Tumoral immune-infiltrate (IF), PD-L1 expression and role of CD8/TIA-1 lymphocytes in localized osteosarcoma patients treated within protocol ISG-OS1

Background We hypothesized that immune-infiltrates were associated with superior survival, and examined a primary osteosarcoma tissue microarrays (TMAs) to test this hypothesis. Methods 129 patients (pts) with localized osteosarcoma treated within protocol ISG-OS1 were included in the study. Clinical characteristics, expression of CD8, CD3, FOXP3, CD20, CD68/CD163 (tumor associated macrophage, TAM), Tia-1 (cytotoxic T cell), CD303 (plasmacytoid dendritic cells: pDC), Arginase-1 (myeloid derived suppressor cells: MDSC), PD-1 on immune-cells (IC), and PD-L1 on tumoral cells (TC) and IC were analysed and correlated with outcome. Results Most of the cases presented tumor infiltrating lymphocytes (TILs) (CD3+ 90%; CD8+ 86%). Tia-1 was detected in 73% of the samples. PD-L1 expression was found in 14% patients in IC and 0% in TC; 22% showed PD-1 expression in IC. With a median follow-up of 8 years (range 1-13), the 5-year overall survival (5-year OS) was 74% (95% CI 64-85). Univariate analysis showed better 5-year OS for: a) pts with a good histologic response to neoadjuvant chemotherapy (p = 0.0001); b) pts with CD8/Tia1 tumoral infiltrates (p = 0.002); c) pts with normal alkaline phosphatas (sALP) (p = 0.04). After multivariate analysis, histologic response (p = 0.007) and CD8/Tia1 infiltration (p = 0.01) were independently correlated with survival. In the subset of pts with CD8+ infiltrate, worse (p 0.02) OS was observed for PD-L1(IC)+ cases. Conclusions Our findings support the hypothesis that CD8/Tia1 infiltrate in tumor microenvironment at diagnosis confers superior survival for pts with localized osteosarcoma, while PD-L1 expression is associated with worse survival.


INTRODUCTION
Ostesarcoma is a rare, aggressive sarcoma. Whereas there is an agreement that surgery, with adjuvant chemotherapy, is paramount for the primary therapy of the cancer, about 30% of patients without evident metastases at presentation will die of disease [1]. The prognosis is even poorer (10% survival at 5-years) in patients with synchronous metastases [1]. Therefore there is an urgent need to identify new targets, different risk groups and predictive factors for tailored treatment for each individual.
Various mechanisms have been proposed for the resistance of human solid tumors to immune recognition and obliteration, including the recruitment of regulatory T cells (T-reg), myeloid derived suppressor cells (MDSC) [2] and up-regulation of immune inhibitory ligands, such as the ligand of programmed cell death protein 1 (PD-L1) [3].
Few data on osteosarcoma microenvironment composition and the PD-1/PD-L1 expression are available, and most of the studies refer to small series with incomplete clinical information, inhomogeneous for stage (metastatic and localized patients), treatments and timing of biopsies [10]. of PD-L1 by macrophages (CD68 + /CD163 + ) occurs at the interface of tumor cell nests with immune infiltrates secreting pro-inflammatory factors such as interferon-γ. The ligation of PD-L1 on macrophage and, in some histotypes, on tumor cells, with programmed cell death protein 1 (PD-1) molecules will down-modulate T cell function, essentially creating a negative feedback loop that reduces antitumor immunity (the so called 'tumor shield' effect), eventually reducing CD8 tumoricidal function. www.impactjournals.com/oncotarget The present study aim is to characterize the immune-infiltrates and PD-L1/PD-1 pathway in sample of chemo-naïve patients with localized osteosarcoma, treated according to the same protocol, in order to assess its prognostic implications and its potential role in cancer immune-evasion.

Tumoral microenvironment components prior chemotherapy (bioptic samples)
Eighty-six out of 129 cases analysed were evaluable for at least 7 markers and were included in the present study.

Tumoral microenvironment components postinduction chemotherapy (surgical samples)
Based on multivariate analysis results on pretreatment samples, a post-hoc analysis on CD8 and Tia1 was performed.
Due to post-treatment changes 33/86 patients were assessable after treatment (excluding 53 patients with massive necrosis and no tumor).
The first study demonstrating the association between CD8 infiltrate and cancer specific survival was published in 2001 [18], and similar findings were subsequently confirmed among many histotypes [19,20].
In the sarcoma field both a series of 33 ostesarcoma, including metastastic and localised patients [11], and a large soft tissue sarcoma French study, including several histotypes and also low grades lesions, failed to demonstrate a prognostic role for CD8+ lymphocytes [21].
The results of our study are in contrast with those data [11,21], in fact a strong advantage in terms of survival has been observed in patients with CD8 + infiltrate.
To investigate the differentiation state of tumorinfiltrating T cells, tissues were analyzed for Tia1 expression, which is a marker of cytotoxic function: the prognostic significance of CD8 + cells was even more relevant when the tumor infiltrate was characterized by the concomitant presence of Tia1 + lymphocytes. Tia1 + lymphocytes might represent a more efficient subset of CD8 + effector cells, playing an important role in immunesurveillance of osteosarcoma [22].
The prognostic role of cytotoxic TILs CD8/ Tia1 was also confirmed after induction chemotherapy, while chemotherapy does not seem to induce significant changes in CD8 + TILs: all cases with no CD8 + TILs prior chemotherapy, were confirmed negative after chemotherapy, while a slight increase on score severity (1 to 2/3) was observed.
In our study, the rate of PD-L1 expression in IC was 14%. Previous studies on osteosarcoma samples reported a higher rate of positive expression of PD-L1(IC) ranging from 25% [11] to 74% [10]. This difference might be related to the heterogeneity of the cases examined, being many of them metastatic, while our patients all had localized disease. In fact, it is well known that PD-L1 expression increases in advanced stages of the diseases as reported by Sundara Y et al: 13% in primary tumours, 25% in local relapses tissue and 48% in metastatic (p = 0.002) [12].  In our study a trend towards an inferior survival for positive PD-L1(IC) patients was observed. Other reports were able to demonstrate a significantly inferior event-free-survival (EFS) for osteosarcoma patients with positive PD-L1(IC) [11]. A prognostic role for PD-L1 was also confirmed by Kim et al, at RNA level [10]. Interestingly, in our series, the PD-L1(IC) expression has a prognostic significance at multivariate analysis in the subgroup of patients with CD8 + immune-infiltrate. It might be hypothesised that "PD-L1"-mediated immunesuppression negatively influences CD8 + lymphocytes function (Figure 1). In addition, it was shown that tumor response to PD-L1 or PD-1 inhibition is directly related to both the level of PD-L1 expression and lymphocytic infiltration of the tumor [23][24][25].
None of the patients had PD-L1 on tumor cells, as shown for other histotypes, such as colon rectal and gastric carcinoma [26,27]. About 7% of osteosarcoma presented  PD-L1 in the neoplastic clone in the study by Koirala et al [11], while head and neck squamous cell carcinoma, melanoma, breast and kidney cancer frequently express PD-L1 on the surface of tumour [4,28]. Such variable expression among different studies may reflect the variable susceptibility of tumour cells and infiltrating immune cells to cytokines and other stromal factors in the tumour milieu [5]. In fact there are two distinct type of PD-L1 expression: the first is a constitutive (innate) expression on tumoral cells membranes, with an homogenous patter; the second is adaptative, and can be found both on tumoral cells or in macrophage [5].
In 67% of the cases we found CD163 + macrophages in microenvironment. CD163 was shown to be a useful marker for M2-like macrophages, which have a "protumoral" activity, in contrast with M1-like macrophages characterized by a "tumoricidal" activity [29].
In our series no difference in survival according with presence of TAMs, characterized by both CD163 and CD68 expression, was observed (CD163 + 81% vs   [13,14]. This might be explained by different statistical analysis design and different stages of patients included [13,14]. Our data suggest a role of immune-infiltrate in progression of localized osteosarcoma, and might support the use of immune-modulating agents in the treatment of this tumor. Of interest, mifamurtide, a modulator of innate immunity, which increases a wide variety of immunomodulatory molecules [15] and favours CD8 and NK cell activation [30], has been approved by EMA for the treatment of patients with localized high-grade osteosarcoma based on the results of a randomized trial [31]. In conclusion, CD8/Tia1 citotoxic T-lymphocytes emerge as an important player in anti-tumor immune response. Also, this study highlights the prognostic role of tumor microenvironment in the setting of localized osteosarcoma. The data are interesting and intriguing, but a clinical application requires confirmations in other series ( Figure 4).

Pre-treatment (bioptic samples)
After EC approval, tissue samples obtained from biopsies performed prior chemotherapy in 129 patients were collected. Patients were prospectively treated at Istituto Ortopedico Rizzoli from 04/2001 to 11/2006 within protocol ISG-OS1, with surgery and a chemotherapy based on methotrexate, cispaltin, adriamicyn and ifosfamide as described [16].
For tissue microarray (TMA) construction, a slide stained with hematoxylin and eosin was prepared from each formalin-fixed, paraffin-embedded (FFPE), and representative tumor regions were morphologically identified and marked on each slide. Tissue cylinders with a diameter of 1.0 mm were punched from the marked areas of each block and brought into a recipient paraffin block. Five TMAs were constructed. Each tumor sample was represented by a minimum of 1 core to a maximum of 5 cores.
Only cores with tumoral component were included in the analysis. The immunehistochemical scores were generally concordant among cores of the same patient, in case of heterogeneity the highest score was considered for the analysis. For survival analysis samples were classified as negative (immunostaining = 0) or positive (immunostaining = 1 to 3) for all markers except: CD68 + cases that were classified in high expressing (severe and moderate expression) and low expressing (focal expression). Score for PD-L1 expression in TC: specimens with >5% membranous expression were considered positive.
OS was estimated according to the Kaplan and Meier method with their respective 95% confidence intervals (CI) and calculated from the first day of chemotherapy administration to death or last follow-up visit.

Post-induction chemotherapy (surgical samples)
In 86/129 patients, FFPE from tumoral masses surgically resected, after neoadjuvant chemotherapy, were collected. The full slides sections from FFPA were investigate by immunohistochemistry for CD8 + and Tia1 expression.
The following factors were correlated with overall survival: tumoral microenvironment components (CD8 and Tia1).

Author contributions
EP, CA substantive intellectual contribution to conception and design of study, drafting and revision of manuscript, statistical analyses and interpretation of data SF: patients selection / inclusion; statistical analyses and interpretation of data SP, PL, PP: substantive intellectual contribution to conception and design of study, final approval of the version to be published MT, KS, ALMSB: drafting and revision of manuscript.