High-risk gastrointestinal stromal tumour (GIST) and synovial sarcoma display similar angiogenic profiles: a nude mice xenograft study

Background Gastrointestinal stromal tumour (GIST) is the most common primary mesenchymal tumour of the gastrointestinal tract. Spindle cell monophasic synovial sarcoma (SS) can be morphologically similar. Angiogenesis is a major factor for tumour growth and metastasis. Our aim was to compare the angiogenic expression profiles of high-risk GIST and spindle cell monophasic SS by histological, immunohistochemical and molecular characterisation of the neovascularisation established between xenotransplanted tumours and the host during the initial phases of growth in nude mice. Methods The angiogenic profile of two xenotransplanted human soft-tissue tumours were evaluated in 15 passages in nude mice using tissue microarrays (TMA). Tumour pieces were also implanted subcutaneously on the backs of 14 athymic Balb-c nude mice. The animals were sacrificed at 24, 48, and 96 h; and 7, 14, 21, and 28 days after implantation to perform histological, immunohistochemical, and molecular studies (neovascularisation experiments). Results Morphological similarities were apparent in the early stages of neoplastic growth of these two soft-tissue tumours throughout the passages in nude mice and in the two neovascularisation experiments. Immunohistochemistry demonstrated overexpression of pro-angiogenic factors between 24 h and 96 h after xenotransplantation in both tumours. Additionally, neoplastic cells coexpressed chemokines (CXCL9, CXCL10, GRO, and CXCL12) and their receptors in both tumours. Molecular studies showed two expression profiles, revealing an early and a late phase in the angiogenic process. Conclusion This model could provide information on the early stages of the angiogenic process in monophasic spindle cell SS and high-risk GIST and offers an excellent way to study possible tumour response to antiangiogenic drugs.


Introduction
High-grade sarcomas can be implanted into immunodeficient mice, where they grow as xenografts supported by the murine stroma blood supply [1]. In general, transplantations of low-grade tumours fail to establish, with successful xenografts deriving mainly from aggressive neoplasms. Although some properties of the original tumours may not be fully represented, xenografts have become very useful models for preclinical experiments in cancer research [2][3][4][5].
In recent years, much research has focused on the role of angiogenesis in tumour development, growth, invasion, and metastasis [6,7]. It has become clear that tumour angiogenesis is the result of an imbalance between proangiogenic and antiangiogenic factors, the threshold of change in favour of proangiogenesis is considered to be the angiogenic switch. Several angiogenic factors and chemokines related with angiogenic mechanisms have been studied in different tumour types [6,8,9]. We recently communicated these findings in a nude mice osteosarcoma model [10] as well as in two high-grade chondrosarcomas (in press) [11]. The angiogenic process presents two different phases of tumour growth. An initial induction phase, in which new unstable vessels are built, followed by a remodelling phase, in which blood vessels are stabilised [12]. At this point, hypoxia occurs and the angiogenic process is activated through the well-known hypoxia-inducible transcription factors (HIF) that induce the expression of several tumour-derived cytokines, such as vascular endothelial growth factors (VEGF) or fibroblast growth factors (FGF) [6] and some chemokines (GRO, CXCL9, and CXCL10) with their respective receptors (CXCR2 and CXCR3) [8,13,14]. More recently, the CXCL12/ CXCR4 axis was reported to be involved in mediating tumour cell invasion and proliferation and to play an important role in tumour angiogenesis, progression, and metastasis [15,16]. Moreover, CXCR4 expression has been associated with poor survival in bone and soft-tissue sarcomas [17,18] and many types of carcinomas [19]. Consequently, considerable interest has been generated in the therapeutic potential of targeting the growth of new vessels (antiangiogenesis) and the capacity to control those that have already been formed (vascular targeting) [13,20].
Soft-tissue sarcomas are an infrequent group of mesenchymal tumours, they may be high grade and display poor survival [21]. Gastrointestinal stromal tumour (GIST) is the most common primary mesenchymal tumour of the gastrointestinal tract and spans a clinical spectrum from benign to malignant; most cases contain KIT-or PDGFRA-activating mutations [22]. Mutations in different genes may also be present, [23] although the main prognostic factors are tumour size, mitotic activity, location and capsular invasion [24,25]. Targeted therapy with imatinib is indicated for high-risk cases and in advanced disease [23].
Synovial sarcoma (SS) is a mesenchymal tumour with a variable degree of epithelial differentiation and a specific chromosomal translocation t(X;18)(p11;q11) that leads to the formation of an SS18-SSX fusion gene. The differential diagnosis with a high-grade GIST may be difficult [26,27]. Both tumour types cause metastasis and display an aggressive behaviour, suggesting that molecular reorganisations such as of SYT-SSX gene translocation and KIT mutations might be similarly essential for the growth of angiogenic factors. A recent publication described an intra-abdominal monophasic spindle cell SS that mimicked the morphology and immunohistochemistry of a high-risk spindle cell GIST [27].
Several animal models have been used in the study of tumour angiogenesis [12,14,28]. Studying angiogenesis through a xenograft model in high-grade sarcomas such as high-risk GIST and synovial sarcomas (SS) may provide a better understanding of this process and increase information regarding potential candidates for effective targeted therapy.
We developed a xenograft nude mice model to clarify the presence of angiogenic factors within the neoformed peritumoral stroma and in the internal tumour blood supply, during the early stages of tumour growth after the transfer into the subcutaneous tissue of the host. To this end, we used two previously established xenotransplanted tumour cell lines of human sarcomas: a high-risk spindle cell GIST and a monophasic spindle cell SS [22,29].
Our aim was to characterise the markers associated with vasculogenesis using histology, immunohistochemistry, and molecular techniques and to search for similarities that may exist between the two tumours.

Samples
Samples were collected from patients treated at the Hospital Clínic Universitari de Valencia. The GIST came from a 63-year-old male with a gastric mass of approximately 26 × 20 × 35 cm diagnosed as a high-risk spindle cell tumour ( Figure 2A). Firstly, the GIST was treated with www.ecancer.org ecancer 2017, 11:726 imatinib (400 mg/day) for six months. The tumour responded partially to targeted therapy and finally resection of the mass was decided upon seven months after diagnosis. No metastasis was seen at the moment of diagnosis, but the patient died of various surgical complications after resection.
The SS came from a 32-year-old male who attended our hospital with a relapse in the right thigh and multiple lung metastases after chemoradiotherapy. The tumour was approximately 10 × 8 × 8 cm and was diagnosed as monophasic spindle-cell SS, the patient died of tumour progression several months after diagnosis.
Molecular biology studies revealed genomic alterations in both tumours. The GIST had the KIT gene mutation and the SS had the typical translocation t(X,18)(SYT-SSX).
The tumours were collected for histopathological, ultrastructural, and genetic characterisation at our Pathology Department. The original tumours were transferred subcutaneously on the backs of nude mice (Nu407 and Nu335) and maintained for several generations (passages). In both tumours, we divided the passages into three time periods, early passages (from 1st to 5th passage), middle passages (from 6th to 10th passage) and late passages (from 11th to 15th passage). We calculated the average speed of tumour growth in both nude mice passages (Nu335 and Nu407) according to the formula (15 mm/days to next passage), 15 mm being the approximate tumour size when mice were sacrificed.
Tumour pieces 3-4 mm in size from the early passages of Nu407 and Nu335 were also xenografted subcutaneously on the backs of two sets of athymic Balb-c nude mice (n = 14 each). The animals were sacrificed at 24, 48, and 96 h; and 7, 14, 21, and 28 days after implantation (neovascularisation experiments). Tissue samples were fixed in 10% formaldehyde, paraffin-embedded, and haematoxylin and eosin (H&E) staining was performed for histological analysis. Moreover non-fixed samples were collected for molecular analysis. Approval for animal experimentation was obtained from the Ethics Committee of the Universitat de València Estudi General (UVEG).

Assembly of TMAs
Tissue microarrays were constructed using a Manual Tissue Arrayer (Beecher Instruments, Sun Praire, WI). Two cores (1 mm thick) of each sample were included, with additional cores in cases with diverse morphologic areas. The TMA contained normal tissue controls, original tumour, and the corresponding xenograft passages. The cores were grouped into early transfers 1-5, middle passages 6-10, and late passages 11-15. After assembly, an initial section from each TMA was stained with haematoxylin-eosin to evaluate the viability of the samples. Several 5-mm sections were also prepared for immunohistochemical (IHC) staining. Table 1 summarises the antibodies used for the IHC.
Immunohistochemistry IHC was carried out by an indirect peroxidase method on paraffin sections following the same methodology, as we discussed in our previous papers [10] and [11].

Molecular biology
RNA was extracted from 50 to 200 mg of tumour samples obtained from the NU335 and Nu407 series.
The whole methodology and studied genes (Table 1S) are also discussed in our previous papers [10] and [11].

TMAs
No morphological changes were observed between passages in GIST; however, a high number of mitoses were clearly observed in the passages in both tumours ( Figures 2B and 2F). The IHC study of GIST showed intense expression of vimentin, CD117, DOG1, desmin, and CD34 ( Figures 2C and 2D) and was negative for PDGFRα and S-100. Ki67 was expressed in 15% of tumour cells in all cores ( Figure 2E).
The IHC study of SS showed positivity for EMA ( Figure 2G), cytokeratin (AE1/AE3) and bcl-2. Intense positivity was also revealed for vimentin and weak expression for TLE1 in all passages. Ki67 increased slightly over the passages, being positive in 20% of tumour cells in the last passages ( Figure 2H), whereas with GIST it was more constant.
Double-immunofluorescence staining demonstrated that chemokine ligand expression in general was slightly higher in the xenograft passages than in the original tumour ( Figure 3). There were very few differences between the two sarcomas with regard to chemokine expression profile. CXCL10 was constantly high in both tumours and GRO was mildly expressed in all passages ( Figures 4A, 4D, and 4E). CXCL9 increased in both tumours over the passages ( Figure 4B). Their receptors CXCR2 and CXCR3 were constantly expressed in all passages, with CXCR2 presenting a higher expression in SS. Finally, the CXCL12/CXCR4 axis was constantly overexpressed in all passages in both tumours ( Figures 4C and 4F).

Neovascularisation experiments
In our neovascularisation experiments, during the first hours after xenografting, peritumoral haemorrhagic areas with inflammatory infiltration compounded by lymphocytes, plasma cells, neutrophils, karyorrhectic, and apoptotic figures were observed in SS and GIST. Small capillaries surrounded the xenograft associated with mesenchymal angioblastic and non-angioblastic cells included in a loose matrix.
Patchy hypoxic necrosis in SS appeared within the first 96 h after implantation, reaching a peak extension in the third week. The SS presented characteristic adipose tissue infiltrate and peritumoral skeletal muscle mouse fibres, but without the pseudocapsule observed in GIST. In GIST, the massive necrosis appeared during the first week, earlier than in the SS. After the fourth week, the histological picture of the GIST was re-established, with features similar to those of the human control, re-establishing also the amount of mitoses and the remission of necrosis, which became patchy and scant. During the third week after xenografting, the inflammatory component decreased. At this time, newly formed capillary vessels were remodelled and penetrated or sprouted into the tumour.
Areas of massive necrosis were associated with a lower proliferative index in both tumours. Ki67 was lower in the early stages after tumour xenografting in both tumours. In the last weeks, the increase in Ki67 expression was also inversely correlated with HIF1α in both neoplasms.
Angiogenic factors represented by the VEGF family and their receptors presented a different expression profile in the two tumours. In SS, maximum VEGF positivity presented 24 h after implantation and was also expressed in the extracellular matrix, while VEGF positivity was lower in GIST and appeared 96 h after xenografting ( Figure 2I). VEGFR2 presented a similar expression profile to its ligand, and VEGFR3 was the most positive receptor in both tumours. HIF1α expression was slightly higher and more constitutively expressed in SS than GIST Double-immunofluorescence staining showed chemokine expression (CXCL9, CXCL10 and GRO) in the tumour cell cytoplasm/ nucleus and deposited in the extracellular matrix. This was also the case for their receptors (CXCR3 and CXCR2) ( Figures 4G and 4H). Chemokine ligand expression was higher during the first 48 h in GIST; however, it appeared later in SS where peak expression occurred during the first week. The CXCL12/CXCR4 axis showed an intense coexpression in both sarcomas at all times throughout the experience. Interestingly, we observed that murine peritumoral stroma expressed CXCR4 but not CXCL12 in the two tumour xenografts ( Figure 4I). It is worth mentioning that the chemokine receptors were expressed more constantly at all times in comparison with their ligands. www.ecancer.org

qRT-PCR low-density arrays of angiogenesis-related genes
Gene expression profiles ( Figure 5) in GIST were similar at 24 h and 7 days but differed from those observed at 48 h, 14 and 21 days. However, SS expression profiles were similar at 24 h and 28 days, differing from those at 48 h and 14 days. In GIST, the early phase appeared 96 h after xenografting and was characterised by the overexpression of genes clearly involved in angiogenesis induction, including VEGF, PDGFA, PDGFB, VEGFC, and their receptors. In contrast, the earlier phase in SS occurred during the first week after xenografting. Finally, in GIST, the late phase of the angiogenic process (remodelling phase) appeared during the first week after xenografting, while in SS, this phase appeared later in the fourth week.
RNA samples corresponding to the third week of SS and to the fourth week of GIST were not viable for analysis. www.ecancer.org

red) and its receptor CXCR3 (green) in GIST tumour cells in early passages (40X). (B) Immunofluorescence staining shows expression of chemokine ligand CXCL9 (red) in GIST tumour cells in later passages (40X). (C) double-immunofluorescence staining shows coexpression of chemokine ligand CXCL12 (green) and its receptor CXCR4 (red) in GIST tumour cells in middle passages (40X). (D) Immunofluorescence staining shows expression of chemokine ligand GRO (red) in SS tumour cells in middle passages (40X). (E) double-immunofluorescence staining shows coexpression of chemokine ligand CXCL10 (red) and its receptor CXCR3 (green) in SS tumour cells in late passages (40X). (F) Doubleimmunofluorescence staining shows high coexpression of chemokine ligand CXCL12 (green) and its receptor CXCR4 (red) in SS tumour cells in late passages (40X). (G) Double-immunofluorescence staining shows coexpression of chemokine ligand GRO (red) and its receptor CXCR2 (green) in SS tumour cells 24 h after xenografting (40X). (H) Double-immunofluorescence staining shows coexpression of chemokine ligand GRO (red) and its receptor CXCR2 (green) in GIST tumour cells in the control tumour (40X). (I) Double-immunofluorescence staining shows coexpression of chemokine ligand CXCL12 (green) and its receptor CXCR4 (red) in GIST tumour cells two weeks after xenotransplantation (40X), observe the different expression between murine stroma (arrow) and tumour cells (asterisk).
A B

Discussion
Angiogenesis is critical for the growth and metastasis of tumours. Early in tumorigenesis, an angiogenic switch is activated by hypoxia, promoting the expression of pro-angiogenic growth factors, such as the VEGF family, its receptors, and HIF1α among others [7]. Recent publications have highlighted the difficulties in differentiating between GIST and monophasic intra-abdominal SS, where molecular mutations are sometimes the only distinguishing feature [27]. It has been suggested that it would be particularly informative to explore possible relationships between the presence of vasculogenic structures and the response to antiangiogenesis therapy [30,31]. Furthermore, it is interesting to speculate that antiangiogenic therapy may result in a selective growth advantage for cells exhibiting vasculogenic mimicry and vascular co-option, promoting drug-induced resistance [30,31].
HIF1α is a principal regulator of cellular and systemic homeostatic response to hypoxia as it activates many genes, including those involved in angiogenesis [6]. In our model, HIF1α was overexpressed in the early stages, indicating that the angiogenic process is constitutively active in the xenotransplanted tumour. HIF1α plays an important role in angiogenic induction and remodelling phases and in an increased VEGF expression [32]. Some recent studies of GIST and chondrosarcoma have shown a correlation between the expression of angiogenic markers, such as VEGF and microvessel density, and a worse prognosis [33,34], suggesting that the development of antiangiogenic chemotherapy might be useful. However, in other tumours, such as osteosarcoma, microvessel density seems to be associated with a longer overall and relapse-free survival [34,35]. Imatinib continues to be used as a first-line medical treatment for advanced GIST, although resistance and non-response sometimes appear. Sunitinib and regorafenib, antiangiogenic drugs, are used as second-and third-line therapies, respectively, and are given in imatinib-resistant GIST cases [23]. Volumetric growth and the development of metastases in cases of GIST appear to be related to the development of a new vascular network [36]. The importance of vascularisation in the context of GIST is the action mechanism of the second-generation drug sunitinib, which is based on the blockade of VEGF activity along with tyrosine kinase receptor blockade that has been used with success in some GIST patients [37,38].
Anthracyclines and ifosfamide, either alone or in combination, are the gold standard treatments for advanced SS [39]. However, after failure of conventional first-line cytotoxic chemotherapy, available treatment options are severely limited because of a high risk-to-benefit ratio in terms of patient tolerability and survival. Recently, it has been demonstrated that pazopanib is a feasible option for patients who have been heavily pre-treated for metastatic SS [40].
Cluster analysis performed on our qRT-PCR expression studies revealed two additional groups of genes clearly separated into two stages, corresponding to early angiogenic induction where VEGF and PDGF family genes among others play an important role, and the later remodelling phases where other angiogenic genes are overexpressed. Apparently the high-risk GIST behaved biologically as high-grade sarcoma in the passages and neovascularisation experiments similar to our previous molecular results [10]. However, induction and remodelling phases of SS appeared later than GIST and other high-grade bone tumours [10]. This difference may be related to a different sensibility and response to antiangiogenic drugs, with GIST being more sensitive. Nevertheless, we cannot be sure that this difference will have any biological translation.
In addition to angiogenic factors, chemokines also play an important role during angiogenic induction. The coexpression of ligands and chemokine receptors in neoplastic cells and extracellular matrix suggests that autocrine and paracrine stimulation by the tumour cells www.ecancer.org ecancer 2017, 11:726 results in production of angiogenic factors in response to hypoxia during the first stages of tumour growth, as reported in other neoplastic and non-neoplastic conditions [8,12,30,41]. Moreover, we found a correlation between high chemokine ligand expression and hypoxic necrosis in both tumours. Few studies of chemokines in SS and GIST have been made [42,43]. CXCR4 expression has been related with poor prognosis in patients with bone and soft-tissue sarcomas in a meta-analysis [17]. High expression of CXCL12/CXCR4 was observed in all passages of both tumours and in the neovascularisation experiment, this could be related with their aggressive clinical behaviour. The CXCL12/CXCR4 axis is related to mediating tumour cell invasion and proliferation and plays an important role in tumour angiogenesis, progression and metastasis [44]. CXCL12/CXCR4 is overexpressed by tumour cells, but not by murine stromal peritumoral cells which only produce CXCR4. Perhaps CXCL12 induces murine stromal cells to generate new vessels in a paracrine effect and may be a good objective for targeted therapy to reduce tumour growth.

Conclusions
This model provides information on the early stages of the angiogenic process in monophasic spindle-cell SS and high-risk GIST. We suggest that different angiogenic molecular profiles could predict different biological and clinical behaviour and determine the response to antiangiogenic treatment. We also demonstrate the importance of chemokine expression as a therapeutic target of tumour growth.
The fact that angiogenesis is a dynamic, changing and multistep process over time should be taken into consideration when developing future therapeutic strategies in soft-tissue tumours. www.ecancer.org ecancer 2017, 11:726 www.ecancer.org ecancer 2017, 11:726