1. Introduction
Soft tissue sarcomas (STS) are a heterogenous group of rare malignant tumours originating from soft tissues that can occur in different parts of the human body. While treatment of localised disease is based on radical resection, evidence from a limited number of trials supports the use of perioperative systemic therapy in high-risk tumours [
1,
2]. Therapeutic options, however, are limited, and in a recent trial, a histology-tailored regimen was inferior to standard chemotherapy with epirubicin and ifosfamide [
2]. Immunotherapy, which has been successfully introduced for several cancers, might be an option to individualise therapy of sarcoma patients [
3].
In normal tissues, expression of cancer-testis antigens (CTAs) is restricted to germ cells and trophoblasts. They are aberrantly expressed in various types of cancer. Following their identification in malignant melanoma, CTAs have been detected in carcinomas of various sites, such as lung, ovary, urinary bladder, liver, and other organs [
4]. Several CTAs, like
MAGE, NY-ESO-1, and PRAME, are expressed in different sarcoma subtypes, such as synovial sarcoma [
5,
6,
7,
8,
9,
10,
11], myxoid/round cell liposarcoma [
6,
11,
12,
13,
14,
15,
16], and other soft tissues sarcomas. CTAs can be highly immunogenic and are considered potential targets for immunotherapy of cancer [
17].
Tumour-infiltrating lymphocytes (TILs) belong to the microenvironment of malignant tumours and constitute an essential part of the human body’s dynamically changing reaction to the tumour. While some TIL populations have been shown to promote tumour progression [
18], others are implicated in killing tumour cells or enhancing the tumour response to certain chemotherapeutics [
19]. Although high TIL counts following neoadjuvant treatment of high-risk soft tissue sarcomas have been associated with a more favourable prognosis [
20], the role of TILs in chemo-naïve STS is less clearly defined [
21].
In the present study, we analyse the expression of CTAs NY-ESO-1, PRAME, and SSX2 in a large and well-characterised cohort of high-risk soft-tissue sarcoma patients with long-term follow-up and correlate our findings with the presence of TILs, clinical tumour characteristics, and survival data.
We show that PRAME, NY-ESO-1, and SSX2 display distinct expression patterns in different STS subtypes. PRAME and NY-ESO-1 expression levels are correlated to patient survival and tumour grade in opposing ways, while both CTAs are correlated with low TIL counts. In multivariate analysis, high PRAME and low SSX2 expression levels as well as metastatic disease, non-radical resections, and not receiving chemotherapy are shown to be independent predictors of shorter overall survival. These results may guide future immunotherapeutic approaches in STS.
3. Discussion
In this study, we analysed the expression of CTAs PRAME, NY-ESO-1, and SSX2 as well as the presence of tumour-infiltrating lymphocytes (TILs) in a well-characterised cohort of high-grade soft tissue sarcoma patients. Although several CTAs have been studied extensively in the past, little is known about PRAME as well as SSX2 on the in-situ protein expression level. Both antigens were chosen because of the lack of data pertaining their presence in sarcoma. NY-ESO-1 was chosen to serve as a reference because it is among the well-studied CT antigens.
We found the expression of PRAME and the presence of a high number of TILs to be associated with shorter survival, while NY-ESO-1 expression was more frequent in patients with a more favourable prognosis. Furthermore, the expression of PRAME was associated with higher grade and a lower number of TILs while NY-ESO-1 expression was correlated with lower grade and low TILs. Finally, high PRAME expression was confirmed as a prognostic factor for overall survival in a Cox proportional hazards model that included clinical parameters like metastatic disease and surgical margins. To our knowledge, the present study of 249 patient cases is the largest study of CTAs in soft tissue sarcoma demonstrating a prognostic significance of these markers.
Significant PRAME overexpression has been described in uterine carcinosarcoma, synovial sarcoma, and multifocal leiomyosarcoma, while other sarcoma subtypes appear to express PRAME less frequently [
22]. In the present study, we found PRAME expression in undifferentiated pleomorphic sarcoma (UPS), malignant peripheral nerve sheath tumour (MPNST), synovial sarcoma, leiomyosarcoma (LMS), angiosarcoma, dedifferentiated liposarcoma (DDLPS), uterine carcinosarcoma, and rhabdomyosarcoma, making PRAME a potential target for immunotherapy in these histological subtypes. Moreover, in our mixed cohort of intermediate and high-grade sarcomas, PRAME expression was prognostic for unfavourable survival, which was not the case in the studies of liposarcoma, leiomyosarcoma, and synovial sarcoma subgroups by others [
9,
22]. Our results, therefore, might help establish PRAME both as a prognostic marker that can be used in nomograms and as a valuable target antigen in soft tissue sarcoma. The fact that we were able to demonstrate PRAME expression in only 3 of 28 synovial sarcomas compared to 100% overexpression in a gene expression study by Roszik et al. [
22] might be due to different methods, detection of PRAME on protein level, and scoring evaluation. The potential of PRAME as a target for immunotherapy may, therefore, be more important than suggested by our results. Future studies should also assess the prognostic value of CTAs in specific histological subtypes.
Currently, PRAME is a target antigen in a clinical phase I trial evaluating multi-tumour-associated antigen-specific cytotoxic T lymphocytes in rhabdomyosarcoma (NCT02239861, TACTASOM). Given that PRAME expression was negatively correlated with lymphocyte infiltration in our study and that PRAME has been described to downregulate antigen-presentation [
22], effective strategies for T-cell recruitment and activation will be needed. These may include a boost of MHC class I expression using, e.g., demethylating drugs and other approaches [
23], combinatorial therapy with NK cells [
24], or checkpoint inhibitors [
25].
Our multivariate regression analysis suggests that the prognostic value of PRAME expression is robust, even in relation to established clinical parameters such as surgical margins or metastatic disease. Nonetheless, while most biopsies were taken before the initiation of neoadjuvant treatment, some patients had received radiotherapy, chemotherapy, and/or hyperthermia before histopathological sampling for this study. Radiotherapy potentially induces gene mutations and alters the expression of CTAs [
26], which may have influenced our results.
NY-ESO-1 has been studied extensively in various cancers. We found it to be expressed in almost half of synovial sarcomas and in some cases of UPS, LMS, DDLPS, and angiosarcoma. This essentially confirms the results of previous studies, although expression in synovial sarcoma has been demonstrated to be as high as 76% to 80% [
10,
27,
28]. In our cohort of soft tissue sarcoma patients, NY-ESO-1 expression was associated with lower grade, and it was prognostic of more favourable survival, as has been shown in a series of high-grade soft tissues sarcoma by Kakimoto et al. [
27]. Interestingly, in non-small-cell lung cancer, high NY-ESO-1 expression was associated with poor prognosis [
29], while in epithelial ovarian cancer [
30] and breast cancer [
31] no relationship with survival has been found. These discrepancies may be caused by the variability of specific interactions between different NY-ESO-1-expressing tumours and their microenvironment, including altered expression of tumour antigens, epitope spreading, and the induction of immunosuppressive cells [
32]. Furthermore, NY-ESO-1 is highly immunogenic, and expression might stimulate a T-cell response in some tumours.
Widespread expression across various cancer types has made NY-ESO-1 an attractive target for different immunotherapeutic strategies such as adoptive cell transfer [
33,
34], cancer vaccines (NCT01883518), and immune checkpoint inhibition [
35]. While its strong immunogenic nature implies that therapies directed against NY-ESO-1 may also boost the natural immune response, its restricted expression in normal tissue presumably limits off-target toxicities [
32].
Interestingly, in our study, NY-ESO-1 and PRAME were expressed more often in tumours with low TILs. The suppression of CD3
+ T-lymphocytes has been described in malignant-melanoma-expressing NY-ESO-1, although its mechanism remains unclear [
36]. In a previous analysis of our cohort, the expression of PD-L1 on STS was associated with an infiltration by PD-1-positive TILs, high tumour grading, and short survival [
37]. These results point to an actual PD-L1 interaction with PD-1-positive TILs in STS and support checkpoint inhibitor therapy in eligible patients. While lymphocyte infiltration is a prerequisite for this approach, it appears that non-T-cell inflamed tumours, which are resistant to PD-1/PD-L1 inhibitors and which were more frequent among NY-ESO-1 and PRAME-positive STS in the present study, can still be treated with adoptive T-cell based immunotherapy [
38]. Targeting CTAs may, therefore, provide a therapeutic option in patients not eligible for checkpoint inhibitor therapy.
Although high SSX2 expression was prognostic of favourable survival in multivariate analysis, the low overall frequency of SSX2 expression in the present study did not allow for a meaningful statistical analysis of this CTA in relation to TILs or grading. The SS18-SSX2 fusion, however, was a frequent finding in synovial sarcomas in the analysis of the Cancer Genome Atlas [
39]. SSX2 may, therefore, be a target in synovial sarcomas, which were SSX2-positive in 25% in our study and which are treated with SSX2-directed cytotoxic T-lymphocytes in the TACTASOM trial (NCT02239861).
Our study has several limitations. First, no conclusions can be drawn about the prognostic value of CTAs in specific histological subtypes due to the limited number of patient cases. Second, in this specific cohort based on the EORTC-STBSG 62961 trial protocol [
40], the majority (78%) of patients received regional hyperthermia, which limits the external validity of our results. Finally, the generalisability of our conclusions is limited by the fact that in 25% of patients, biopsies used in this study had been taken after the initiation of neoadjuvant treatment. Although we have found no indication in our data that systemic therapy influenced CTA expression, results may not be transferable to patients not having undergone neoadjuvant treatment.