Introduction
Osteosarcoma is the most common primary high-grade bone malignancy, primarily affecting children and adolescents [
1]. It commonly arises in the metaphyseal plates of the long bones of the extremities (i.e. distal femur, proximal tibia), while tumours developing in the axial skeleton, pelvis or craniofacial bones tend to occur in older individuals. Osteosarcoma patients are treated with curative intent, consisting of surgery of the primary tumour and any resectable metastatic lesions, in addition to pre- and post-operative chemotherapy [
2]. This multimodal therapeutic approach greatly improves the disease-free survival probability, from 10 to 20% with the surgery alone, to more than 60%. However, survival for patients with relapsed or metastatic disease remains dismal and unchanged over the last three decades, as efforts in developing novel active agents have been generally disappointing [
3,
4]. Osteosarcoma’s genomic complexity is one of the major explanations for the lack of specific targetable mutations or molecular pathways. On the other hand, this genomic characteristic and the associated high mutational burden may generate specific tumour neoantigens, therefore providing potential targets for T-cell-based immunotherapies [
5‐
7].
Historically, the first successful example of immunotherapy was in 1891 when William B. Coley injected a mixture of streptococcal bacteria into unresectable bone sarcomas, resulting in an immunological reaction and tumour regression [
8,
9]. Recent advances in cancer immunology have now revealed the importance of a spontaneous antitumour immune reaction to predict response to immunotherapy, mostly carried out by cytotoxic CD8-positive T cells [
10]. Antigen presentation by surface HLA expression on tumour cells is an important prerequisite for antitumour immunity since loss or down-regulation of HLA molecules is a common mechanism deployed by tumour cells to escape immune surveillance [
11]. Another immune escape mechanism is the dysregulation of immune checkpoint pathways such as PD-1/PD-L1 axis, which has been actively studied in epithelial malignancies [
12]. So far, data regarding the immune microenvironment in osteosarcoma are rather limited. In order to evaluate the feasibility of T-cell-mediated immunotherapies, we assessed HLA class I expression, PD-L1 and T-cell infiltration, as well as their changes during osteosarcoma progression using immunohistochemistry.
Discussion
In this study, we investigated HLA class I and PD-L1 expression, as well as T-cell infiltration in osteosarcoma, to assess the potential application of T-cell-based immunotherapies. To characterise the HLA class I expression, we used the monoclonal antibodies HCA2 and HC10, which recognise HLA-A and HLA-B/C heavy chains, respectively, and an antibody specific for the light chain β2-microglobulin. Overall, we observed defects in HLA class I expression in 44% of the osteosarcomas, mainly with a heterogeneous pattern instead of a complete lack of HLA class I. Previous studies have reported a loss or a down-regulation of HLA class I expression in approximately 50–62% of osteosarcomas, using a pan-HLA class I antibody (EMR8-5) and a cut-off of 50% positive tumour cells [
16,
17]. In our series, negative or heterogeneous expression of HLA-A molecules was found more frequently compared to the other heavy chains HLA-B/C and β2-microglobulin. The clinical and therapeutic relevance of this finding, if any, has to be further investigated, since HLA-A and HLA-B both are antigen presenting molecules, and required for T-cell-based immunotherapies.
Another major clinical finding in our series is the high density of tumour-infiltrating T cells in metastatic osteosarcoma lesions compared to primary tumours and local relapses. CD3
+CD8
− and CD3
+CD8
+ T cells were found in the same proportion within the tumour microenvironment while the density of CD3
+CD8
−FOXP3
+ T cells was higher in local relapses, probably reflecting local immune escape mechanisms. Expression of the light chain β2-microglobulin together with the HLA-B/C was an important predictor of T-cell infiltration. In other cancers, particularly with high mutational burden such as melanoma, lung and colorectal cancers, a high T-cell infiltrate commonly associates with better clinical outcome and may predict response to immunotherapy [
10,
18,
19]. Although not statistically significant, high T-cell infiltrate in primary osteosarcomas also tended to have survival benefit for the patients included in our study. It has been demonstrated that tumour-infiltrating lymphocytes could be easily isolated from adult osteosarcomas and exhibited in vitro a high cytotoxic activity, suggesting that therapies based on these tumour-infiltrating T cells could be an efficient strategy in osteosarcoma [
20]. However, some tumours are poorly infiltrated by T cells and the mechanisms associated with reduced T-cell trafficking and infiltration are poorly understood. A potential mechanism might be the loss of the tumour suppressor gene
PTEN, frequently reported in osteosarcomas [
21,
22].
PD-L1 positivity in metastases, while mainly negative in the associated primary tumours, emphasises the dynamics of an adaptive mechanism of immune escape. Overall, PD-L1 expression was found in almost half of the metastatic osteosarcoma lesions (48%), on both osteosarcoma cells and immune cells, mainly macrophages. Our results are in accordance with other groups who reported PD-L1 positivity using another monoclonal antibody only in metastatic osteosarcomas, as well as a higher PD-L1 mRNA expression in metastases, which correlated with the T-cell infiltrate [
23,
24]. An increased PD-1 expression on peripheral CD4
+ T cells was also observed in patients with metastases, strengthening the idea that PD-L1/PD-1 axis may play a role during osteosarcoma progression [
25]. Recently, Koirala et al. reported the discrepancy between whole tumour sections and tissue micro-arrays as a result of the heterogeneity of PD-L1 expression in osteosarcoma, confirming our idea that this immune marker should be evaluated on whole sections [
26]. Moreover, they identified PD-L1 positivity as a potential prognostic marker for poorer survival, which was not the case in our cohort. However, it should be noted that we mostly included patients with poor outcome, in order to assess the immune changes during disease progression. Therefore, our conclusions regarding the prognostic value of these different immune markers in primary tumours cannot be generalised in all osteosarcoma patients.
What is the explanation for the observed low and/or heterogeneous expression of HLA class I? In other cancers, defects in the antigen presentation pathway were reported at different levels (mutations in
HLA-
A,
HLA-
B and
HLA-
C genes, β2-microglobulin or defects in components of antigen-processing machinery) and have been regarded as mechanisms to escape from T-cell immune recognition [
27‐
29]. Tumour cells that have a loss/down-regulation of HLA class I may gain a selective clonal advantage in a process called immunoediting, enabling them to escape from the CD8+ T-cell-mediated destruction [
30].
However, in osteosarcoma, the question can be raised regarding the mechanisms involved in the frequent heterogeneous expression of HLA class I. First of all, there is scarce data on baseline HLA class I expression in normal bone cells. Therefore, it is not clear whether low HLA class I expression reflects the normal situation or, in contrast, ‘down regulation’ as a consequence of escape from T-cell immune recognition. An alternative explanation for heterogeneity of HLA class I expression could be the consequence of the extreme genomic instability of osteosarcoma. But another interesting interpretation would be the induction of HLA class I expression by interferon-γ secreted by T cells. In this study, we observed that in tumours with HLA class I negative and positive regions, the T cells spatially colocalised with the HLA class I positive tumour areas. Additionally, we observed a strong correlation between high numbers of tumour-infiltrating T cells and PD-L1 expression. Together, this could imply that, in areas of T-cell infiltration, immune activation leads to induction of HLA class I and PD-L1 expression. Although it would have been interesting to look at the difference in immune profile of primary tumour between patients with and without metastasis at diagnosis, we could not answer the question due to the low number of samples.
Importantly, and despite the limited number of patients, HLA class I expression in primary osteosarcomas correlated with a better disease-free survival, which is consistent with previous results [
16,
17]. In addition, we observed that HLA class I expression in metastases was frequently positive, and this was in contrast to Tsukahara et al. [
17] who reported that HLA class I loss or down-regulation occurred more commonly in metastatic osteosarcoma lesions (7 out of 8 cases; 88%). Consequently, the HLA data from our series can have two therapeutic implications. First, the frequent HLA class I expression in both primary tumours and metastases highlights the potential of (neo)antigen presentation by osteosarcoma cells, which can be exploitable in developing personalised immunotherapies. Second, identifying the rare patients with a negative or weak expression of HLA class I molecules on primary tumour is of clinical interest, to consider a different therapeutic approach (i.e. strategy based on NK cells) [
31].
Altogether, the increased number of tumour-infiltrating T cells and PD-L1 expression during disease progression, associated with a frequent classical HLA class I expression, suggest that T-cell-based immunotherapy with adoptive cell transfer, peptide vaccines or immune checkpoint blockade could be a suitable treatment for metastatic osteosarcoma patients. Preclinical data demonstrated the benefit of PD-L1/PD-1 blockade antibodies, alone or in combination with anti-CTLA-4 in a mouse model of metastatic osteosarcoma and the efficacy of pembrolizumab, a monoclonal anti-PD-1 antibody, is currently being investigated for bone sarcomas in the phase II SARC028 study (NCT02301039) [
23,
32]. Considering the limited therapeutic options currently available in advanced diseases, enhancing this pre-existing antitumour immune response in metastastic lesions may offer clinical benefit in osteosarcoma patients.