Background
Cherubism is a rare, autosomal dominant, benign fibro-osseous disease of the jaws with incomplete penetrance and variable expressivity [
1]. The first signs appear between 2 and 5 years of age, progress until puberty, and then usually regress in adulthood [
2]. Patients present a painless bilateral and multilocular granuloma of the jaws, affecting facial appearance and teeth eruption. Cherubism may also affect facial nerves, orbits and breathing [
3,
4]. Histologically, cherubism granulomas are composed of a dense fibrous connective stroma with fibroblasts and randomly distributed multinucleated giant cells (MGC), considered to be osteoclast-like cells [
5,
6].
At the molecular level, cherubism is caused by mutation of the
SH3BP2 gene (SH3 domain-binding protein 2), located on chromosome 4p16.3 [
7]. SH3BP2 is an adaptor protein involved in lymphocyte activation, osteoclast differentiation and bone remodeling, through pathways involving Src, Syk and Vav-family protein kinases, and NFATc1 (nuclear factor of activated T cell 1) [
8‐
13]. Most of the autosomal dominant mutations identified in cherubism lead to a single amino-acid change [
7]. Recent genetic and biochemical studies have provided critical insights into the pathogenic mechanism of cherubism thanks to the creation of knock-in (KI) mouse models with the most common
SH3BP2 mutations [
14]. However, unlike human
SH3BP2 heterozygotes, heterozygous mice do not exhibit any cherubism phenotype, and homozygous mutants develop severe bone loss due to osteoclast hyperactivity. Despite this important difference in genetic expression,
Sh3bp2 KI mice are considered a cherubism model [
14]. According to Ueki’s mouse model, cherubism is associated with a high level of TNF-α (Tumor Necrosis Factor α) that is responsible for maintaining the phenotype: hyperactive macrophages secrete a high level of TNF-α that drives systemic inflammation, stimulates secretion of RANK-L (Receptor Activator of Nuclear factor Κ B Ligand) and M-CSF (Macrophage Colony Stimulating Factor) (osteoclastogenesis-associated proteins) by stromal cells, and ultimately results in bone loss [
14]. In vitro, upon stimulation by RANK-L,
Sh3bp2 KI myeloid progenitor cells induce the activation of the NFATc1 signaling pathway, leading to hyperactive osteoclasts [
14,
15]. In vivo,
Sh3bp2 KI mice develop systemic inflammation as a result of systemic infiltration by macrophages into tissues, as well as bone loss [
14], defining cherubism as an auto-inflammatory bone disease [
16‐
18].
The main objective of the present study was to determine if this auto-inflammatory bone disease paradigm could also be applied to human cherubism. To do so, we systematically examined the types of cells present in granulomas from 7 cherubism patients to look for evidence of chronic inflammation. We then characterized the osteoclastic features of the MGC both in vivo and in vitro. We also explored the potential role of TNF-α in the pathogenesis of human cherubism, and searched for potential biomarkers of the disease. Thus, we showed that in human cherubism, osteoclasts are the major myeloid cell type embedded within a fibrous stroma. The characteristics of these CD68-positive cells (macrophage vs. osteoclast) may predict the aggressiveness of the disease. Moreover, we demonstrated that first human cherubism granuloma is heterogeneous according to the patient and second the mechanism underlying human cherubism appeared to be different from that of mice.
Methods
Patients
This study included 7 patients (5 children and 2 adults) treated and followed in the maxillo-facial surgery department of Necker Hospital, Paris, through the MAFACE reference center for rare facial malformations. The 5 children were already part of our previously published cohort [
19]. All patients gave their written informed consent for this study and for genetic analysis. Age, sex, age at diagnosis, age at first surgical treatment, radiologic extent and evolution of the lesions were recorded for each patient at the biopsy time. Medical treatment and past history were noted. Mutations in the gene encoding the binding protein SH3BP2 on chromosome 4p16.3 were sought for each patient. Direct Sanger sequencing of exons 2 to 13 of the
SH3BP2 gene was performed for 7 patients. All patients underwent intraosseous cherubism granuloma curettage.
The cherubism cases were classified according to patient age at the time of surgery (children were classified as group 1, adults as group 2) and sub-classified according to their aggressiveness based on the usual radiologic classification [
20] and disease evolution after surgery [
19]. For the child cases, radiologic grade I with favorable evolution after surgery was classified as group 1-A (low degree of aggressiveness: radiologic grade I with a favorable evolution); 1-B (moderate degree of aggressiveness: radiologic grade II-IV with a favorable evolution); 1-C (high degree of aggressiveness: radiologic grade V-VI and/or an unfavorable evolution, recurrence or extension); the adult patients were sub-classified as 2-A (remodeling bone) and 2-B (acute exacerbation) (Table
1) [
20].
Table 1
Patient classification
Child Patients | 1-A | Low degree of aggressiveness: radiologic grade I with a favorable evolution after surgery | 1-A | 16 | M | I.2 | c.1244 G > A | Favorable |
1-B | Moderate degree of aggressiveness: radiologic grade II-IV with a favorable evolution after surgery | 1-B1 | 9 | F | II.1 | c.1244 G > A | Favorable |
1-B2 | 8 | M | II.1 | c.1244 G > A | Favorable |
1-C | High degree of aggressiveness: radiologic grade I-VI with an unfavorable evolution after surgery | 1-C1 | 8 | M | V | c.1244 G > A | Unfavorable (recurrence and extension) |
1-C2 | 7 | M | VI | c.1253 C > G | Unfavorable (recurrence and extension) |
Adult Patients | 2-A | Remodeling bone | 2-A | 19 | F | N/A | c.1244 G > A | N/A |
2-B | Acute exacerbation | 2-B | 45 | F | VI.3 | c.1244 G > A | N/A |
Sample collection
Granuloma samples were obtained at the time of surgery and divided into 2 parts, one for pathological examination and the second for the biochemical and biomolecular studies outlined below. Tissues for pathological examination were fixed in 10% formalin and embedded in paraffin. Four samples of normal alveolar bone, collected during third molar extraction, were included after written informed consent as controls for biomolecular analysis.
HES staining
4-μm thick sections were cut from each paraffin block. Section staining with hematoxylin/eosin/safranin (HES) was automated by using a Leica Autostainer (Nussloch, Germany): after deparaffinization (in successive baths of xylene and alcohol), staining was performed by successive baths in hematoxylin GILL2 (Thermo Shandon, Pittsburgh, PA, USA), 1% eosin (Ral Diagnostics, Martillac, France), 0.12% safranin in 1% hydrochloric acid solution (Ral Diagnostics, Martillac, France) and 95% ethanol. The sections were then mounted in synthetic resin (WVR, Radnor, PA, USA). Two pathologists blinded to clinical history scored the slides, according to the previous description of cherubism [
5,
21,
22]. For each specimen, we evaluated presence of intracytoplasmic vacuoles in MGC, presence of collagen (semi quantitative evaluation: 0, +, ++, +++), cells subtype (fibroblast, round cells, MGCs) semi-quantitative evaluation (0, +, ++, +++), perivascular hyalinosis. From 5 randomly selected areas at 200 high-power field (HPF), we evaluated the number of MGC and the number of nuclei per MGC.
Immunohistochemistry
Vimentin, AE1-AE3, CD68, CD4, CD8, CD3, CD5, CD20 automated immunohistochemistry (Additional file 1)
Immunohistochemical staining was performed using a Ventana Benchmark® XT automated slide preparation system (XTUltraviewDABv3, Ventana, Tucson, AZ, USA). 4-μm thick sections were incubated with primary antibody (Table
2). After washing in phosphate-buffered saline (PBS), sections were incubated with streptavidin/horseradish peroxidase (Biolegend, San Diego, CA, USA). Sections were immersed in 3% H
2O
2 to quench endogenous peroxidase activity. Staining was visualized in brown color using a diaminobenzidine tetrahydrochloride chromogen substrate (DAB, SK-4105 Vector Laboratories, Burlingame, CA, USA). Sections were counterstained with hematoxylin. Slides were scored by two pathologists blinded to clinical history and the primary antibody used. The number of positive cells was evaluated for each assay from five randomly selected areas (200 HPF), except for vimentin. Vimentin expression was semi-quantitatively evaluated (0 = vimentin negative; + few positive cells; ++ less than the half positive cells; +++ more than half positive cells).
Table 2
RANK/RANKL/OPG triad and NFATc1 expression by cherubism-granuloma cells
1-A | – | – | – | – | – | – | – | – |
1-B1 | + | + | – | – | – | – | + | – |
1-B2 | – | – | + | – | – | – | – | – |
1-C1 | + | + | + | – | + | – | + | + |
1-C2 | + | + | + | – | + | – | + | + |
2-A | – | – | – | NA | + | – | – | – |
2-B | – | – | + | – | – | – | + | + |
RANKL, RANK (receptor of activated nuclear factor Κ B), OPG (osteoprotegerin), TNFR1(tumor necrosis factor receptor 1)., IL6 (interleukin), IL17, NFATc1 manual immunohistochemistry (Additional file 2)
For immunohistochemistry staining, 4-μm thick sections were first deparaffinized with xylene for 30 min, post-fixed with 90% ethanol for 10 min and then washed in distilled water for 5 min. For antigen retrieval, the sections were incubated in a 78 °C water bath for 30 min in pH 6 or pH 9 buffer, cooled at room temperature for 20 min, washed in PBS for 10 min and finally incubated with reagents from an avidin/biotin kit (Vector Laboratories, Perterborough, UK). Endogenous perodixase activity was blocked by incubating the sections with 3% H2O2 followed by a wash in PBS. Sections were blocked in a 5% normal human serum for 30 min before incubation with the primary antibodies or isotype control for 60 min at room temperature. After a PBS wash, the sections were then incubated with the secondary antibody for 30 min, then with streptavidin/horseradish peroxidase for 30 min (Biolegend, San Diego, CA, USA). Staining was visualized as a brown color by using a diaminobenzidine tetrahydrochloride chromogen substrate (Vector Laboratories, Perterborough, UK). Slides were scored by two pathologists blinded to clinical history and the primary antibody used. The signal for each antibody was evaluated for MGCs and stromal cells. For NFATc1, the numbers of nuclear and cytoplasmic positive cells were evaluated.
TRAP (tartrate resistant acid phosphatase) activity assay
A TRAP activity assay was performed on paraffin-embedded tissue sections and cell cultures from granulomas on Lab-Tek™ chamber slides (Dutscher). Paraffin sections were deparaffinized and rehydrated, stained in pH 5.2 acetate buffer containing 2.5 mM Naphthol AS-TR phosphate, 0.36 M N–N dimethylformamide, 0.1 M sodium tartrate and 4 mM Fast Red TR Salt (Sigma-Aldrich). Hematoxylin was used for nuclear staining. TRAP-positive giant multinucleated (> 3 nuclei) cells were considered as osteoclasts. The number of osteoclasts (MGC > 3 nuclei) and number of nuclei per osteoclast were evaluated from five randomly selected areas (200 HPF).
Total RNA was extracted from the granuloma samples, bone controls or cell cultures using Trizol Reagent (Life technology, Saint Aubin, France) according to the manufacturer’s instructions. The total RNA yield (ng) was determined fluorometrically using a Qubit fluorometer (Life Technologies, Saint Aubin, France). Total RNA (1 μg) was reverse transcribed using SuperScript II™ reverse transcriptase (Life Technologies, Saint Aubin, France) according to the manufacturer’s instructions. Real-time quantitative PCR was carried out using SYBR-green master mix (Life Technologies, Saint Aubin, France) in a BCR Bio-Rad Opticon thermocycler (Bio-Rad, Marne La Coquette, France). PCR conditions were: 98 °C for 30 s followed by 40 cycles of 95 °C for 10 s, 60 °C for 20 s and 72 °C for 20 s. Primer sequences of all the analyzed genes are shown in (Additional file
3). Cq was transformed into quantity values using the formula (1 + Efficiency)
-Cq as previously described [
23].
GAPDH,
SDHA,
TBP and
HPRT were used as reference genes.
Granuloma cell culture
Samples from only 5 cherubism cases yielded cell cultures. Tumor tissues were washed in PBS. Fragments of the tumors were incubated at 37 °C for 60 min in a hyaluronidase and collagenase solution mix (1 mg/ml for each enzyme) in PBS. The number of cells was counted manually using a Mallassez counting chamber. Cells were cultured either on a dentin slice (ids, Paris, France) (at a density of 0.5 × 106 cells/ml) for the resorption assay, Lab-Tek II chamber slides 1 well (ThermoScientific, Rochester, NY, USA) (at 0,6 × 106 cells/cm2) for the ELISA supernatant study, or Lab-Tek II chamber slides 8 wells (at 0,6 × 106 cells/cm2) for the TRAP activity assay with standard medium. After 12 h of incubation at 37 °C in a 5% CO2 humidified atmosphere, half of the cell culture was incubated in standard medium and the other half in osteoclastogenic medium. Standard medium contained minimal essential medium (αMEM, without red phenol) and 1% Penicillin-streptomycin, 1% L-glutamine (all from Life Technologies, Saint Aubin, France) and 10% fetal calf serum (Hyclone, South Logan, UT, USA). For the Lab-Tek cultures, osteoclastogenic medium contained standard medium supplemented with RANKL at 30 ng/ml and M-CSF at 25 ng/ml (Peprotech, Neuilly-sur-Seine, France). For the dentin slices cultures, osteoclastogenic medium contained standard medium supplemented with RANKL at 60 ng/ml and M-CSF at 25 ng/ml (Peprotech, Neuilly-sur-Seine, France). Half of the cell culture was maintained for 3 days, and the other half was maintained for 7 days. In the 7-day cultures, the culture medium was changed at day 3. Supernatants were collected at day 3 and day 7 for ELISA. Cultures in Lab-Tek 8 chambers were fixed in 4% paraformaldehyde for TRAP activity measurement, and cultures in Lab-Tek 1 chambers were treated with Trizol Reagent for RNA extraction. Cultures on dentin slices were washed in distilled water and sonicated to remove cells. The slides were then stained with 0.5% toluidine blue to reveal lacunar resorption areas by light microscopy.
Osteoclast in vitro differentiation from peripheral blood mononuclear cells (PBMC)
Whole blood (15 ml) was obtained from the French Transfusion Establishment (EFS). Blood was diluted with 1X PBS, layered on 12 ml of Ficoll (Euromedex, Souffersheim, France), and then centrifuged (× 1500 g, 10 min, room temperature, with the brake off). The PBMC layer was collected and washed in PBS, and then counted in a Malassez counting chamber. The cells were plated in Lab-Tek 8 chambers (at 0,6 × 106 cells/cm2), in standard medium and osteoclastogenic medium for 14 days.
ELISA
Levels of RANK-L, OPG, M-CSF, TNF-α and IL-6 in culture supernatants were determined by commercially available specific ELISAs, according to the manufacturer’s protocols (R&D Systems, Minneapolis, MN, USA). The lower and higher detection limits for each cytokine are shown in Additional file
1. It is important to note that the system only detects free, unbound RANK-L. Absorption was determined with an ELISA reader at 450 nm (Additional file
4).
Statistical analysis
Results are expressed as mean ± 2 SEM. Statistical comparisons were made using ANOVA, with p < 0.05 being considered significant.
Discussion
Cherubism is a rare disease described as an auto-inflammatory disease from the mouse model point of view. However human cherubism, and specially cherubism intraosseous lesion is poorly described in the literature. In the present study, we extensively (in situ and in vitro) explored cherubism intra-osseous lesions in order to determine if human cherubism could be considered as an auto-inflammatory bone disease. For this purpose, we explored cells in presence, and characterized the osteoclastic characteristics of MGC. Moreover, we examine the expression of inflammatory cytokines by granuloma cell cultures. Thus, we showed that human cherubism granuloma is composed mainly of osteoclasts or macrophages and relatively few immune cells, mainly CD68-positive cells, within a fibroblastic environment. The characteristics of these CD68-positive cells (macrophage vs. osteoclast) may predict the aggressiveness of the disease. Moreover, we demonstrated that the mechanism underlying human cherubism is different from that of mice; since, contrary to mice, it does not seem that TNF-α underlies the physiological mechanism of human cherubism, and that RANK-RANKL-OPG triad is not playing their usual role.
This study is the first large cellular and molecular analysis carried out on human cherubism granuloma samples. Our results show that MGC are derived from a monocyte lineage within a fibrous stroma with few inflammatory cells. The myeloid cells present in cherubism granuloma are mainly macrophages and osteoclasts, and the osteoclast phenotype may determine the aggressiveness of the disease. In addition, this study demonstrated that these osteoclast-like MGC are truly functional osteoclasts, able to resorb bone. Moreover, we showed a low expression of TNF-α pathway protein both in granuloma and in culture, while IL-6 is more widely expressed. In addition, it seems that the RANK/RANKL/OPG triad is disturbed in cherubism.
In this study, the seven unrelated patients presented an array of phenotypes, including both adult and childhood cases and non-aggressive to highly aggressive granulomas. Moreover, the study included a wide variety of analyses not previously performed in any study of human cherubism: ex vivo and in vitro studies, and pathological, immunohistological, biomolecular and cellular analyses.
The immunohistochemical analysis permitted us to characterize the types of cells present within the cherubism granuloma. Fibroblast cells and monocyte lineage cells are the most prevalent cells, with some rare lymphoid cells, principally CD8+ cells. These results support the definition of cherubism as an auto-inflammatory disease [
17]. Auto-inflammatory bone disease is characterized by a chronic noninfectious inflammation which induces bone resorption, and results from aberrant activation of the innate immune system [
16,
17]. Lymphoid cells and acquired molecules of the adaptive immune system do not play a role in this branch of auto-inflammatory bone diseases [
26]. In cherubism mice, Ueki et al. previously demonstrated that cherubism is a myeloid (lymphoid independent) inflammatory bone-resorbing disease [
14]. In our series, we detected four different cell types of myeloid cells: immature macrophages (CD68
+, TRAP
− mononucleated cells), multinucleated macrophages (CD68
+, TRAP
−, nuclear NFATc1
− multinucleated cells), osteoclast precursors (CD68
+, TRAP
+ mononucleated cells), and osteoclasts (CD68
+, TRAP
+, nuclear NFATc1
+ multinucleated cells). First, these results confirmed that MGC are derived from a monocyte lineage, as previously described [
21]. In this study, the myeloid cell types present in granulomas are likely to predict the disease aggressiveness: i.e., osteoclast fate in aggressive cherubism and macrophage fate in non-aggressive cherubism, as we previously suggested [
19]. Moreover, our in vitro analysis showed that the osteoclasts are functional and able to resorb bone, as was previously demonstrated by Southgate et al. [
22].
Here, we discovered that RANKL has an unexpected expression in cherubism. We showed that the cherubism granulomas contained very few RANKL-positive cells (by immunohistochemistry), and these RANKL-positive cells were localized in regions poor in MGC. Moreover,
RANKL mRNA expression was low in the granulomas; and in cultured cells, RANKL protein expression was also low. Furthermore, it seems that a negative feedback process is activated, with increased secretion of OPG. By immunohistochemistry, we showed numerous OPG-positive cells around MGC, in aggressive cherubism (1C1, 1C2). Liu et al. [
21] showed similar results with extensive expression of OPG (by in situ hybridization). However, these results contrast with observations in mice. In
Sh3bp2 KI models, the observed osteopenia is due to an osteoblast functional defect and a reduction in OPG synthesis [
27,
28]. In human cherubism, we hypothesize that osteoclast/MGC differentiation is induced through a RANK-L-independent pathway (because of a potential perpetual RANK activation), with inefficient regulation by OPG or a RANK-L bypass.
In the present study, one adult patient (2-A) in the remodeling phase was included. Interestingly, the patient’s granuloma displayed characteristics of osteogenesis: a RANK-L/OPG ratio less than 1, and high expression of ALP and OC (osteoblast markers).
Interestingly, cherubism cells differentiate into functional osteoclasts upon exogenous stimulation by RANK-L (Fig.
5c, d), but are also capable of differentiating without secreted or recombinant RANK-L stimulation (Fig.
5a, b) (Additional file
7). As suggested by Ueki et al. [
14], mutated myeloid cells are hypersensitive to RANK-L and differentiate into osteoclasts, but may also differentiate through a RANK-L-independent pathway. Supporting these results, Mukai et al. [
29] showed that
Sh3bp2KI/+ bone marrow-derived M-CSF-dependent macrophages are highly sensitive to TNF-α and can differentiate into osteoclasts independently of RANK-L. Moreover, Wang et al. [
30] suggested that cherubism bone phenotype may arise from a direct cross talk between osteoclast and osteoblast. Many mechanisms may explain the development of cherubism: other cytokines may be involved and may favor osteoclastogenesis in human cherubism independently of RANK-L; the RANK pathway may be permanently stimulated; or MGC may be stimulated through an autocrine pathway. All these theories may explain a supposed more accessory role of TNF-α in human cherubism.
These results suggest that cherubism cells produce a cytokine(s) that stimulates osteoclast differentiation (reflected in the increased number of osteoclasts from day 3 to day 7) and fusion (increased osteoclast size and nuclei from day 3 to day 7) in standard medium. Studies of mouse models of cherubism [
14,
29,
31] suggest that TNF-α plays a key role in the disorder by enhancing osteoclast and macrophage differentiation. In humans, previously published immunohistochemical analyses showed TNF-α expression by MGC [
29,
32]. In our study, immunohistochemical analysis did not show expression of TNF-R1 or R2 proteins, and biomolecular analysis showed
TNFR1,
TNFR2 and
TNF-α expression comparable to that in normal alveolar bone. Moreover, in vitro, the level of TNF-α in the culture supernatant was low. This result is in line with the dubious role of TNF-α in human cherubism. Indeed, Amaral et al. [
33] showed that the relative expression of
TNF-α transcript was low in cherubism, and others reported that anti-TNF-α therapy failed to improve cherubism lesions [
32,
34]. Thus, osteoclast and macrophage differentiation must be stimulated by other cytokines. From our analysis of culture supernatant, IL-6 may be a likely candidate. IL-6 is widely recognized as a stimulator of bone loss in the context of inflammation, through pathways both dependent and independent of RANK-L [
35‐
37].
This study is a first step to understand the molecular pathways leading to human cherubism. Our results suggest that osteoclast activation is mediated by translocation of NFATc1 into the nucleus, leading to the transcription of osteoclastogenic effectors such as TRAP. The role of NFATc1 had already been identified in the mouse model. However, in the
Sh3bp2 KI models, NFATc1 is implicated in bone loss but not in inflammation [
15]. In human cherubism, three previous studies demonstrated the involvement of NFATc1 [
4,
33,
38], showing an increase in NFATc1 transcription in cherubism granuloma and nuclear expression of NFATc1 in MGC. Moreover, we recently showed that nuclear NFATc1 expression is correlated with the prognosis for disease aggressiveness [
19]. In the present study, we further described the osteoclast phenotype of MGC, the increased size of MGC (due to osteoclast fusion), and their self-maintenance. All these characteristics may be explained by the activation of NFATc1. Indeed, NFATc1 promotes its own amplification, the expression of osteoclastogenesis genes (such as TRAP), and osteoclast fusion [
39].
Many challenges remain. First, because of the rarity of cherubism, a large prospective study is difficult to undertake. Second, the granulomas are highly variable not only between patients, but also within each granuloma itself. Thus, it is difficult to draw firm conclusions based on a single biopsy, and serial coring may be necessary. Finally, since MGC are osteoclasts (terminally differentiated cells), secondary cultures are impossible; indeed, serial analyses in cherubism cultures are difficult. Addressing these challenges will require investigation of a much larger series of patients, probably international in scope owing to the rarity of the disease, and creation of a human in vitro model.