Introduction

Giant cell lesions of bone are relatively rare tumors with low metastatic potential but they can be locally aggressive [2, 3, 29]. Giant cell lesions include central giant cell lesions of the maxillofacial skeleton and of the axial and appendicular skeleton [42]. Clinical and radiographic criteria are used to stratify giant cell lesions into aggressive and nonaggressive tumors as histologic criteria have not been correlated with tumor behavior [2, 18, 38]. Surgery remains the dominant approach but systemic therapy targeting receptor activator of nuclear factor kappa-B ligand (RANKL) with a monoclonal antibody (denosumab) has been shown to be an effective adjuvant [6]. However, the long-term effect of systemically targeting RANKL is poorly understood and it is not curative as giant cell lesions recur once the therapy is stopped [22]. This has prompted further investigation in the pathophysiology of giant cell lesions with the hopes of identifying other targets. Histologically, there are mainly two cell populations in giant cell lesions that can act as potential targets for novel therapies. These are multinucleated osteoclastlike giant cells and mononuclear, spindle-shaped fibroblastlike, stromal cells [20, 23, 27]. Mononuclear stromal cells are thought to be the neoplastic cell as the multinucleated giant cells disappear with denosumab treatment yet the mononuclear stromal cells persist, and mononuclear stromal cells can be indefinitely propagated in culture [57]. For these reasons, it is useful to consider mononuclear stromal cells separately from multinucleated giant cells.

The role of immune surveillance in giant cell tumors of bone was suggested in an in vitro analysis of lymphocytic activation where radiologically more-aggressive tumors had decreased lymphocytic activation [35]. The role of immune surveillance has regained importance as tumor infiltrating lymphocytes have been correlated with outcomes in several cancers. Unfortunately, many cancer cells have evolved tumor escape mechanisms to avoid immune surveillance such as defects in human leukocyte antigen (HLA) class I antigen processing machinery components leading to ineffective presentation of tumor antigen-derived peptide complexes to their cognate T cell receptor [9], and blocking activation of cytotoxic lymphocytes via the interaction of immune modulators such as B7-H3 [41].The HLA class I molecule is a heterodimer that consists of two polypeptide heavy α chains and a β2-microglobulin light chain. However, multiple intracellular molecules are needed to effectively present antigens to their cognate T cells, and defects in the intracellular processing of antigens and other stages of antigen processing machinery also have been correlated with poor survival in several malignancies [32, 47]. Despite having functional HLA class I antigen, some tumors can still escape the immune system by expressing coinhibitory molecules that deactivate cytotoxic lymphocytes cells. B7-H3 is a member of the B7 family of molecules involved in immune regulation and is overexpressed on a wide variety of solid tumor types [41]. It has been shown that B7-H3 exerts an immunosuppressive action in various cancers [11, 55, 56]. The aims of our study were to assess the frequency of HLA class I antigen defects in a large collection of well-annotated giant cell lesions, and to correlate these results with lymphocyte infiltration and the clinical course of the disease. In addition, we explored the immunologic role of B7-H3 and its association with local invasion in giant cell lesions.

Therefore we asked: (1) Was an immune response elicited by giant cell lesions? (2) Did clinically relevant HLA defects exist in giant cell lesions? (3) Was B7-H3 a clinically relevant immune modulator?

Methods

Study Design and Setting

This was a longitudinal study of a cohort of patients with giant cell lesions of bone treated at Massachusetts General Hospital during a 15-year period (January 1993 to June 2008).

Participants and Study Subjects

Subjects were identified through the Massachusetts General Hospital giant cell patient registry [38, 39]. Patients were included if they had histologically confirmed central giant cell lesions of the maxillofacial skeleton that were treated by the Department of Oral and Maxillofacial Surgery or axial and appendicular giant cell tumors that were treated by the Department of Orthopaedic Surgery, with a minimum followup of 6 months. Histologic examination results were confirmed as reported by the official interpretation of our musculoskeletal pathologist (WCF) during clinical care of the patient. Patients with followup less than 6 months (n = 26) were excluded. Patients with systemic diseases (n = 4), such as hyperparathyroidism, and giant cell lesions associated syndromes (n = 4), such as cherubism, and Noonan syndrome, also were excluded as these conditions are known to be associated with giant cell lesions and may represent a separate pathogenesis from a solitary giant cell lesion. Patients also were excluded if insufficient tumor tissue was available (n = 2), or if insufficient documentation was encountered (n = 7). A total of 43 patients with mean age of 36 ±14 years (52% were female patients) were excluded. Ninety-three patients with a mean age of 28 ± 12 years (56% were female patients) who underwent surgical resection of their giant cell lesions were included in this study (Table 1). Any bias inherent in this process was unintentional. It seems unlikely that bias inherent in this retrospective study altered the outcome sufficiently. The project was approved by the institutional review board (protocol 2008-P-000563).

Table 1 Patient and tumor characterization
Full size table

Description of Experiment, Treatment, or Surgery

Selected tissue blocks and accompanying hematoxylin and eosin-stained slides were reviewed by a senior musculoskeletal pathologist (WCF) to assess for adequate tumor tissue. A representative region of each tumor block was identified and extracted to build the tissue microarray. One-millimeter cores were extracted with a maximum of three cores per block. Cores from human cartilage, human spleen, human liver, human lymph node, human melanoma metastasis, melanoma xenograft, and mouse liver also were included in the tissue microarray. Monoclonal antibodies (mAb) hydroxycarboxylic acid receptors (HCA2), which recognize β2-microglobulin-free HLA-A (excluding-A24), -B7301, and -G heavy chains [48, 49]; mAb HC-10, which recognizes β2-microglobulin-free HLA-A3, -A10, -A28, -A29, -A30, -A31, -A32, -A33, and all β2-microglobulin-free-HLA-B (excluding -B5702, -B5804, and -B73) and HLA-C heavy chains [40, 48, 49]; and mAb NAMB-1, which recognize HLA class I heavy chain- associated and free β2 microglobulin [7]; and mAb LGII-612.14 which recognizes a monomorphic epitope expressed on the β chain of HLA-DR, -DQ, and -DP antigens [51] were developed and characterized as described. mAbs were purified from ascitic fluid by affinity chromatography on a Protein G column (GE Healthcare Life Sciences, Pittsburgh, PA, USA). The purity and activity of mAb preparations was monitored by sodium dodecyl sulfate polyacrylamide gel electrophoresis and by binding assays with the cognate antigen, respectively. The B7-H3-specific mAb 1027 was obtained from R&D Systems (Minneapolis, MN, USA) [10, 15, 53]. A binding assay using purified recombinant B7-H3 was used to validate the affinity of the B7-H3-specific mAb 1027 to B7-H3. The human CD8 (clone 4B12) and CD4-specific (EPR6844) mAbs were obtained from DAKO (Carpinteria, CA, USA) and Abcam (Cambridge, MA, USA), respectively.

Secondary Antibody System

The DAKO EnVisionTM+System horseradish peroxidase-labeled polymer antimouse (K4001) kit was obtained from DAKO.

Immunohistochemical Staining

Four micrometers of formalin-fixed, paraffin-embedded tissue sections from the giant cell lesions tissue microarray block were used as substrates in immunohistochemical reactions. Tissue microarrays were stained with HLA class I-, HLA class II- and B7-H3-specific mAb as previously described [37]. Staining intensity and percentage of stained tumor cells in each lesion were reviewed and enumerated by two investigators (CH, LC) who had no knowledge of the patients’ characteristics or clinical outcomes. Results were graded as positive, heterogeneous, or negative when the total score in an entire lesion was greater than 75, 75 to 25, and less than 25, respectively [54]. Staining with CD8- and CD4-specific mAbs was performed according to the manufacturer’s (DAKO) instructions, and the results of staining were calculated by counting the number of stained infiltrating cells in four high-powered fields (×40) of the giant cell lesion cores. Giant cell lesions are formed principally by two cell types which are multinucleated osteoclastlike giant cells and mononuclear, spindle-shaped fibroblastlike, stromal cells [57]. Immunohistochemical staining was scored individually for each of the two cellular subtypes.

Clinical Data Stratification of Aggressive Versus Nonaggressive Lesions

We used Enneking’s staging system for benign bone tumors [18]. This system focuses on the clinical behavior of the tumors. Inactive lesions are designated Stage I, whereas tumors that are active (show progressive growth, radiologic deformation of the bony cortex) are Stage II, and aggressive lesions (destructive lesions with a soft tissue mass) are Stage III. For the purpose of our study, we classified the tumor lesions as either aggressive or nonaggressive, as described previously [42]. Briefly, for axial and appendicular lesions all Enneking Stage III tumors and any Stage II tumors with pathologic fracture and/or recurrence were considered aggressive. All other axial and appendicular tumors were categorized as nonaggressive [42]. Central giant cell lesions of the maxillofacial skeleton were characterized as aggressive or nonaggressive according to the clinical and radiographic criteria defined by Chuong et al. [13] in 1986 and modified by Kaban et al. [28] in 2007.

Research Questions

The first question we asked is whether the immune system is recognizing and responding to the giant cell lesions. One way to assess whether the host is responding to tumor cells is to assess whether CD8+ lymphocytes are infiltrating the tumors. Many of our normal cells express HLA class I antigens on their cell surface and it is here that they present antigens to CD8+ lymphocytes. If the antigen presented by the HLA expressing cell is abnormal, then the CD8+ lymphocyte will destroy the tumor cell. This is part of the normal immunosurveillance function of our immune system [16]. This leads to our second question, which is whether HLA class I antigen expression was low in our aggressive giant cell lesions. If HLA class I antigen expression is low in our tumor cells, CD8+ lymphocytes will not be able to detect the tumor cells, and therefore they will not be activated nor will they mount an immune response and the tumor cells will be able to evade this aspect of the immune system leading to a more aggressive phenotype [9]. The last question we asked pertains to the B7-H3 molecule which is an immune-modulator [35]. Immune-modulators are an important mechanism by which our host prevents an overzealous response by lymphocytes, which, if left unchecked, could lead to autoimmune disease. These immune-modulators are essentially check points that can suppress or stimulate the host’s response. Unfortunately, some cancers have adopted the immune suppressive effects of inhibitory immune-modulators, such as B7-H3, which allows them to escape the immune response and invade surrounding tissues [34]. We examined the association between B7-H3 and tumor aggressiveness and local tumor invasion.

Statistical Analysis and Study Size

Statistical analyses were performed using IBM SPSS statistical software Version 20.0 (IBM Corporation, Armonk, NY, USA). A chi-square test was used to assess association between the different parameters, and the Mann-Whitney U test was used to compare means of lymphocyte cell infiltration. Spearman’s rank correlation coefficient was used to assess correlations. A probability less than 0.05 was considered statistically significant.

Results

Was an Immune Response Elicited by Giant Cell Lesions?

Tumor infiltrating lymphocytes in primary giant cell lesions CD4+ and CD8+ cell infiltration were found in 90% (76 of 85) and 84% (71 of 85) of giant cell lesions, respectively. The number of CD4+ cells ranged between 0 and 93 (mean number, 14.3), whereas that of CD8+ cells ranged between 1 and 184 (mean number, 24.2). Staining patterns of giant cell lesions with CD4+ and CD8+ tumor infiltrating lymphocytes are shown (Fig. 1). There was a weak positive correlation between CD4+ and CD8+ cell infiltration (correlation coefficient, 0.224; p = 0.039). Interestingly, the number of CD8+ cells was lower in aggressive tumors (median, 4.8; interquartile range [IQR], 0.4–13.4) when compared with nonaggressive tumors (median, 15.8; IQR, 4.3–46.3; p = 0.007) (Fig. 2). We then investigated the prognostic value of HLA class I antigen expression in combination with CD8+ T cell infiltration level. Low expression of HLA-B and -C antigens in combination with low CD8+ T cell infiltration was associated with highest risk of tumor aggressiveness (odds ratio [OR],7.81; 95% CI, 1.4–43.5; p = 0.011) when compared with high HLA-B and -C antigen expression and high CD8+ T cell infiltration (Fig. 3).

Fig. 1
figure 1

Representative chromogenic immunohistochemistry staining results of the giant cell lesion tissue microarray are shown for CD4+ and CD8+ tumor infiltrating lymphocytes. The staining was performed using CD8+ and CD4+ specific mAbs according to the manufacturer’s instructions (DAKO, Carpinteria, CA, USA). The results of the staining were calculated by counting the number of stained infiltrating cells in four high-powered fields (×40) in each tissue microarray core. Representative high T cell infiltration and low T cell infiltration are shown. The red arrows point to the stained T cells.

Full size image
Fig. 2
figure 2

The boxplot shows a comparison of CD8+ tumor infiltrating cell count between aggressive and nonaggressive giant cell lesions of the bone. Aggressive tumors show a significantly (p = 0.007) low number of CD8+ tumor infiltrating cells (median, 4.8; IQR, 0.4–13.4) compared with nonaggressive tumors (median, 15.8; IQR, 4.3–46.3).

Full size image
Fig. 3
figure 3

The bar graph shows the relative number of giant cell lesions with high HLA-B and -C expression and high CD8+ tumor infiltrating cells (TILs) versus low HLA-B and -C expression and low CD8+ TILs stratified by tumor aggressiveness. HLA-B and -C antigen low expression in combination with low CD8+ T cell infiltration was associated with highest risk of tumor aggressiveness (OR, 7.81; 95% CI, 1.4–43.5; p = 0.011) when compared with high HLA-B and -C antigen expression and high CD8+ T cell infiltration.

Full size image

Did Clinically Relevant HLA Defects Exist in Giant Cell Lesions?

Expression of HLA-A antigens and HLA-B and -C antigens in mononuclear stromal cells were low and heterogeneous in 68% (56 of 83) and 53% (45 of 84) of the tumors, respectively. Similarly, expression of HLA-A antigens and HLA-B and -C antigens in multinucleated giant cells were low and heterogeneous in 65% (49 of 75) and 73% (56 of 77) of the tumors, respectively. There was no difference in the percentage of HLA-A and HLA-B and -C expression in mononuclear stromal cells (percentage difference, 1%; 95% CI, 12%–14%; p = 0.85) and multinucleated giant cells (percentage difference, 2%; 95% CI, −17% to 20%; p = 0.83). Representative staining patterns of giant cell lesions with HLA-A-specific mAb HCA2 and with HLA-B and -C specific mAb HC-10 are shown (Fig. 4). Beta 2-microglobulin expression was low and heterogeneous in 89% (75 of 84) of the mononuclear stromal cells and 51% (39 of 76) of the multinucleated giant cells. Low β2-microglobulin expression level in mononuclear stromal cells was associated with low HLA-A heavy chain (OR, 22.05; 95% CI, 4.00–121.58; p < 0.001) and low HLA-B and -C heavy chain (OR, 13.00; 95% CI, 2.79–60.49; p < 0. 001). Similarly, low β2-microglobulin expression level in multinucleated giant cells of β2-microglobulin was associated with low HLA-A heavy chain (OR, 7.81; 95% CI, 1.4–43.5; p = 0.011) and low HLA-B and -C heavy chain expression (OR, 7.81; 95% CI, 1.4–43.5; p = 0.011). Low expression of HLA-B and -C heavy chain, but not HLA-A heavy chain was associated with aggressive tumors (OR, 4.30; 95% CI, 1.50–12.35; p = 0.005) (Table 2). Similarly, low and heterogeneous β2-microglobulin expression was associated with aggressive tumors (OR, 4.67; 95% CI, 1.29–12.35; p = 0.031). However, when adjusting the data for HLA-B and -C heavy chain expression, we observed no association between low and heterogeneous β2-microglobulin and aggressive tumors (OR, 0.25; 95% CI, 0.03– 2.35; p = 0.383). HLA class II antigens were expressed by mononuclear stromal cells in 69% (58 of 84) of the giant cell lesions; the expression was scored as positive in 26% (22 of 84) lesions and heterogeneous in 43% (36 of 84). HLA class II antigen expression in mononuclear stromal cells was not associated with HLA class I antigen expression in mononuclear stromal cells. However, high expression of HLA class II in mononuclear stromal cells was associated with local tumor invasion to surrounding soft tissue (OR, 6.56; 95% CI, 1.40–30.56; Fisher’s exact test, p = 0.019). HLA class II antigen expression was low or negative in 75 of the 77 samples tested. Two samples were heterogenous and none of them had high expression. HLA class II antigen expression was not associated with HLA class I, or with either CD4+ or CD8+ T cell infiltration.

Fig. 4
figure 4

Representative chromogenic immunohistochemistry staining results of the giant cell lesions tissue microarray are shown for the five immunologic markers HLA class I HLA-A heavy chain (stained using hydroxycarboxylic acid receptors (HCA2) mAb), HLA class I HLA-B and -C (stained using HC-10 mAb), HLA class I β-2 using microglobulin (β2m) light chain (stained using namb-1 mAb), HLA class II (stained using lgii-612.14 mAb), and B7-H3 (stained using B7-H3-specific mAb 1027). Each tissue microarray core was scored as low, heterogeneous, or high if the percentage of the stained area lesion was less than 25%, 25% to 75%, and greater than 75%, respectively.

Full size image
Table 2 Association between tumor immunologic markers and clinical parameters
Full size table

Was B7-H3 a Clinically and Immunologically Relevant Marker?

B7-H3 was expressed in all giant cell lesions (100%; n = 83). The expression was low and heterogeneous in 63% (52 of 83), heterogeneous in 60% (50 of 83), and low in 3% (two of 83); and in multinucleated giant cells in 51% (39 of 76), heterogeneous in 41% (31 of 76), and low in 10% (eight of 76) of tumors. High B7-H3 expression in mononuclear stromal cells was associated with high HLA-B and -C heavy chain expression (OR, 2.9; 95% CI, 1.16–7.33; p = 0.022), but not with HLA-A heavy chain expression. High B7-H3 expression in mononuclear stromal cells was associated with high CD8+ cell infiltration (OR, 3.64; 95% CI, 1.4–9.38; p = 0.006), but not with CD4+ cell infiltration. High B7-H3 expression in mononuclear stromal cells was associated with high HLA class II antigen expression (OR, 6.32; 95% CI, 1.70–23.50; p = 0.003). B7-H3 expression, by itself, was not associated with tumor aggressiveness as defined in Methods (OR, 1.42; 95% CI, 0.51–3.97; p = 0.50). However, high B7-H3 expression was associated with local tumor invasion (soft tissue mass) (relative risk [RR], 3.12; 95% CI, 2.13–4.54; p < 0.001). When stratifying the data according to CD8+ cell infiltration (high versus low), the association between high B7-H3 and local tumor invasion was seen only in the high CD8 T cell infiltration group (RR, 1.97; 95% CI, 1.2–4.27; p = 0.028). When stratifying the data according to the expression of B7-H3 and CD8+ tumor infiltrating lymphocytes in mononuclear stromal cells, the high B7-H3 expression combined with low CD8+ tumor infiltrating cells stratum was associated with increased risk of having an aggressive tumor when compared with low B7-H3 expression and high CD8+ tumor infiltrating cells stratum (OR, 8.89; p = 0.0491) (Fig. 5).

Fig. 5
figure 5

The bar graph shows the relative number of giant cell lesions and their respective expression of B7-H3 and CD8+ tumor infiltrating cells stratified by tumor aggressiveness. The giant cell lesions with low B7-H3 and low CD8+ tumor infiltrating cells were statistically more likely to be aggressive tumors (p < 0.001).

Full size image

Discussion

The role of the immune system in giant cell lesions is unknown. However, in general, there has been renewed interest in the role of the immune system in how cancers progress with the hopes of improving our understanding and possibly identifying targets for therapy [16]. Clinical trials have shown encouraging results in a subset of patients with various types of malignancies [17, 21, 25, 43, 50]. Specifically, the administration of mAbs which blocks interaction of immune regulatory checkpoint molecules such as programmed cell death protein 1 (PD-1) with their ligands programmed cell death ligand 1 (PD-L1), have shown a major effect on the clinical course of various types of malignancies [1, 5, 12, 25, 26, 52]. However the efficacy of this novel immunotherapeutic strategy is limited to less than 30% of patients [45]. These findings have stimulated an interest in exploring other immune regulatory checkpoints such as B7-H3 and characterizing the host’s immune response. Currently there are three ongoing Phase 1 clinical trials (NCT02381314, NCT02475213, NCT02982941) that are evaluating enoblituzumab, which is a monoclonal antibody that targets B7-H3 that is overly expressed by tumor cells and its associated vasculature to mount antibody-dependent cell-mediated cytotoxicity [33]. We sought to identify the role of the immune system in giant cell lesions by studying the expression of components of HLA classes I and II and the immune modulator B7-H3. Our study provides, for the first time to our knowledge, evidence that giant cell lesions may escape immune surveillance secondary to defects in HLA class I expression and the inhibitory effect of B7-H3.

Limitations

This study has several limitations worth noting. First, we chose to combine central giant cell lesions of the maxillofacial skeleton with giant cell tumors of the axial and appendicular skeleton. This has been a controversial choice and there have been reports that suggest that these tumors are molecularly distinct. One study failed to show a histone mutation in central giant cell lesions (n = 9) [24] that was reported in 92% of giant cell tumors (49 of 53) [4]. However, these two studies were not methodologically equivalent, as one study separated the giant cells from stromal cells [4] and the other did not [24]. Furthermore, it is possible that other mutations in histones would be seen if whole genome sequencing had been performed. Furthermore, it is unclear what the functional importance of these histone mutations is as neither group assessed the corresponding protein expression. We believe that there are enough similarities between these groups that combining them makes clinical sense. This is particularly true for aggressive giant cell lesions.

It was reported that giant cell lesions in the axial and appendicular skeleton are more aggressive than those seen in the jaw [19]. However, in the jaw, there are many, very small nonaggressive lesions that, if lumped together with aggressive lesions, lead to an underestimation of the aggressiveness of the latter group. By excluding the indolent groups from the jaw, the two groups become comparable, which is what we have seen [38].

In a previous study, Resnick et al. [42] showed that if you look at aggressive giant cell lesions of the jaws and aggressive giant cell lesions of the axial and appendicular skeleton, they are similarly aggressive—except that a low percentage of the axial and appendicular lesions metastasize. In that study, the classification of Chuong et al. [13] was used which categorizes giant cell lesions in two categories as aggressive and nonaggressive. Using this model, clinical and radiologic features of the Enneking classification [18] were converted to a binary system (aggressive and nonaggressive). The theory was that when you categorize giant cell lesions based on their behavior (ie, aggressive versus nonaggressive), they are found to behave similarly in both groups. Resnick et al. [42] hypothesized that “maxillofacial and axial and appendicular giant cell lesions represent the same entity when assessed by comparable phenotypic parameters, comparing the clinical behavior of lesions from each group might reconcile some of differences noted in previous studies, and differences noted between maxillofacial and axial and appendicular lesions are a function of their anatomic locations, which results in the earlier diagnosis of maxillofacial lesions because of exposure of the face and the prevalence of screening dental examinations and radiographs.” They found that the lesions were similar when grouped in aggressive and nonaggressive categories.

In the study by Peacock et al. [39] on gene expression and tissue microarray analysis, the same argument was made by looking at gene expression from laser capture microdissection experiments confirmed with tissue microarray. Again, the lesions were found to be similar. In addition, in another study, Peacock et al. [38] asked pathologists to differentiate between aggressive and nonaggressive giant cell lesions based on histology and on location. The pathologists were not able to differentiate the phenotypic behavior of the tumors nor their site of origin based on histology alone [38]. In another study [46], the aggressive nature of giant cell lesions of the jaw was described in a 20-year followup study of the adjuvant interferon protocol. This study [46] corroborates the aggressive nature of giant cell lesions occurring in the jaw.

A second limitation is that our study proves an association but it does not definitely show causation. However, we feel that our findings are compelling when read in the context of the cited studies. Third, while we have a large series of tumors, there is inherently a selection bias that we cannot control given that not all of our patients were included in the study. The reasons why some patients were not included were many but they were unique to the practice of each surgeon whose patients were eligible for the study. It is does not seem likely that this selection bias inherently flaws the data we have collected.

Was an Immune Response Elicited by Giant Cell Lesions?

Muscolo and Ayerza [35] identified the potential role of a T cell immune response in giant cell lesions when they were able to show a graded activation of T lymphocytes when cocultured with cells taken from giant cell tumors. The Enneking system [18] was correlated with activation of the lymphocytes where more aggressive Stage III tumors did not elicit a robust T cell activation when cocultured with giant cell tumor cells, whereas Stage I tumors did elicit T cell activation. Their study provided evidence that behavior of giant cell lesions may be inversely related to T cell activation [35]. Our study provides a mechanistic framework, which helps to explain the lack of T cell activation seen in the study by Muscolo and Ayerza [35]. Our study showed that a T cell immune response is present in giant cell lesions as lymphocyte infiltrates were identified in all tumors. Low CD8+ T cell infiltration was associated with aggressive tumors, further suggesting that the immune system plays an active role in determining disease status.

Did Clinically Relevant HLA Defects Exist in Giant Cell Lesions?

Lymphocytes appear to impose selective pressure on tumor cell populations, as HLA class I antigen defects have been found in more than 50% of the mononuclear stromal cells analyzed. These defects reflect the outgrowth of tumor cells which escape from recognition and destruction by the host immune system because of abnormalities in HLA class I antigens. The latter may cause a defective expression of HLA class I antigen-tumor antigen derived peptide complexes which mediate the interaction of tumor cells with cognate cytotoxic T lymphocytes. While defects in β2-microglobulin were detected in almost 90% of mononuclear stromal cells, they did not show any correlation with tumor aggressiveness when stratified by HLA class I expression. Similarly, there was no association between β2-microglobulin expression and CD8+ infiltration. This suggests that a defect in β-2 microglobulin light chain is unlikely to play a role in tumor immune evasion in giant cell lesions. HLA class II antigens have classically been thought to be expressed on antigen processing cells with limited expression in other normal cells, although there is considerable evidence to suggest that HLA class II antigens are expressed on many normal cells other than those of the hematopoietic lineage [9]. In addition, HLA class II antigens are thought to be induced in some cancer cells. These “nonprofessional” antigen-presenting cells can present antigens to CD4+ T lymphocytes. However, there is evidence that this antigen-presenting cell and CD4 interaction leads to immunosuppression rather than activation of the immune response [8]. Our data show that HLA class II antigens are present in the majority of mononuclear stromal cells and that HLA class II expression is associated with local tumor invasion. The same was not true with HLA class II expression in multinucleated giant cells where 97% of the samples were negative for HLA class II expression. This is interesting in a historical context where osteoclasts were once thought to be the giant cells in giant cell tumors [14]. Mature osteoclasts express HLA class II antigens yet the giant cells in our giant cell lesions did not express HLA class II antigens. HLA class II antigen expression was not associated with HLA class I, or with either CD4+ or CD8+ T cell infiltration.

Was B7-H3 a Clinically Relevant Immune Modulator?

Defects in HLA class I antigen expression are not the only escape mechanism used by giant cell lesion cells to avoid destruction by host immune cells. High expression of B7-H3 expression, which we found on mononuclear stromal cells in 37% of the giant cell lesions, is associated with local tumor invasion. This is similar to what has been shown in glioblastoma where B7-H3 expression by tumor cells was associated with greater local tumor invasiveness [31]. Similarly, a previous study showed silencing of B7-H3 reduced migration and invasion of melanoma and breast cancer cell lines [36]. B7-H3 has been associated with immune suppression and worse prognosis in multiple cancers including lung, glioblastoma, renal cell carcinoma, prostate, and pancreatic ductal adenocarcinoma cancer cells [11, 15, 31, 44, 58]. B7-H3 exerts an inhibitory immune effect by preventing activation of CD8+ T lymphocytes [30]. Our data indicate that high B7-H3 expression in the presence of high CD8+ tumor infiltrating cells is associated with local tumor invasiveness suggesting an inhibitory immune function for B7-H3 in giant cell lesions.

Conclusions

Locally aggressive giant cell lesions are associated with lower HLA class I expression, lower CD8+T cell infiltration, and higher expression of the immune modulator B7-H3 when compared with nonaggressive lesions. Giant cell lesions may become aggressive because they evade immune surveillance. The data presented here provide a mechanistic view into how some giant cell lesions may progress and become infiltrative as they overcome the host’s immune system. Exploring antigen processing machinery could shed some light on how tumors can regulate their HLA class I. It also offers potential sites one could target with immune therapy such as blockage of B7-H3.