ and IL-6 in tumorigenesis in colon, hepatocellular, pancreatic, and esophageal carcinomas (43, 44). Thus, our combined data indicate that Ikkβ is vital for the IL-6 autocrine/paracrine activation of Stat3 and NF-κB in angiosarcoma. Moreover, we provide key novel data showing that the mechanism by which suppression of Ikkβ results in a reduced response to exogenous IL-6 is through reduced expression of gp130 and Jak2 (Fig. 7E).

Mounting evidence indicates that inflammation promotes malignant transformation. In the cancer microenvironment, cross-talk between cancer cells and infiltrating immune cells is linked by both extrinsic (cytokine mediated paracrine or autocrine) and intrinsic (gene mutation) pathways. In this mouse model NF-κB/IL-6/Stat3 signaling becomes a major extrinsic pathway for cancer inflammation in association with an intrinsic loss of p16^Ink4a/p19^Arf. IL-6 drives Jak mediated activation of Stat3, presumably in both angiosarcoma cells and in immune cells recruited into the tumor microenvironment. NF-κB activity is maintained by the feed-forward (autocrine) loop. Thus, both NF-κB and Stat3 have central and integrated roles in the inflammation of cancer. These transcription factors are involved in both procarcinogenic inflammation and in antitumor immune responses (45). Here we show that myeloid Ikkβ is linked to antitumor immunity. Moreover, in the in vivo tumor mediated myeloid chemotaxis assay, deletion of myeloid Ikkβ also impairs the migration of both neutrophils and macrophages. Thus, NF-κB and Stat3 interact at multiple levels during tumorigenesis.

The clinical prognosis for angiosarcoma is poor and current therapeutic options offer limited hope. There is a crucial need for an effective therapy. Recent phase II clinical trials targeting of Raf/MAPK pathway with sorafenib (46) and BCR-ABL/VEGF pathway with imatinib (47) have been conducted in patients with advanced or metastatic angiosarcoma. The efficacy was limited, with response rates between 12% and 13%. Large clinical trials with the chemotherapeutic agents doxorubicin (26, 48) and taxane (49) show only transient or ineffective responses in metastatic angiosarcoma. We observed that targeting Ikkβ systemically with small molecule inhibitor or with antibody to IL-6 in angiosarcoma bearing immunocompetent mice resulted in dramatic reduction of tumor growth. Together with other evidence that IL-6/Stat3/HIF plays major role in human clear carcinoma (50), data presented here support the Ikkβ/NF-κB/IL-6/Stat3 pathway as novel molecular targets for angiosarcoma therapy and enhance our understanding of how disrupting the balance between IKK/NF-κB in tumor and inflammatory cells promotes tumor development and growth.

No potential conflicts of interest were disclosed.

Conception and design: M. Boothby, A. Richmond

Development of methodology: J. Yang, A. Lazar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Yang, J. Sai, O. Hawkins, E. Young, E.G. Demicco, A. Lazar, D. Lev

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Kantrow, J. Sai, M. Boothby, G.D. Ayers, E.G. Demicco, A. Lazar, D. Lev, A. Richmond

Writing, review, and/or revision of the manuscript: J. Yang, S. Kantrow, M. Boothby, G.D. Ayers, E. Young, A. Lazar, D. Lev, A. Richmond

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.D. Ayers, A. Lazar

Study supervision: A. Richmond

We thank Linda W. Horton, Krystle Fordyce and Melissa Downing for excellent technical assistance and immunostaining. We thank Michael Karin (UCSD) for plasmids of pLSLPw shRNA Ikkα and Ikkβ and for the IKKβ floxed mice, Lynda Chen (DFCI) for the INK4a/ARF null mice, and Li Yang and Hal Moses for the LysM-Cre mice and Fiona Yull for guidance on genotyping mice. We thank Gary A. Sulikowski in the VUMC Chemical Biology Core for synthesizing the BMS-345541. This work was supported by research grants from the NIH: CA 116021 (AR), and 5P30CA068485 and a Merit Award the Department of Veterans Affairs (AR, 1IO1BX000196) and a VA Senior Research Career Scientist Award (AR).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.