Angiosarcoma is a malignant vascular tumor originating from endothelial cells of blood vessels or lymphatic vessels. The specific driver mutations in angiosarcoma remain unknown. In this study, we investigated this issue by transcriptome sequencing of patient-derived angiosarcoma cells (ISO-HAS), identifying a novel fusion gene NUP160–SLC43A3 found to be expressed in 9 of 25 human angiosarcoma specimens that were examined. In tumors harboring the fusion gene, the duration between the onset of symptoms and the first hospital visit was significantly shorter, suggesting more rapid tumor progression. Stable expression of the fusion gene in nontransformed human dermal microvascular endothelial cells elicited a gene-expression pattern mimicking ISO-HAS cells and increased cell proliferation, an effect traced in part to NUP160 truncation. Conversely, RNAi-mediated attenuation of NUP160 in ISO-HAS cells decreased cell number. Confirming the oncogenic effects of the fusion protein, subcutaneous implantation of NUP160–SLC43A3-expressing fibroblasts induced tumors resembling human angiosarcoma. Collectively, our findings advance knowledge concerning the genetic causes of angiosarcoma, with potential implications for new diagnostic and therapeutic approaches. Cancer Res; 75(21); 4458–65. ©2015 AACR.
Angiosarcoma is a malignant vascular tumor originating from endothelial cells of blood vessels or lymphatics vessels. The tumors frequently occur in the liver, breast, and skin. Cutaneous angiosarcoma usually arises on the scalp of elderly individuals, and rapidly metastasizes. We previously established cultured human cutaneous angiosarcoma cell line, ISO-HAS, to clarify the malignant characteristics of the tumor, and found that the cells harbor a TP53 point mutation (1). However, there may be more specific driver mutations in angiosarcoma. Trisomy 5 and a loss of the Y chromosome were reported in a single case of angiosarcoma arising within a venous malformation (2). Antonescu and colleagues (3) reported some patients with breast angiosarcoma who harbored KDR mutations. More recently, mutations in PTPRB and PLCG1 were identified in the tumor tissues of angiosarcoma patients (4). Although these changes seem to be functional, they are not found in all patients, and the mechanism by which they contribute to tumorigenesis is unclear.
In the present study, we describe a novel frequent fusion gene, NUP160–SLC43A3, in cutaneous angiosarcoma patients.
Materials and Methods
ISO-HAS was isolated from a tumor tissue as described in 1992 (1). Adult human dermal microvascular endothelial cells (HDMEC) or NIH3T3 cells were obtained from Lonza or ATCC in 2009, respectively. Cells were tested and authenticated by DNA fingerprinting in May 2015.
High-quality RNA for transcriptome sequencing was obtained from cultured cells using the RNeasy Mini Kit. The RNeasy FFPE Kit was used for RNA extraction from paraffin-embedded sections. Institutional review board approval and written informed consent were obtained.
Library preparation and transcriptome analysis
Transcriptome analysis was performed in accordance with the protocol provided by Riken Genesis using the Illumina TruSeq RNA Sample Preparation Kit. Final cDNA libraries were sequenced on Illumina HiSeq2000 platform. The clean and trimmed reads were aligned to GRCh37/hg19 using TopHat. The differential expression between the samples was analyzed with Cuffdiff by calculating the fragments per kilobase per million map reads (FPKM). Bioinformatic analysis for detecting mutations was performed using Samtools. Potential gene fusion transcripts were identified and filtered by TopHat-Fusion.
For RT-PCR, cDNA, and primers were mixed with SYBR Premix Ex Taq II. PCR was performed for 50 cycles of denaturation for 5 seconds at 95°C, annealing for 30 seconds at 60°C, and extension for 1 minute at 68°C. Mutations of TP53 or KDR were detected as described previously (3, 5, 6).
For PCR array, cDNA was mixed with RT2 SYBR Green/Rox qPCR Master Mix, and the mixture was added to Human Angiogenesis PCR Array.
Fluorescence in situ hybridization and G-banding
FISH analysis was performed according to the protocol provided by Nihon Gene Research Laboratories. The cells were fixed with freshly made Carnoy's solution. The probes were designed by Empire Genomics. The slides were treated with probes, and were denatured by heating at 72°C for 4 minutes. Hybridization was performed by incubation at 37°C overnight. The nuclei were stained with DAPI.
For G-banding, slides were treated with 0.2% trypsin in Hanks' Balanced Salt Solution. The cells were then washed immediately with PBS, and were stained for 5 minutes with Giemsa solution.
Migration assays were performed using 24-well plate Transwell inserts with a polycarbonate filter membrane. Cells were detached by trypsin-EDTA, and were seeded into the upper compartment of Transwell chamber in 100 μL of serum-free medium. The lower chamber was filled with 10% FBS-containing culture medium as a chemoattractant. After 24 hours, the cells migrated to the undersurface of filter membrane were counted in five different 400-fold magnification fields under an inverted microscope.
Cell count and BrdUrd ELISA
Cells were detached from the wells by trypsin treatment and counted using Coulter Particle Counter. Cell proliferation activity was confirmed using the bromodeoxyuridine (BrdUrd) ELISA kit.
Cell lysates were precleared, and then incubated with an anti-SLC43A3 antibody and Protein A/G Plus Agarose overnight at 4°C. Agarose-bound proteins were extracted by incubation in sample buffer at 95°C. The immunoprecipitates were subjected to electrophoresis on sodium dodecyl sulfate-polyacrylamide gels and transferred onto polyvinylidene difluoride (PVDF) filters. The PVDF filters were incubated with primary antibodies, followed by incubation with secondary antibodies.
Lentiviral gene transfer
Lentiviral vector–mediated gene transfer was performed using CSII-EF-RfA, pCMV-VSV-G-RSV-Rev, and pHIVgp (7).
For reverse transfection, siRNAs mixed with Lipofectamine RNAiMAX was added when the cells were plated.
NIH3T3 cells were implanted into 6-week-old female athymic nu/nu mice by s.c. injection using 25-gauge needles. The implants were removed 5 weeks after xenografting, fixed in 10% buffered formalin, embedded in paraffin, and sliced into sections of 5 μm in thickness.
Data presented as bar graphs are the means + SD from at least three experiments. Values of P < 0.05 were considered to be statistically significant.
Other methods are described in Supplementary Materials and Methods.
Identification of angiosarcoma-specific fusion gene
In gene-expression study by transcriptome sequencing, 94 genes were estimated to be significantly upregulated and 237 genes were estimated to be significantly downregulated in ISO-HAS compared with HDMEC (Supplementary Table S1). These included many angiogenesis-related genes (Supplementary Table S2). Several changes in the expression of angiogenesis-related genes seen in ISO-HAS by the present study (e.g., downregulation of ITGB3, CTGF, and CLDN1) were compatible with the findings of previous reports (8–10).
Next, we determined the presence of ISO-HAS–specific point mutations in five angiogenesis-related genes and 21 oncogenes selected on the basis of previous studies (3, 4). Among six putative angiosarcoma-specific mutations identified by bioinformatic analysis (Supplementary Table S3), the point mutation of TP53, a T-to-A transition in exon7 at codon 240, was already described in ISO-HAS (1). On the other hand, the T-to-C mutation found in PLCG1 of ISO-HAS was different from previously reported PLCG1 mutation (4).
Putative fusion transcripts were predicted by TopHat-Fusion (Supplementary Table S4). Among the 14 candidates, 35 spanning reads and five spanning mate pairs indicated the in-frame fusion between exon26 of NUP160 and exon7 of SLC43A3, which are located at 11p11.2 and 11q12.1, respectively (Fig. 1A). The fusion gene was not found in HDMEC. Specific primers designed for NUP160 exon26 and SLC43A3 exon7 were used for RT-PCR experiments to validate the finding (Fig. 1B). The expected fragment was amplified by PCR using RNA from ISO-HAS, but not RNA from HDMEC, cultured human keratinocytes or dermal fibroblasts. Sanger sequencing of the amplified fragments confirmed the in-frame fusion between NUP160 and SLC43A3 (Fig. 1A). The gene-expression analysis indicated that levels of NUP160 and SLC43A3 were not significantly altered in ISO-HAS compared with HDMEC; the fold-changes in expression were 0.90 (q value; 0.99) and 2.41 (q value; 0.82), respectively.
To characterize the rearrangements at the chromosomal level, multicolor FISH experiments were performed. In HDMEC, two separate green (NUP160) and two separate red signals (SLC43A3) were visible (Fig. 1C). On the other hand, the fusion of one pair of signals, appearing as a single yellow signal, was seen in ISO-HAS. These data confirmed the presence of the NUP160–SLC43A3 fusion, which was observed in 244/300 ISO-HAS (81.3%). Increased single red/green signals were also observed in ISO-HAS, indicating their aneuploidy. G-banding showed the presence of aneuploid nuclei in many chromosomes, including Ch11, of ISO-HAS (Fig. 1D).
To determine whether the NUP160–SLC43A3 fusion was recurrent, we performed fusion-specific RT-PCR using RNA obtained from paraffin sections of human tumors. We screened a series of 25 primary angiosarcomas on the scalp (Supplementary Table S5), and detected the fusion gene in nine cases: For example, amplification by fusion gene–specific primer pair was observed in the RNA samples from patient nos. 16, 17, and 19 (Fig. 2A), which were all confirmed by Sanger sequencing. On the other hand, patient no. 14 had the fusion gene in the tumor tissue, but not in the peripheral blood cells or serum (Fig. 2B), indicating that the rearrangement was somatically acquired. NUP160–SLC43A3 was not found in normal skin samples or in other tumor tissue specimens (Supplementary Table S5).
The clinical and histopathologic characteristics were basically similar between the patients with and without the fusion gene; both poorly (Fig. 2C) and well-differentiated lesions (Fig. 2D) were seen in patients with the fusion. On the other hand, the duration of the disease (between symptom onset and the first hospital visit) was significantly shorter in patients with the fusion gene than in those without (2.9 vs. 9.1 months, P = 0.028 by Mann–Whitney U test), suggesting that the patients with the fusion gene had a more rapid progression and/or more severe subjective symptoms (Supplementary Table S5). In additional, at the first visit, metastasis was less frequently detected in patients with the fusion than in those without (0% vs. 37.5%, P = 0.045 by Fisher's exact probability test), which also indicated that patients with the fusion gene visited the hospital earlier (before metastasis).
Functional analysis of the fusion gene
To clarify the function of the fusion gene, it was amplified by PCR and cloned into a lentiviral vector. HDMECs were stably transfected with the control vector or fusion gene, full-length NUP160 or full-length SLC43A3. To note, the cells transfected with full-length NUP160 did not grow, and exhibited inhibited cell proliferation. HDMECs transfected with the control vector, fusion or SLC43A3, and ISO-HAS showed no apparent differences in cell shape (Fig. 3A).
We determined the effects of the fusion gene on the gene expression using a PCR array (Supplementary Table S6: the complete dataset is available at www.ncbi.nlm.nih.gov/geo/, GSE69725). The transfection of the fusion increased or decreased the expression of several angiogenesis-related genes. Twelve genes were included in both the transcriptome analysis (Supplementary Table S2) and PCR array (Supplementary Table S6). Among these, the up- or downregulation of nine genes was common to both experiments. Thus, there is a possibility that the gene-expression pattern of HDMEC stably transfected with the fusion gene mimicked that of ISO-HAS. We expected that cells transfected with the fusion gene would obtain an “angiosarcoma-like” phenotype.
We then examined the phosphorylation levels of the NUP160–SLC43A3 protein, because NUP214–ABL1 fusion reportedly causes its autophosphorylation. Full-length NUP160 and SLC43A3 have a serine phosphorylation site in exon29 and a tyrosine phosphorylation site in exon9, respectively (Fig. 3B). NUP160–SLC43A3 preserves the SLC43A3 phosphorylation site, but loses the NUP160 site. Immunoprecipitation revealed that all cell types, including HDMEC, ISO-HAS, HDMEC transfected with the control vector and HDMEC transfected with the fusion gene expressed the full-length SLC43A3 protein (Fig. 3C). ISO-HAS and HDMEC transfected with fusion gene expressed the fusion protein, but tyrosine phosphorylation of the fusion was not detected in these cells. Taken together, these findings indicate that the fusion protein is not likely to exhibit increased autophosphorylation.
The migration of ISO-HAS was almost identical to that of HDMEC (Fig. 4A). On the other hand, although the cell growth of ISO-HAS and HDMEC was similar until passage 14, the growth of HDMEC was significantly reduced after passage 18 (Fig. 4B). Considering the potential role of mitogenesis in the development of angiosarcomas described previously (11), we hypothesized that the sustained cell growth of ISO-HAS plays a role in its malignant phenotype, and therefore focused on cell proliferation. Stable overexpression of the fusion gene in HDMEC at higher passages led to the significant recovery of cell numbers and the BrdUrd incorporation (Fig. 4C), whereas full-length SLC43A3 overexpression did not. Thus, the fusion gene may contribute to sustained cell proliferation. Of note, truncated NUP160 (exons1-26) could also increase the number of HDMEC (Fig. 4D). Taken together, our results suggest that NUP160 truncation, and not the phosphorylation of NUP160–SLC43A3, is the cause of the sustained cell proliferation of ISO-HAS. On the other hand, NUP160 siRNA inhibited the expression of the fusion gene (Fig. 4E), as well as the expression of full-length NUP160 (not shown) in ISO-HAS, and significantly decreased the cell number (Fig. 4F), indicating therapeutic value of the siRNA.
Finally, we performed a tumorigenicity assay using NIH3T3 according to a previous study (12). At 5 weeks after subcutaneous implantation, NIH3T3 transfected with the fusion gene showed tumor formation, whereas the cells with the control vector did not (Fig. 4G). The histopathologic findings of the tumor tissue showed a mass of pleomorphic and hyperchromatic cells (Fig. 4H), the features of which were similar to poorly differentiated angiosarcoma (Fig. 2C). We confirmed that the tumor tissues, as well as the cells cultured from them, were positive for the fusion gene by RT-PCR with fusion-specific primers (Fig. 4I).
Among skin tumors or vascular tumors, fusion genes are found in dermatofibrosarcoma protuberans (COL1A1-PDGFB), epithelioid hemangioendotheliomas (WWTR1-CAMTA1 and YAP1-TFE3), and primary cutaneous hidradenomas (TORC1-MAML2, CRTC1-MAML2, MECT1-MAML2, EWSR1-POU5F1, and EWS-Oct-4B; refs. 13–15). Because COL1A1-PDGFB is highly specific to dermatofibrosarcoma protuberans, its detection is clinically useful for the diagnosis. Furthermore, EML4–ALK-driven lung carcinoma shows dramatic response to ALK inhibitors. Thus, the identification of fusion genes in cancers is also important for the treatments. Although the detection of fusion genes, including EWSR1–ATF1 or CEP85L–ROS1, has also been reported in angiosarcoma (16), these were seen in single cases, and were not specific to angiosarcoma. Furthermore, their functions were not determined.
On the basis of the orientation of the genes and the appearance of Ch11, we suppose that interchromosomal or interchromatid rearrangement may cause the NUP160–SLC43A3 fusion. Fusion genes involving the NUP or SLC family have been reported in several tumors. NUP160 is one of the constituent nucleoporins of NUP107-160, the largest subunit of nuclear pore complex (NPC). The NPC is the regulator of molecular bidirectional traffic between the cell nucleus and cytoplasm, and NUP107-160 is a subcomplex essential for various processes, including interphase NPC assembly and mitotic chromosome segregation. Although NUP160 itself is originally implicated in nucleoplasmic transport, the molecule is also associated with hybrid inviability, female sterility, morphologic anomalies, and slow development in vivo (17). On the other hand, SLC43A3 was originally predicted as a transmembrane protein that transports nutrients or metabolites in rapidly growing and/or developing tissues (18). However, its functions have not been confirmed in animal models. Recent articles indicate SLC43A3 as a candidate gene associated with microvascularization, systemic inflammation links to atherosclerosis and thyroid cancers (19, 20). Taken together, although the normal functions and pathways of NUP160 and SLC43A3 are still unclear, they can be involved in the mitosis, vascular formation, and carcinogenesis, which may contribute to the pathogenesis of angiosarcoma. Our results suggest that full-length NUP160 inhibits cell proliferation, and the truncation of NUP160 stimulates it. Thus, lack of phosphorylation residue due to the NUP160 truncation in the fusion gene may be the cause of sustained cell proliferation of ISO-HAS, at least partly.
Patients with NUP160–SLC43A3 did not have the previously reported mutations in KDR or TP53 (Supplementary Table S5; refs. 1, 3), suggesting the possibility that the fusion may occur independently of KDR/TP53 mutations, that KDR/TP53 is an alternate oncogenic mechanism for NUP160–SLC3A3 fusion, and that these molecules may be associated with each other. With regard to the clinical significance of the fusion gene, the detection of NUP160–SLC43A3 by RT-PCR or variant assays may be useful for the diagnosis of challenging cases or the evaluation of surgical margins, especially in well-differentiated angiosarcoma. Because the patients with tumors expressing the fusion gene visited the hospital earlier than those without, and thus did not have metastasis, immediate and intensive treatments may improve their prognosis. Furthermore, NUP160 siRNA significantly decreased the cell number of ISO-HAS, indicating that full-length NUP160 and/or the NUP160–SLC43A3 fusion may represent potential therapeutic targets. Such specific treatments would be beneficial to avoid side effects in elderly patients.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: M. Jinnin, H. Ihn
Development of methodology: N. Shimozono, M. Jinnin, Z. Wang, A. Hirano, Y. Tomizawa, T. Etoh-Kira, M. Harada
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Shimozono, M. Jinnin, M. Masuzawa, Z. Wang, A. Hirano, Y. Tomizawa, T. Etoh-Kira, I. Kajihara, M. Harada
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Shimozono, M. Jinnin, Z. Wang, I. Kajihara, S. Fukushima, H. Ihn
Writing, review, and/or revision of the manuscript: N. Shimozono, M. Jinnin, M. Masuzawa, M. Masuzawa, H. Ihn
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Masuzawa, Z. Wang, A. Hirano, Y. Tomizawa, T. Etoh-Kira, I. Kajihara, M. Harada
Study supervision: M. Masuzawa, H. Ihn
The authors thank Dr. H. Miyoshi for kindly providing the lentiviral vector and Dr. A. Irie and C. Shiotsu for their valuable technical assistance.