Identification of a blood-borne miRNA signature of synovial sarcoma

The screening cohort of patients with active synovial sarcoma (n = 5) as analyzed in a microarray as well as the independent validation cohort of synovial sarcoma patients (n = 3) did not differ significantly from the healthy controls (n = 5), synovial sarcoma in complete remission (n = 10) and active leiomyosarcoma (n = 5), MPNST (n = 3), Ewing sarcoma (n = 3) and liposarcoma (n = 6) group concerning age, BMI, hemoglobin (Hb) level, platelet count and leukocyte count (Additional file 1 and Additional file 2). Two patients with active synovial sarcoma presented the SS18-SSX2 fusion gene phenotype, while five presented the more common SS18-SSX1 phenotype [6, 18]. Tumor tissue of one patient was not available for analysis. Information regarding disease and therapy status of sarcoma patients is depicted in Additional file 3. Whole blood RNA from patients with active synovial sarcoma (n = 5) and healthy controls (n = 5) was analyzed by Affymetrix miRNA 4.0 Arrays. Microarray data are available in the ArrayExpress database (www.​ebi.​ac.​uk/​arrayexpress) under accession number E-MTAB-3273 (http://​www.​ebi.​ac.​uk/​arrayexpress/​experiments/​E-MTAB-3273). Hierarchical clustering of all covered human mature miRNAs and human pre-miRNAs separated sarcoma samples from control samples (Fig. 1a). Following hierarchical clustering, the array results were narrowed down to deregulated miRNAs with a fold change > |3.0|, a p-value <0.005, a q-value (false discovery rate (FDR)-corrected p-value) [19] <0.2 and a similar deregulation of related miRNAs with the same base sequence throughout the array. Of the ten miRNAs, which met these criteria (miR-99a-5p, miR-146b-5p, miR-148b-3p, miR-183-3p, miR-195-5p, miR-223-3p, miR-500b-3p, miR-505-3p, miR-589-5p and miR-1225-5p), nine miRNAs showed an increased expression in tumor patients and one miRNA was decreased (Table 1). Unsupervised hierarchical clustering of this panel could again separate sarcoma patients from healthy controls (Fig. 1b).

Subsequently, seven miRNAs were further validated by qRT-PCR (Fig. 2): miR-99a-5p, miR-146b-5p, miR-148b-3p, miR-195-5p, miR-223-3p, miR-500b-3p and miR-505-3p. All seven miRNAs were significantly upregulated in the screening cohort of patients with synovial sarcoma (n = 5) compared to healthy controls (n = 5). Moreover, five of the analyzed miRNAs were shown to be significantly upregulated in the screening cohort of active synovial sarcoma patients compared to patients with active leiomyosarcoma (n = 5): miR-146b-5p, miR-148b-3p, miR-223-3p, miR-500b-3p and miR-505-3p. MiR-99a-5p and miR-195-5p were not significantly deregulated (Fig. 2). Comparing the screening cohort of active synovial sarcoma patients to patients with active MPNST (n = 3), six of the analyzed miRNAs were shown to be significantly upregulated: miR-99a-5p, miR-146b-5p, miR-148b-3p, miR-223-3p, miR-500b-3p and miR-505-3p. MiR-195-5p was not significantly deregulated (Fig. 2). Analyzing active synovial sarcoma patients and patients with active Ewing sarcoma (n = 3), five of the analyzed miRNAs were shown to be significantly upregulated: miR-146b-5p, miR-148b-3p, miR-223-3p, miR-500b-3p and miR-505-3p. MiR-99a-5p and miR-195-5p were not significantly deregulated. Compared to patients with active liposarcoma (n = 6), all seven analyzed miRNAs were shown to be significantly upregulated (Fig. 2). These results suggest that the detection of these miRNAs in whole blood of synovial sarcoma patients provides a non-invasive way to detect distant metastasis or local recurrence and this “liquid biopsy” could potentially develop into a further keystone of tumor diagnosis and staging. We therefore validated the expression levels of all seven miRNAs by qRT-PCR in an independent cohort of active synovial sarcoma patients (n = 3): Five of the analyzed miRNAs were shown to be significantly upregulated when compared to the group of healthy donors (miR-146b-5p, miR-148b-3p, miR-195-5p, miR-500b-3p and miR-505-3p). MiR-99a-5p and miR-223-3p were also expressed in higher quantities, however differences were not statistically significant (Fig. 3). Expression of this miRNA panel was further analyzed in samples of synovial sarcoma patients in complete remission (n = 10). All seven miRNAs were significantly upregulated in the independent cohort of active synovial sarcoma patients compared to patients with synovial sarcoma in complete remission (Fig. 3)

To exclude a bias of fundamental expression levels of miRNAs among the individual patients with synovial sarcoma and to illustrate the change of miRNA expression at remission and in active disease, we performed a matched-pair analysis in two individual sarcoma patients. Patient 1 initially presented with localized disease of the lower extremity. The miRNA-expression before surgical resection of the tumor as well as two weeks, three months and ten months after complete surgical removal of the tumor was analyzed. The miRNA expression levels decreased rapidly after tumor resection (Fig. 4). Patient 2 was part of the synovial sarcoma in remission cohort at first with unsuspicious miRNA levels. However, she developed pulmonary metastases four months after the initial blood withdrawal. The 2nd blood withdrawal was carried out 5 month after the 1st blood withdrawal. At this time, the miRNA expression levels were significantly elevated when compared to their expression at the time of remission (Fig. 4). Thus, by performing a matched-pair analysis in these two individual sarcoma patients, we were able to illustrate the downregulation of miRNA expression at remission as well as the upregulation of the presented miRNAs in reoccurring active disease.

In summary, the detected miRNAs might provide sensitive biomarkers to distinguish patients with active synovial sarcoma from healthy controls and patients in complete remission by a simple and non-invasive blood test. They might also be useful to distinguish synovial sarcoma from other soft tissue sarcoma subtypes.

Expression profiling of miRNAs in whole blood may be altered by differences in circulating blood cells. In our two cohorts, patients with active synovial sarcoma did not differ significantly from the healthy controls, synovial sarcoma in complete remission and active leiomyosarcoma, MPNST, Ewing Sarcoma and liposarcoma group concerning age, BMI, hemoglobin-level, platelet count and leukocyte count (Additional file 1 and Additional file 2), thus showing that different miRNA expression in the patient groups evaluated was not due to changes in cell populations such as cancer- and chemotherapy-induced anemia. Furthermore, miRNA-expression levels did not differ whether patients received anticoagulation or chemotherapy. The fact that the patient receiving radiotherapy showed the lowest miRNA-expression levels of all active synovial sarcoma patients for most of the detected miRNAs (Additional file 4) was probably due to the fact that this patient was also the only patient presenting with localized disease, thus suggesting that high miRNA expression levels reflect the dissemination or possibly the activity of the disease. When analyzing miRNA expression levels in synovial sarcoma tissue and whole blood of patients with active synovial sarcoma, the majority of the detected miRNAs did not show correlations with SS18-SSX1 and 2 fusion gene phenotypes (Additional file 5). However, the sample number in these subgroups are to low to allow for a valid conclusion. A major advantage of blood sample collection in PAXgene Blood RNA Tubes is the immediate stabilization of intracellular RNA, thus preserving the gene expression profile for reliable downstream gene expression analysis. Without immediate stabilization, degradation of RNA and upregulation or downregulation of transcripts may alter the results of further analysis. Furthermore, blood samples collected in PAXgene Blood RNA Tubes can be safely stored or transported at room temperature for up to 72 h and at −20 °C or −70 °C for at least 50 months. This might reduce pre-analytical errors and facilitate future multi-centre studies, as immediate processing, which is for example needed for the detection of plasma biomarkers, is not necessary, thus avoiding potential changes in miRNA levels by different processing conditions [20].

Keller et al., analyzing 454 whole blood samples collected in PAXgene Blood RNA Tubes from human individuals with different cancers or non-cancer diseases, were able to predict the correct disease in 67.45 % of all individuals by their miRNA expression profiles. Furthermore, they were able to distinguish lung cancer from chronic obstructive pulmonary disease (COPD) with an accuracy of 91.7 %, thus providing support for the feasibility of peripheral whole blood miRNA expression patterns in spite of the fact that changes in cell populations may affect overall miRNA profiles. They also found that only a few of the miRNAs deregulated in blood were also previously reported as deregulated in solid tissues derived from individuals with the same diseases [21], which explains why the miRNAs found to be deregulated in whole blood of synovial sarcoma patients in our study did not correlate with previous findings of deregulated miRNAs such as miR-183, miR-200b∗, miR-375, [16] let-7e, miR-99b, and miR-125-3p, [18] miR-200b, miR-200c and miR-141 [15] in synovial sarcoma tissue. This might be due to the fact that the miRNA expression patterns found in whole blood depict the miRNA expression patterns of peripheral blood mononuclear cells (PBMC) rather than the miRNA expression patterns of circulating tumor cells, as Mookherjee et al. proved in their study [22]. Furthermore, Leidinger et al. showed that especially the miRNA signature of cells of the innate immune system allowed separation between lung cancer patients and healthy controls, however pointing out that cancer-specific miRNA expression of whole blood samples are not determined by a single cell type but influenced by other blood constituents such as platelets or erythrocytes [23]. These studies point out that the miRNA signature of synovial sarcoma detected in our study represents a response of the innate immune system to the tumor rather than the miRNA expression profile of the tumor itself, thus explaining why none of the presented miRNAs match the highly expressed miRNAs in synovial sarcoma tissues as described by Fujiwara et al. [24], Hisaoka et al. [18] and Renner et al. [16]. This is further suggested by the following findings: When comparing the expression of the detected miRNAs in whole blood RNA of patients with active synovial sarcoma with the corresponding tumor tissue (n = 7), we found that most miRNAs were significantly higher expressed in whole blood RNA of patients with active synovial sarcoma (Fig. 5). MiR-99a-5p was the only miRNA being higher expressed in synovial sarcoma tissue than in whole blood. Comparing the miRNA expression levels in synovial sarcoma tissue (n = 7) with healthy skeletal muscle (n = 5), we found no significantly different expression of most of the detected miRNAs (Fig. 5). Only MiR-223-3p was significantly higher expressed in healthy skeletal muscle tissue (Fig. 5e). When comparing the miRNA expression levels of lymphocytes (Jurkat; human T lymphocyte cell line) and monocytes (THP-1; human acute monocytic leukemia cell line) to synovial sarcoma cells (SYO-1 und 1273/99 cell line), we found that most miRNAs were significantly higher expressed in human lymphocytes compared to SYO-1 and 1273/99 synovial sarcoma cells and human monocytes compared to SYO-1 and 1273/99 synovial sarcoma cells (Additional file 6). Interestingly, miR-99a-5p is the only miRNA being higher expressed both in synovial sarcoma tissue compared to whole blood cells (Fig. 5a) and synovial sarcoma cells compared to THP-1 and Jurkat cells (Additional file 6A), thus indicating that the high expression levels of miR-99a-5p in synovial sarcoma patients might not result from a miRNA expression shift of peripheral blood cells but might actually be related to the tumor itself. However, miR-99a-5p shows overlapping expression values in whole blood of patients with active synovial sarcoma compared to the different control groups (Figs. 2 and 3), thus it is not an ideal biomarker for synovial sarcoma. Furthermore, it was shown that human macrophages transfer miR-223 to hepatocarcinoma cells through intercellular contact, leading to decreased expression of stathmin-1 and insulin-like growth factor-1 receptor and decreased proliferation of tumor cells [25], which adds to the point that the detected upregulation of miR-223-3p in whole blood cells might primarily be a defense of circulating mononuclear cells against the proliferation of synovial sarcoma.

Interestingly, we found that four of the detected seven miRNAs seem to interact with the nuclear factor kappa B pathway (NF- κB), thus possibly playing an important role in toll like and cytokine receptor signaling. MiRNAs are well described regulators of the immune response [26]. The expression of miR-146b has been both shown to be upregulated by inflammation-related induction of NF-κB through a MyD88-dependent pathway, and also downregulating TNF receptor-associated factor 6 (TRAF6) and IL-1 receptor-associated kinase 1 (IRAK1), thus reducing the activity of the pathway by a negative feedback regulation loop and maintaining the balance of toll like and cytokine receptor signaling [27]. In consistency with this study, Hulsmans et al. showed that silencing of miR-146b-5p resulted in more NFκB p65 DNA binding activity and TNFα. Moreover, low levels of miR-146b-5p in monocytes of obese persons have been shown to be associated with increased mitochondrial reactive oxygen species (ROS) [28], thus showing that miR-146b inhibits NF-κB – mediated inflammation. An upregulation of miR-146b-5p might thus reflect an immune evasion of the tumor, leading to increased tumor proliferation and progression. MiR-195 also seems to interact with NF-κB, as it was shown to decrease the expression of multiple NF-κB downstream effectors [29]. Another of the NFκ-B related miRNAs is miR-223; as the transcriptional activity of miR-223 promoter was shown to be activated by Notch and NF-κB [30]. Finally, miR-500 was shown to induce gastric cancer cell proliferation and tumorigenicity by activating NF-κB. [31]. Thus, although finding the exact biological significance of the miRNAs in tumorigenesis of synovial sarcoma is a complex matter, it seems that the miRNAs detected in our study control various activities of the immune system by interacting with different pathways such as the NFκ-B signaling pathway. This further underlines our hypothesis that the detected miRNA expression levels reflect changes of the immune system as a defense to tumor progression as well as immune evasion mechanisms of the tumor.

Biomarkers for synovial sarcoma are urgently needed to identify patients with local recurrence and metastatic disease, as imaging studies are often inconclusive. Although validated miRNAs were not always significantly different in patients with active disease compared to the different control groups, miR-146b-5p, miR-500b-3p and miR-505-3p allow discrimination of patients of active synovial sarcoma patients from the different control groups, showing clear cut-off-values at approximately 18.5 (miR-146b-5p), 17.5 (miR-500b-3p) and 11.5 (miR-505-3p) 2^-ΔΔCt ; while miR-99-5p, miR-148b-3p, miR-195-5p and miR-223-3p show overlapping expression values. These miRNAs are promising candidates as diagnostic biomarkers for synovial sarcoma, potentially allowing the detection of local recurrence or distant metastasis by a simple blood test: A rise of these miRNAs beyond the cut-off value - while initially lying underneath the cut-off value following complete removal of the tumor - might indicate local recurrence or distant metastasis, as shown in Fig. 4b (Patient 2). Moreover, these miRNAs could be applied in order to distinguish between synovial sarcoma and other types of sarcoma, which may be of therapeutic importance if chemosensitive sarcoma such as the rare extraosseous Ewing’s sarcoma with spindle cell sarcoma-like pattern, often showing histological overlap with synovial sarcoma [32], are considered in differential diagnosis. Although our results proved to be highly significant, the main limitation of this study consists in the low number of patients within the different study groups. Nonetheless, one has to take into account that soft tissue sarcoma is a rare malignancy which is furthermore divided into a variety of different subtypes, thus making the collection of a large number of patients within one study group difficult. We preferred to identify a miRNA panel in a small group of patients with synovial sarcoma as an initial “proof of concept” rather than including a large number of patients with different sarcoma subtypes. As synovial sarcoma is characterized by the presence of the SS18-SSX fusion gene, which can be detected by fluorescence in situ hybridization (FISH) of tumor tissue, this sarcoma subtype can be specifically diagnosed without the risk of being confounded with other sarcoma subtypes. Through this approach, we achieved highly significant miRNA array and qRT-PCR results and a clear separation of sarcoma samples from control samples by hierarchical clustering. The fact that the miRNA signature established in this pilot study even distinguishes synovial sarcoma from leiomyosarcoma, MPNST, Ewing sarcoma and liposarcoma samples further elucidates the specificity of our results.

Further studies are warranted to validate the panel of deregulated miRNAs in a larger cohort of independent synovial sarcoma patients and to identify miRNA signatures of other sarcoma subtypes. Moreover, the existing group of patients with synovial sarcoma in complete remission needs to be closely monitored in order to evaluate a possible rise of the established miRNAs with local recurrence or distant metastasis.