Tumor microenvironment in giant cell tumor of bone: evaluation of PD-L1 expression and SIRPα infiltration after denosumab treatment

Giant cell tumor of bone (GCTB) is an intermediate malignant bone tumor that is locally aggressive and rarely metastasizes. Denosumab, which is a receptor activator of nuclear factor kappa B ligand (RANKL) inhibitor, can be used to treat GCTB. We focused on potential immunotherapy for GCTB and investigated the tumor microenvironment of GCTB. Programmed death-ligand 1 (PD-L1) and indoleamine 2,3-dioxygenase 1 (IDO1) expression and signal-regulatory protein alpha (SIRPα), forkhead box P3 (FOXP3), and cluster of differentiation 8 (CD8) infiltration were assessed by immunohistochemical studies of 137 tumor tissues from 96 patients. Of the naive primary specimens, 28% exhibited PD-L1 expression and 39% exhibited IDO1 expression. There was significantly more SIRPα+, FOXP3+, and CD8+ cell infiltration in PD-L1- and IDO1-positive tumors than in PD-L1- and IDO1-negative tumors. The frequency of PD-L1 expression and SIRPα+ cell infiltration in recurrent lesions treated with denosumab was significantly higher than in primary lesions and recurrent lesions not treated with denosumab. PD-L1 expression and higher SIRPα+ cell infiltration were significantly correlated with shorter recurrence-free survival. PD-L1 and SIRPα immune checkpoint inhibitors may provide clinical benefit in GCTB patients with recurrent lesions after denosumab therapy.


Results
Clinical results. The clinicopathological features of the subjects are shown in Table 1. The median age at initial diagnosis was 33  years. The subjects included 54 females and 42 males. The tumors were mainly located in the femur or tibia (Table 1). Four patients did not undergo surgery and were treated only with denosumab. Ten patients received denosumab (six for neoadjuvant therapy and four for recurrence). Recurrenceand metastasis-free survival data were available for 78 patients, with follow-up ranging from 1 to 332 months (median: 58 months). In this study, 18 patients (23%) had local recurrence and eight (10%) had distant metastasis. All patients with metastasis developed pulmonary metastases. There were no patients with tumor-related death.
Radiological features. Radiological images, including plain radiographs, computed tomography scans, and/or magnetic resonance imaging scans, or medical records were available for 76 patients. Maximum diameters (N = 68) and pathological fractures (N = 76) were investigated. The radiological characteristics are shown in Table 1. Radiologically, the median of the maximum diameters was 4.3 cm (range: 0.6-7.9 cm). A pathological fracture was apparent in six of 76 cases (8%).
Alteration of PD-L1 and IDO1 expression and SIRPα + macrophage and FOXP3 + and CD8 + lymphocyte infiltration after denosumab treatment. Differences in PD-L1 and IDO1 expression between primary, ND-Rec, and D-Rec lesions. We evaluated PD-L1 and IDO1 expression in primary lesions, recurrent lesions after denosumab treatment (D-Rec), and recurrent lesions not treated with denosumab (ND-Rec). The frequency of PD-L1 expression in D-Rec lesions was significantly higher than in primary and ND-Rec lesions (P = 0.0243) (Fig. 3a). There were no significant differences in IDO1 expression between primary, ND-Rec, and D-Rec lesions (Fig. 3b).
Differences in SIRPα, FOXP3, and CD8 infiltration between primary, ND-Rec, and D-Rec lesions. We compared SIRPα + , FOXP3 + , and CD8 + cell counts between primary, ND-Rec, and D-Rec lesions. SIRPα + cell infiltration in D-Rec lesions was significantly increased compared with that in primary lesions and ND-Rec lesions (P = 0.0074 and P = 0.0188, respectively) (Fig. 4a). Representative figures of PD-L1 and SIRPα that compared pre-and post-denosumab treatment belong to same patient were shown in Fig. 5. There was no significant difference in SIRPα + cell infiltration between primary tumors and ND-Rec lesions. FOXP3 and CD8 infiltration in primary, ND-Rec, and D-Rec lesions did not reach statistical significance (Fig. 4b,c). We assessed the prognostic significance of clinico-radio-pathological characteristics, PD-L1 and IDO1 expression, and SIRPα + , FOXP3 + and CD8 + cell infiltration by using Kaplan-Meier survival analysis (Table 3). PD-L1 expression and a high number of SIRPα + cells were correlated with shorter recurrence-free survival (P = 0.0355 and P = 0.0243, respectively) ( Table 3, Fig. 6a,b). There was no significant correlation between recurrence-free survival and preoperative den-Table1. Clinico-radio-pathological features.  Localization of SIRPα on mononuclear cells of GCTB. Moreover, to verify which cells express SIRPα, double immunofluorescence was performed for SIRPα and CD14 (n = 10). Representative images are shown jn Fig. 8. SIRPα positive cells were diffusely found and some of these positive cells also expressed CD14 (monocyte marker).

Discussion
Classically, GCTB is treated with chemical adjuvant and/or surgery to reduce local recurrence. However, the rate of local recurrence after chemical adjuvant therapy is 15-50% 2,14 . Local recurrence significantly debases patients' activities of daily living and quality of life. Thus, prevention of local recurrence is an important part of GCTB treatment. Denosumab, which is a human monoclonal anti-RANKL antibody that inhibits the RANK/RANKL signaling pathway, has been introduced to treat GCTB 12,13,34 . Pathological findings have previously revealed osteoclast-like multinucleated giant cell and mononuclear stromal cell depletion and new bone formation after denosumab treatment 11,35,36 . GCTB sometimes shows dramatic histological changes and mimics primary malignant bone tumors after denosumab treatment 35,36 . Kato et al. reported that after denosumab therapy, mononuclear tumor cells with H3F3A mutations could survive, but osteoclast cells could not survive without RANK/RANKL signaling 6 . Some reports have recently suggested that preoperative denosumab administration might increase the recurrence rate after operation 34,37 . New bone formation may be difficult to distinguish from preexisting bone trabeculae and make it difficult to identify true surgical margins 34 . No consensus has been reached about whether preoperative denosumab treatment might be useful to prevent local recurrence. Some GCTB cases are refractory to denosumab and have frequent recurrence. Thus, novel therapeutic targets, such as immunotherapy against GCTB, are required.
There have been few studies concerning tumor immunity in GCTB and immune microenvironment alterations after denosumab therapy. To understand the tumor microenvironment of GCTB, we established this study to focus on the tumor immune microenvironment of primary lesions and recurrent lesions with and without denosumab therapy in GCTB patients, and to investigate an immune checkpoint inhibitor treatment strategy for GCTB patients. A prior immunohistochemical study showed PD-L1 expression by tumor cells and    (34) 12 (13) 27 (29) 16 (17) 20 (22) 23 (25) 19 (21) 24 (27) 18 (20) 25 (28) Gender Male 27 (29) 13 (14)        www.nature.com/scientificreports/ multinucleated giant cells in 28.3% of GCTB specimens, and this was associated with shortened disease-free survival and high Ki-67 positivity 22 . Moreover, this previous study focused on two immune-related genes, TLR8 and LCK, which are related to innate immunity activation and CD4 + and CD8 + lymphocyte development. Lower TLR8 and LCK expression were correlated with PD-L1 immunoexpression positivity, but after denosumab treatment, TLR8 and LCK expression increased 22 . Our results showed PD-L1 immunoexpression in about 28% of primary specimens without denosumab treatment, and this was significantly related to shortened recurrence-free survival. Our double immunofluorescence staining study showed that PD-L1 expression in both neoplastic cells   [16][17][18][19][20][21] . Our results suggested that tumor cells with H3F3A mutation and mononuclear histiocytoid cells might cooperate to express PD-L1 and escape tumor immunity because neoplastic cells and non-neoplastic cells expressed PD-L1. Moreover, our double immunofluorescence staining showed that SIRPα positive cells were diffusely found and some of these positive cells also expressed CD14 (monocyte marker). In GCTB, histologically mononuclear cells are tumor cells or monocytes 11 . Therefore, we suggested that tumor cells also would express SIRPα, although double staining for H3G34W and SIRPα was not possible in our study. There was the possibility of immune alterations associated with recurrence and denosumab treatment. More specifically, PD-L1 and SIRPα were more upregulated in recurrent tumors than Table 3. Univariate and multivariate for recurrence-free survival. Bold value indicates significant differences. * There was no case with more than 5 years follow-up. PD-L1; programmed death ligand-1, IDO-1: indoleamine 2,3-dioxygenase-1, SIRPα: signal-regulatory protein α, FOXP3: forkhead box P3, CD8: cluster of differentiation 8, R-S survival; Recurence-free survival. www.nature.com/scientificreports/ in primary tumors due to resistance to treatment. It is suggested that tumor cells and/or non-neoplastic cells escaped tumor immunity by expressing PD-L1 and SIRPα. Thus, patients with uncontrollable recurrent lesions may be treated with anti-PD-L1 or anti-PD-1 inhibitors after denosumab therapy.

Factors 5 yr R-FS(%)
On the other hand, the use of widely variable combinations of immune checkpoint inhibitors has gained great interest to improve immunotherapy outcomes 31 . For instance, macrophage-related immune checkpoints and CD47/SIRPα interactions have been focused on as new therapeutic targets of immunotherapy 31,32,38,39 . We showed that high SIRPα + infiltration in primary specimens was associated with shorter recurrence-free survival   www.nature.com/scientificreports/ by using univariate and Cox multivariate analyses. SIRPα + macrophages were frequently seen in recurrent specimens treated with denosumab rather than in primary specimens with frequent PD-L1 immunopositivity. Previous studies have reported that SIRPα + macrophage infiltrates were correlated with shorter overall survival and progression-free survival in other malignancies, such as diffuse large B-cell lymphoma 32 and melanoma and renal cell carcinoma 39 . However, other reports have also shown a positive correlation between SIRPα expression and good prognosis 33 . In our immunohistochemical studies, it was suggested that SIRPα enabled tumor cells to evade phagocytosis, leading to promoted tumor growth and progression, similar to PD-L1 expression in GCTB patients. Treatment inhibiting CD47 and SIRP interactions has been investigated in previous studies, with clinical trials in hematopoietic and solid cancers [40][41][42] . We suggest that anti-SIRPα inhibitors will become a treatment for GCTB patients, especially patients with frequent recurrence. Several studies have reported that RANK/RANKL inhibitors have the potential to increase the effectiveness of immune checkpoint inhibitors [43][44][45][46][47][48] . In a mouse study, Ahern et al. showed that combination therapy with anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) inhibitors and anti-RANKL inhibitors increased CD8-positive T-cell infiltration compared to anti-CTLA-4 inhibitors or anti-RANKL inhibitors alone 46 . Our immunohistochemical study results showed no significant differences in CD8 infiltration between primary, ND-Rec, and D-Rec specimens. However, it was suggested that in GCTB, CD8 + lymphocyte infiltration would increase with combined immunotherapy and denosumab therapy. In a previous study, patients with lung cancer or melanoma who received both anti-CTLA-4 and/or anti-PD-1 inhibitors and denosumab had good disease control rates compared to patients not receiving denosumab 48 . Moreover, the authors of this previous study claimed that longer use of combined therapy was preferable to control tumor progression 48 . Although denosumab is widely used for GCTB treatment, the potential of recurrence remains. In our study, PD-L1 and SIRP expression in recurrent GCTB with denosumab treatment was higher than in primary GCTB and recurrent GCTB without denosumab treatment. Our findings suggested that the combination of PD-L1/PD-1 and/or SIRP inhibitors and denosumab might be effective for controlling recurrent GCTB. Further studies using a larger number of cases are required to confirm the effectiveness of these treatments. To the best of our knowledge, this study is the first to report SIRP expression in GCTB and investigate alterations in primary specimens and recurrent specimens with and without denosumab treatment. From the immunohistochemical and immunofluorescence results in this study, it can be concluded that PD-L1 and SIRPα immune checkpoint inhibitors may provide clinical benefits in GCTB patients with uncontrollable recurrent lesions after denosumab therapy.

Materials and methods
Patients and tissue samples. We used samples of GCTB patients registered from 1984 to 2019 in the database of the Department of Anatomic Pathology, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan. All cases were reviewed based on histological examinations with hematoxylin and eosin staining and immunohistochemical studies using antibodies specific to GCTB (anti-H3.3G34W, anti-H3.3G34R, and anti-H3.3G34V) 4,7,11 .
This study included 137 formalin-fixed, paraffin-embedded samples from 96 patients. The samples included primary conventional GCTB, recurrent conventional GCTB, post-denosumab GCTB, and lung metastasis of conventional GCTB. The lesions were collected by biopsy or resection. Clinical data, including age at diagnosis, sex, and tumor site, were collected. The presence of pathological fractures was investigated via plain radiographs, computed tomography scans, and/or magnetic resonance imaging scans that were examined by a radiologist (J.M.). Morphological features, mitotic figures, osteoclastic giant cells, foamy macrophages, bone formation, spindle cell features, and secondary aneurysmal bone cystic changes were also evaluated and investigated by three pathologists (Y.T., K.K., and Y.Y.). The institutional review board at Kyushu University approved this study (approval codes: 29-625 and 29-429). Written informed consent was obtained from the patients and their parents/ guardians prior to tissue collection. All experiments were performed in accordance with guidelines provided by the Ethics Committees and Institutional Review Boards.
Immunohistochemical staining. For the immunohistochemical and immunofluorescence studies, formalin-fixed, paraffin-embedded tissues were sliced into 4-μm sections. The immunohistochemical studies were performed as previously described 20,29 . The following rabbit and mouse monoclonal antibodies were used as the primary antibodies: anti-PD-L1, anti-IDO1, anti-SIRPα, anti-FOXP3, anti-CD8, anti-H3.3G34W, anti-H3.3G34R, and anti-H3.3G34V (Supplementary Table S1). Appropriate controls were used throughout. Three pathologists (Y.T., K.K., and Y.Y.) independently evaluated the immunohistochemical staining results for each sample. For the immunohistochemical evaluation of PD-L1 and IDO1, the membrane PD-L1 expression and cytoplasmic IDO1 expression were defined by combined proportion scores, which evaluates on tumor cells and tumor-associated immune cells. Cases with a combined proportion score ≥ 1% were defined as positive. Moreover, SIRPα + , CD8 + , and FOXP3 + cells were counted per high power field in five dependent fields for each case. This process was based on previous studies, with modifications 32,33,49-51 . Statistically, the median numbers of SIRPα + , CD8 + , and FOXP3 + cells were determined as the cut-off points.