Background
Sarcomas, characterized by high heterogeneity, are the main types of malignant bone and soft-tissue tumors [
1], and neoadjuvant chemotherapy combined with surgery is the mainstream treatment strategy for most sarcomas. However, chemotherapy for sarcomas has entered the platform stage, and more than 40% of sarcoma patients ultimately experience tumor metastasis or recurrence with a poor prognosis [
2]. Moreover, when traditional treatment fails for aggressive disease, few treatment choices are left. However, some sarcomas, such as chondrosarcoma, are not sensitive to chemotherapy or radiotherapy, and surgery is the only effective treatment. Therefore, when the tumor metastasizes or cannot be resected, both the patients and doctors are faced with a devastating dilemma. Therefore, novel and effective therapies for sarcomas are urgently needed to mitigate these desperate situations.
PD-L1 and PD-L2 are both ligands of PD-1, and these interactions transduce co-inhibitory signals for T cell activation, suppress T cell function, which is called T cell exhaustion, and ultimately promote tumor evasion of the immune system [
3,
4]. For the past few years, immune checkpoint inhibitors (ICIs) have led to a breakthrough in solid tumor immunotherapy by relieving the immunosuppressive state of the tumor microenvironment and restoring the immune function of T cells to kill tumor cells [
5‐
7]. Previous studies have shown that PD-1, PD-L1, and PD-L2 have different prognostic roles in various tumors [
8‐
19]. The combination between PD-L2 and PD-1 also has a negative effect on T cell activation, which may be an important reason why some tumors express little to no PD-L1 yet still respond to PD-1 immunotherapy. This observation has aroused the interest of numerous scholars. However, current studies on sarcoma immunotherapy are limited, and the existing studies mainly focus on PD-L1 or PD-1 expression and their clinical implications in sarcomas [
20‐
27]. To our knowledge, there have been no studies of PD-L2 expression in sarcomas.
In our study, we systematically investigated the expression patterns of PD-L1, PD-L2, and PD-1 in sarcomas including osteosarcoma, chondrosarcoma, synovial sarcoma, and giant cell tumors (GCTs) and further evaluated the association between PD-L1, PD-L2, and PD-1 expression and clinical prognosis of osteosarcoma to provide a therapeutic strategy guide. Then, we investigated the therapeutic effect of nivolumab on osteosarcoma and its underlying mechanism.
Methods
Tissue microarray construction
Three tissue microarray (TMA) slides were used to evaluate the expression patterns of the PD-1 axis. One was constructed using samples acquired from the Musculoskeletal Tumor Center, Peking University People’s Hospital (Beijing, China), and the relevant tumor tissues, including osteosarcoma (62 cases) and dedifferentiated chondrosarcoma (4 cases), were acquired at the time of definitive surgery and disease recurrence (either local or metastatic) with several paired samples included on the array. Core tissue (3 mm in diameter) was obtained from each donor block and placed in the recipient tissue array block. TMA sections (5-μm thickness) were sliced and preserved properly at room temperature for subsequent experiments. Informed consent was obtained from each patient, and the study was approved by the ethics committee of Peking University People’s Hospital. Clinical and histopathologic data were collated through a retrospective review of patient records. The other two TMAs (OS803 and SS1501) were purchased from US Biomax, Inc. Among them, the SS1501 TMA included chronic synovitis (9 cases), giant cell tumor (14 cases), and synovial sarcoma (127 cases); the OS803 TMA included 27 cases of chondrosarcoma. Some core tissues were removed from the slide during staining for immunohistochemistry (IHC); thus, the presented results only included the samples that remained on the slide and could be graded.
Cell culture and reagents
HOS, KHOS, 143B, MNNG, U2OS, SAOS-2, MG63, and NIH3T3 cells were obtained from American Type Culture Collection (ATCC). The KHOS cell line used for in vivo experiments was recently authenticated in Beijing Microread Genetics Co., Ltd. by STR analysis and was passaged for less than 3 months after resuscitation. HOS, KHOS, and U2OS cells were cultured in RPMI 1640 medium (HyClone). 143B, MNNG, SAOS-2, MG63, and NIH3T3 cells were maintained in DMEM (HyClone). Cell culture media were supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Invitrogen). All cell lines were cultured at 37 °C with 5% CO2.
Western blot
Western blotting was performed as previously described [
28]. Briefly, equal amounts of protein were collected from various cell lysates, loaded onto 15% SDS-PAGE gels, resolved using a NuPAGE system (Invitrogen), and transferred onto PVDF membranes. After blocking in non-fat milk for 1 h, the membranes were incubated with corresponding primary antibodies overnight at 4 °C. The bands were probed with the western blot detection system (Bio-Rad, Hercules, CA, USA). Antibodies against PD-L1 (sc-50298) and GAPDH (sc-25778) were purchased from Santa Cruz Biotechnology. Anti-PDL2 (ab187662) was purchased from Abcam.
Flow cytometry
All osteosarcoma cell lines were analyzed for PD-L1 and PD-L2 expression by flow cytometry. The cells were prepared and incubated with the primary antibody for 30 min at 4 °C and then washed with phosphate-buffered saline (PBS) according to the manufacturer’s instructions. After washing, cells were assayed using an Accuri C6 flow cytometer (BD Biosciences, San Diego, CA, USA). Fluorescent antibodies, including PE-PDL1 (12-5983), APC-PDL2 (17-5888) and the corresponding isotype controls (17-4714 and 12-4714), were purchased from eBioscience. The single-cell suspensions isolated from the mouse tumors were similarly examined for human lymphocyte infiltration by flow cytometry, and the fluorescent antibodies included APC-mouse CD45 (BioLegend, cat#103112), PE-human CD45, PerCP-human CD3, FITC-human CD4, and PE-human CD8a (cat#555483, 347344, 561005, and 340046; BD Pharmingen).
Immunohistochemistry and immunofluorescence assay
Paraffin sections were incubated with the corresponding antibodies and stained with nonimmune serum in PBS instead of the primary antibody as the negative control. Based on the average percentage of positive cells calculated from at least 10 representative fields (× 400 magnification), positive staining was defined as a positive cell percentage ≥ 10%. Staining intensity was classified as follows: 0, no staining or staining in < 10% of tumor cells; 1+, weak to moderate staining in 10 to 20% of tumor cells; 2+, strong staining in 10 to 20% of tumor cells or weak staining in 20 to 50% of tumor cells; 3+, moderate to strong staining in 20 to 50% of tumor cells or staining in 50% of tumor cells. More than 10 representative areas (× 400 magnification) were calculated for the tumor-infiltrating lymphocyte analysis. The immunostaining assessment was conducted by two independent pathologists without any previous knowledge of the clinical characteristics and outcomes. Antibodies for IHC against PD-L1 (M442) and PD-1 (M569) were purchased from Spring Bioscience. Anti-PD-L2 (82723) was purchased from Cell Signaling Technology, and anti-CD4 (19068-1-AP), anti-CD8a (17335-1-AP), anti-granzyme B (13588-1-AP), and anti-interferon gamma (15365-1-AP) were purchased from Proteintech Group Inc.
For immunofluorescence assay of colocalization of PD-L1 and PD-1 or PD-L2 and PD-1, paraffin sections were incubated with anti-PD-L1 and anti-PD-1 or anti-PD-L2 and anti-PD-1 antibody overnight at 4 °C, then washed three times with PBS and incubated with Alexa Flour 594-conjugated goat anti-mouse IgG and Alexa Flour 488-conjugated goat anti-rabbit IgG for 1 h at room temperature. The sections were viewed using confocal microscopy (FV10i, Olympus, Tokyo, Japan).
Quantitative RT-PCR
Total RNA was isolated using Trizol (Invitrogen), and cDNAs were synthesized with purified RNA and OligdT primers using SuperScript III First-Strand Synthesis SuperMix (Invitrogen). Real-time quantitative PCR was performed using the SYBR-Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) on Bio-Rad CFX96 (Applied Biosystems, CA, USA). Relative transcript expression was normalized to GAPDH. All protocols were conducted as per the manufacturer’s instructions.
The primer sequences were as follows: PD-1 forward 5′-AAGCTTATGTGGGTCCGGC-3 and PD-1 reverse 5′-GGATCCTCAAAGAGGCC-3′; PD-L1 forward 5′-ACGCATTTACTGTCACGGTTCC-3′ and PD-L1 reverse 5′-CGATGGGGTTCCGGCTTCAG-3′; PD-L2 forward 5′-AAAGAGCCACTTTGCTGGAG-3′ and PD-L2 reverse 5′-GAGGACGTAGTAACGAAAGT-3′; GAPDH forward 5′-GCACCGTCAAGGCTGAGAAC-3′ and GAPDH reverse 5′-ATGGTGGTGAAGACGCCAGT-3′.
The osteosarcoma dataset from the Gene Expression Omnibus (GEO) [
29] (accession no. GSE21257 [
30]) was used for the datamining and bioinformatic analyses in this study. Clustering and heat map visualization was performed using the MeV software, version 4.9. Gene set enrichment analysis [
31] (GSEA) was employed to demonstrate the association between our genes of interest and defined gene sets. The gene annotation network analysis was performed using the GATHER [
32], BINGO [
33], and REVIGO [
34] software.
Establishment of the human PBMC-engrafted mouse model
This study was approved by the Institutional Review Board of Peking University People’s Hospital. Blood samples were collected from donors with written informed consent. NPG mice (NOD prkdc
scidIl2rg
null) were purchased from Beijing Vitalstar Biotech. Co. Ltd. The two- to threefold diluted blood samples were subjected to centrifugation on a lymphocyte separation medium (Tianjin Haoyang Biological Manufacture Co. Ltd.) at a density of 1.077 g/ml, and the nucleated cell layer between the plasma and separation medium was gathered. After two washes with RPMI 1640 medium, the PBMC pellets were suspended in RPMI 1640 medium at a density of 5 × 10
7 cells/ml. Then, 1 × 10
7 cells were injected via tail vein for each NPG mouse. Mice were housed in an SPF facility and accessed food and water ad libitum. The PBMC-transplanted mice were bled retro-orbitally every week, and the human CD45-positive cell rate in the mouse peripheral blood was analyzed by flow cytometry. The mice with more than 25% human CD45-positive cells in their blood were considered successful human PBMC-engrafted mouse models (see Additional files
1 and
2) and subjected to tumor cell inoculation.
Generation of xenografts
To evaluate the effect of the nivolumab treatment on primary tumor growth and spontaneous metastasis, 5 × 106 KHOS cells were injected subcutaneously into the right flanks of the humanized mice. On day 5 after the injection, the mice were randomly divided into two groups (n = 5 per group) and injected intraperitoneally with sterile saline or nivolumab at a dose of 10 mg/kg every 5 days for a total of five injections. The volume of the xenograft was measured every 5 days (tumor volume = (length × width2)/2). The mice were sacrificed after 30 days. At the termination of the study, the lungs were processed for routine hematoxylin-eosin (H&E) staining, and the number of metastatic nodules in the lung was determined. The tumors collected from the mice were chopped into small pieces and digested with 1 mg/ml collagenase type IV (Worthington, NJ) solution for 30 min. The dissociated tissues were processed through a 70-μm strainer, and the single-cell suspensions were subjected to flow cytometric analysis for human CD4+ and CD8+ lymphocyte infiltration.
Statistical analysis
All statistical analyses were performed using the SPSS v.21.0 software (SPSS, Chicago, IL, USA) and the GraphPad Prism software. For survival analysis, overall survival (OS) was defined as the time interval between the confirmed diagnosis and death or last follow-up. Survival analysis was per the Kaplan-Meier method with the log-rank test. The association between the expression levels of the PD-1 axis effectors and the clinicopathological variables, along with the relationships among the expression levels of the PD-1 axis effectors, were assessed by using the chi-square analysis. Statistical evaluations were performed using Student’s t tests. Data are expressed as the mean ± S.D. In all statistical analyses, a P value < 0.05 was considered statistically significant in the two-sided test.
Discussion
Known as the B7 family members, PD-L1 and PD-L2 both provide negative costimulatory signals during antigen-specific T cell activation by binding to the PD-1 receptor. Therefore, these three effectors play important roles in forming the tumor immunosuppressive microenvironment. Several studies have shown that inhibition of the PD-1 axis restores and enhances the immune response in vitro and in vivo [
35,
36]. Meanwhile, multiple clinical trials have suggested that blocking the interaction between PD-1 and PD-L1 can effectively inhibit tumor progression and improve patient prognosis [
5]. It is important to note that patients with high levels of PD-L1 expression have higher response rates to PD-1 antibody immunotherapy than patients with low PD-L1 expression [
37], but some patients with little or no expression of PD-L1 also respond to immunotherapy. This phenomenon may be due to PD-L2 expression in this cohort of patients; the PD-1 antibody may block the interaction between PD-L2 and PD-1. Therefore, many scholars have begun regarding PD-L1, PD-L2, and PD-1 as a whole and systematically studying the clinical value of these effectors as biomarkers to predict patient prognosis and evaluate the potential effects of the ICI treatment [
13‐
17].
Few studies have analyzed the clinical significance of the PD-1 axis, especially PD-L2, in sarcomas. In our study, we examined the expression levels of PD-L1, PD-L2, and PD-1 in multiple sarcomas, as indicated in the “
Results” section. The existing studies indicate that PD-L1 mRNA expression is detectable in osteosarcoma and exhibits a negative borderline trend with OS [
20]. Moreover, metastatic, but not primary, osteosarcoma tumors express PD-L1 and PD-1 [
21,
22], while a recent study has shown that PD-L1 is detectable in both primary and metastatic osteosarcomas, with no significant differences between them [
23]. In our study, no significant differences were observed in the expression levels of PD-L1, PD-L2, and PD-1 among the primary, recurrent, and metastatic osteosarcomas, which may be due to the relatively few metastatic osteosarcoma patients (5 patients). The positivity rates of PD-L1 and PD-1 in synovial sarcoma were similar to those in previous studies [
24,
25], while PD-L2 expression was detected in 26 cases (20.5%) for the first time. Furthermore, PD-L1 expression was significantly different according to clinical stage in synovial sarcoma, while PD-1 expression had a borderline difference. Similarly, a recent study reported that PD-L1 and PD-1 expression in tumor-invasive margins was significantly higher in metastatic synovial sarcoma than in primary synovial sarcoma and that PD-1 expression in the tumor-invasive margin was negatively associated with progression-free survival [
27]. We also found that PD-L1 was not expressed in conventional chondrosarcoma, while PD-L1 was detectable in DDCS, which was consistent with a previous study [
26]; this PD-L1 expression difference is possibly because the dedifferentiated component in DDCS may be osteosarcoma, which can express PD-L1. A previous study highlighted PD-L1 expression in some sarcoma patients with a history of GCT [
25], so we further investigated the expression levels of PD-L1, PD-L2, and PD-1 in GCT. As mentioned above, we detected high levels of PD-L1, PD-L2, and PD-1 expressions in GCT for the first time. In our study, a significant association was observed between expression of the combined PD-1 ligands and expression of PD-1 not only in sarcomas as a whole but also in osteosarcoma, synovial sarcoma, and chondrosarcoma, which indicates that PD-L1 and PD-L2 should be regarded as a whole for systematic investigations. Moreover, the expression patterns of PD-1 axis are varying in different tumors: the positive rates of PD-L1 and PD-L2 in lung adenocarcinoma are about 50% [
14]; PD-L1 and PD-L2 expressions are observed in 9.4 and 49.6% of renal cell carcinoma patients, respectively [
15]; in metastatic melanomas, positive rates of PD-L1 and PD-1 are 49 and 25%, respectively [
17]; high expressions of PD-L1 and PD-1 are detected in 38.4 and 50% of colorectal cancer patients, respectively [
19]. Variety of PD-1 axis expression in different tumors may lead to their different responses to immunotherapy. Previous study also indicated a statistically significant correlation between PD-L1 and PD-L2 in lung cancer [
18] and showed that PD-1, PD-L1, and PD-L2 have different prognostic roles in various tumors [
8‐
19].
Because the PD-1 axis effectors are differentially expressed in various sarcomas, the immunotherapeutic efficiency may vary widely due to the pathological type of the sarcoma, so further investigation of each sarcoma is urgently needed. Based on our findings, osteosarcoma, which is the most common primary malignant bone tumor with a high mortality and disability rate, exhibits relatively high expression levels of PD-1 axis effectors. Furthermore, the IHC experiments and data mining both indicate that PD-L1 was negatively correlated with prognosis, while PD-1 had a negative correlation tendency with OS. Meanwhile, PD-L2 had a positive correlation trend with OS.
To investigate the potential mechanisms that underlie the PD-1, PD-L1, and PD-L2 associations with prognosis, mRNA expression of 10 immune checkpoint-related genes in the osteosarcoma samples were clustered and visualized through datamining and bioinformatic analyses. The expression patterns, GSEA, and gene annotation network analysis of the genes associated with PD-1, PD-L1, or PD-L2 have also been presented. In addition to immune suppression, our results indicate that PD-1 may be correlated with docetaxel resistance and activation of MAPK and metastasis-related pathways. The PD-L1-associated poor prognosis may due to immune suppression, cisplatin resistance, and metastasis-related pathway activation, whereas PD-L2 may slow osteosarcoma progression by repressing DNA repair, stem cell-related pathways, and doxorubicin resistance.
Based on the negative prognostic role of PD-1/PD-L1 and the immune response associated with the PD-1 axis in the datamining analysis, we investigated whether blockade of the PD-1 axis could generate an antitumor effect in osteosarcoma. In our study, we revealed that treatment with nivolumab resulted in effective control of pulmonary metastasis in an osteosarcoma model in humanized mice, while no obvious effect was evident on localized osteosarcoma. Moreover, the tumor-infiltrating lymphocyte (TIL) investigation indicated that nivolumab increased CD4+ and CD8+ lymphocytes in the lung but not in the primary lesion. Furthermore, nivolumab enhances the cytolytic activity of CD8 lymphocytes in the lung. The limitation of this animal model was that the mature human lymphocytes gradually initiated severe graft-versus-host disease in the mouse due to the PBMC injection, which resulted in a relatively short survival time. Therefore, we could not determine whether nivolumab could control tumor growth over the long term, even though the tumor growth rate of the nivolumab-treated group had begun to slow down in our study.
As we know, ICI inhibits tumor development by restoring the functions of T cells to kill the tumor cells, and the quantity of TILs plays an important role in the immunotherapy effect. In our study, nivolumab inhibited osteosarcoma metastasis by increasing the number of lymphocytes in the lung but have been ineffective towards primary osteosarcoma. Interestingly, in recent studies [
38,
39], patients with PD-L1/PD-1 monoclonal antibody therapy showed a pattern of rapid disease progression. One reason suggested by the authors was that the PD-1/PD-L1 axis mediated inherent functions in the tumor cells and that the PD-1/PD-L1 blockade may have affected the tumor cell-intrinsic signaling network and subsequently enhanced tumor growth or progression. This indicates that the treatment effect of ICI may be associated with tumor cell-intrinsic signaling of PD-L1 and PD-L2. Several studies have found that PD-L1 and PD-L2 are associated with multiple cellular biological behaviors, such as the epithelial-mesenchymal transition (EMT), proliferation, and autophagy [
40‐
43]. At present, no relevant studies have been conducted to address the tumor cell-intrinsic effects of PD-L1 or PD-L2 in osteosarcoma; thus, further investigations should be undertaken to improve the treatment effect of ICI.