1. Introduction
Although rare, osteosarcoma is the most common malignancy of bone predominantly affecting children and young adults [
1,
2]. It often occurs in the metaphyses of long bones such as the femur or tibia, and is pathologically characterized by cells with high-grade atypia and aberrant osteoid formation [
1]. Jaw osteosarcoma accounts for only 6% of all osteosarcomas and develops mainly in the mandible. It is diagnosed approximately two decades later than long bone osteosarcoma and entails a lower risk of lung metastases reported to be 20–25% versus 44–49% for long bone osteosarcomas [
3].
According to the 2020 WHO classification, osteosarcoma can be classified into high-grade conventional osteosarcomas, the most common histologic subtype (75% of all cases), periosteal osteosarcoma of intermediate-grade, and low-grade osteosarcomas (paraosteal and intramedullary osteosarcomas). Microscopically, conventional osteosarcoma is highly heterogeneous with cells that produce varying amounts of osteoid/chondroid matrix [
2].
The high complexity of the osteosarcoma genome has not allowed for the identification of key molecular therapeutic targets so far [
4,
5,
6]. As a consequence, osteosarcoma treatment still involves a combination of surgery with either neoadjuvant or adjuvant chemotherapy. The 5-year survival rate has not evolved over the past four decades, and does not exceed 70% with combined surgery and chemotherapy. It falls to 25% in patients with metastatic or relapsed osteosarcoma [
7,
8,
9,
10,
11].
There are a number of reported mechanisms for treatment resistance in osteosarcoma including genomic aberrations, noncoding-RNA post-transcriptional regulations or tumor microenvironmental factors such as hypoxia [
12]. Hypoxia is a feature of many solid human tumors as the aberrant cell proliferation rate is associated with a disequilibrium between oxygen supply and consumption [
13]. The many effects of hypoxia on cancer biology include not only promotion of progression and metastasis [
14] but also resistance to radiotherapy or chemotherapy [
15]. The transcription factor, hypoxia-inducible factor 1α (HIF-1α), has been reported as the master driver of adaptation to hypoxia. A wealth of reports based on immunohistochemical studies of human tumor sections indicate that HIF-1α is overexpressed in the majority of human cancers and these elevated levels correlate with cancer-related death [
16]. A number of studies have analyzed the association of hypoxic markers (HIF-1α, GLUT-1 or Glucose Transporter-1, and VEGF or Vascular Endothelial Growth Factor) with prognosis and clinicopathological characteristics of osteosarcoma. Recent meta-analyses suggest that hypoxia would be associated with lower survival rate, higher microvessel density, metastasis, higher pathologic grade, tumor stage and poor response to chemotherapy [
17,
18,
19].
Sphingosine 1-phosphate (S1P) is a pleiotropic phospholipid that regulates proliferation, migration, inflammation or angiogenesis [
20]. S1P is produced by phosphorylation of sphingosine by two isoforms of sphingosine kinases (SphK1 and SphK2), and irreversibly degraded into hexadecenal and ethanolamine phosphate by S1P lyase (SPL), an enzyme that has been shown to be downregulated in cancer [
21]. Sphingosine kinase 1 (SphK1) is commonly overexpressed in cancer cells and correlates with poor patient prognosis in a number of tumor types [
22,
23]. As SphK1/S1P signaling contributes to malignant progression by controlling proliferation and metastatic potential of cancer cells, it represents a potential target for anticancer therapy [
24]. We previously identified SphK1/S1P signaling as a new modulator of HIF-1α and HIF-2α activities under hypoxia in a wide array of cancer cell models (prostate, glioma, breast, lung and renal cell carcinoma) both in vitro and in vivo [
25,
26,
27]. FTY720 or fingolimod (
Figure S1) is an analogue of sphingosine that was approved by the US Food and Drug Administration (FDA) in 2010 for the treatment of relapsing–remitting multiple sclerosis after two phase III clinical studies establishing its efficacy and global safety [
28,
29]. It is currently under trial for the treatment of breast carcinoma, glioblastoma and anaplastic astrocytoma (NCT03941743 and NCT02490930). It can be phosphorylated in vivo to form FTY720-phosphate (FTY720-P), a mimetic of S1P, interacting with S1P receptors (excepted S1P
2 subtype) but preferentially inducing internalization and degradation of S1P
1, which has been associated with prevention of lymphocytes to egress from lymph nodes, reducing the amount of lymphocytes in peripheral blood [
30]. Some of its actions are also attributed to its unphosphorylated form, notably inhibition of SphK1 activity associated with proteasomal degradation of SphK1, as suggested by several independent studies [
31,
32,
33]. Antitumoral properties of FTY720 have been evidenced in a number of cancer animal models, including inhibition of angiogenesis and tumor vascularization [
32,
34,
35,
36,
37]. In that respect, we previously reported that FTY720 decreases both HIF-1α and HIF-2α expression and subsequent expression of GLUT-1 and VEGF in a number of cellular and animal models of renal cell carcinoma [
38]. Collectively, these data suggest that targeting of SphK1/S1P signaling represents a strategy that could potentially be exploited in therapeutic approaches to decrease hypoxia in cancer [
39,
40].
The objectives of this work were to assess (i) the role of the SphK1/S1P signaling in various cellular models (U-2OS, SaOS-2, MG-63 and 143B) of osteosarcoma cell lines under hypoxia, (ii) the SphK1 enzymatic activity in human osteosarcoma samples versus nontumoral bones, and (iii) the expression of HIF-1α target gene GLUT-1, SphK1 and S1P1 in 130 cases of osteosarcomas (long and flat bone samples) as the link between SphK1/S1P signaling and hypoxia has never been investigated in human tissues.
2. Materials and Methods
2.1. Patient and Tumor Characteristics
Tissue microarrays (TMAs) were prepared from the diagnostic biopsies of 130 patients coming from Toulouse (AGB), Lille (SA), Nantes (LG), Marseille (CB), Paris (FL), and Tours (GDP). TMAs (triplicate sampling of 1 mm) were established at the Institut Universitaire du Cancer biobank of Toulouse (AGB). For all cases, the TMA cores have been selected in the most cellular areas based on the diagnostic biopsy slide stained with hematoxylin–eosin (H&E). All OS samples were reviewed and reclassified by the accredited pathologists (SA, CB, FL, GDP, AGB) of the GFPO (French Group of Bone Pathologists), according to the WHO 2020 classification. The TMAs were then stored at the certified NF 96–900 cancer biobank of Toulouse (BB-0033-00014) where the immunohistochemistry study was conducted (GLUT1, SphK1, S1P
1). In compliance with French law, the biobank cancer collection was declared to the Ministry of High Education and Research (DC-2020-4074) and a transfer agreement was obtained (AC-2020-4031) by ethical committees. All patient records and information were anonymized and deidentified before analysis. Informed consent was obtained from all patients and the use of the biological specimens was approved by the local institutional review board. Patient and tumor characteristics and response to treatment are described in
Table 1 for the 130 patients included in the tissue microarray study. We studied 59 flat-bone osteosarcomas (mandibular and other locations: pelvis, ribs) and 71 long-bone osteosarcomas. To complete our exploration, we used 12 other surgical biopsies coming from the OS Toulouse collection and 10 normal bone samples also stored at the certified NF 96–900 cancer biobank of Toulouse (BB-0033-00014). These samples were used to quantify the SphK1 kinase enzymatic activity.
2.2. Immunohistochemistry
SphK1 immunohistochemistry was performed manually on deparaffinized FFPE tissue sections heat-pretreated (95°) using Envision TRS Low-pH buffer (pH 6, GV80511-2, Agilent Technologies, CA, USA) for 30 min. Sections were subsequently incubated with 3% hydrogen peroxide for 20 min to block endogenous peroxidase. The primary polyclonal SphK1 antibody [
41], used at 1/300 (in Envision Flex diluent (K800621-2), Agilent Technologies), was applied for 1 h at room temperature and visualized using the Envision Flex DAB detection kit (Agilent Technologies). The tissue slides were counterstained using hematoxylin (Agilent Technologies) for 20 min at room temperature. S1P
1 immunohistochemical stains were automated using the Discovery ULTRA (Roche, Ventana Medical Systems, Innovation Park Drive, Tucson, AZ, USA). After dewaxing, tissue slides were heat-pretreated using a CC1 (pH8) buffer (05424569001, Roche) for 32 min at 100 °C. The slides were blocked for endogenous peroxidase activity using the CM inhibitor (32 min at 37 °C) (Roche). The primary S1P
1 antibody [
42], used at 1/100 in Ventana Diluent (Roche) was incubated for 20 min at 36 °C. The target was then linked using the OmniMap anti-rabbit (05269679001, Roche, Tucson, AZ, USA) HRP-conjugated secondary antibody and visualized using the ChromoMap DAB detection kit (05266645001, Roche). The tissue slides were counterstained using hematoxylin (05277965001, Roche) for 8 min followed by postcoloration using the Bluing reagent for 4 min at room temperature (05266769001, Roche). Finally, GLUT1 immunohistochemistry was performed using the Benchmark ULTRA (Roche, Ventana Medical Systems, Tucson, AZ, USA). After dewaxing, tissue slides were heat-pretreated using a CC1 (pH 8) buffer (05424569001, Roche) for 64 min at 98 °C. The slides were blocked for endogenous peroxidase activity and incubated with ready-to-use primary anti-GLUT1 polyclonal antibody (06419178001, Roche) for 32 min. The target was then visualized using the OptiView DAB detection kit (06396500001, Roche). The tissue slides were counterstained using hematoxylin (05277965001, Roche) for 8 min followed by postcoloration using Bluing reagent for 4 min at room temperature (05266769001, Roche). All slides were then dehydrated (ethanol and xylene) and mounted using xylene-based mounting. Isotype negative-control immunoglobulin stains were included for all markers as quality control. The staining for each target was evaluated using two methods: the percentage of stained cells, without considering the intensity of the staining, altered by the preanalytical treatment (decalcification) of histological sections and the IRS Score (semiquantitative immunoreactive score). This latter scoring system multiplies staining intensity (4 grades) by the percentage of cells positive for the marker (5 grades) and results in a scoring of 0 to 12 [
43,
44]. Immunoreactivity was considered positive if detected in >1% of cells per core of 1mm, irrespective of staining intensity. A double-blind examination by two pathologists, one an expert in bone sarcoma, was performed for the interpretation of the immunohistochemistry results.
2.3. Chemicals and Reagents
Culture medium was obtained from Life Technologies (Saint Aubin, France). Serum was from Perbio (Brebières, France). [γ-32P]ATP was from Perkin-Elmer (Courtaboeuf, France). Silica gel 60 TLC plates were from VWR (Fontenay sous Bois, France). FTY720 (Fingolimod) was from Enzo Life Science (Villeurbanne, France). SKI II (CAS Number 312636-16-1) SphK1 inhibitor, doxorubicin and all other reagents were from Sigma-Aldrich (Saint-Quentin Fallavier, France).
2.4. Cell Lines
Human U-2OS, MG-63 and SaOS-2 osteosarcoma cell lines were obtained from ATCC (Molsheim, France). Human 143B cell line was kindly supplied by Dr F. Lecanda (CIMA, Pamplona, Spain). Cells were cultured in RPMI containing 10% fetal bovine serum at 37 °C in 5% CO2 humidified incubators. Cell lines were routinely verified by the following tests: morphology examination, growth analysis and mycoplasma detection (MycoAlert TM, Lonza, Basel, Switzerland). All experiments were started with low-passaged cells (<25 times). Hypoxia (0.1% O2, 5% CO2, 94.5% N2) was achieved using an InVivo2 hypoxic workstation (Ruskinn, Bridgend, UK).
2.5. Sphingosine Kinase-1 Enzymatic Activity
SphK1 activity was performed as previously described [
45], and determined in the presence of 50 µM sphingosine, 0.25% Triton X-100 and [γ-32P] ATP (10 µCi, 1 mM) containing 10 mM MgCl
2. The labeled S1P was separated by thin-layer chromatography on silica gel 60 with 1-butanol/ethanol/acetic acid/water (80:20:10:10,
v/v) and visualized by autoradiography. Activity was expressed as picomoles of S1P formed/min/mg of protein.
2.6. Western-Blot Analysis and Antibodies
Mouse anti-HIF-1α (BD Biosciences, Le Pont de Claix, France) was used as primary antibody. Proteins were visualized by an enhanced chemiluminescence detection system (GE Healthcare, Vélizy-Villacoublay, France) using anti-mouse horseradish peroxidase-conjugated IgG (Bio-Rad, Hercules). Equal loading of protein was confirmed by probing the blots with anti-α-tubulin or anti-ß-actin antibodies (Sigma-Aldrich). Densitometry quantitation was determined with Image J software 1.53i using the area under the peak method (NIH, Bethesda, MD, USA).
2.7. Quantitative Real Time PCR
Total RNA was isolated using the RNeasy minikit (Qiagen, Courtaboeuf, France) and 1 μg was reversed transcribed to cDNA using the SuperScript Frist Stand Synthesis System (Invitrogen). Quantitative real-time PCR was performed using the MESA Blue PCR Master mix (Eurogentec). Reactions were performed using hSphK1 (forward primer 5′-CTGGCAGCTTCCTTGAACCAT-3′; reverse primer, 5′-TGTGCAGAGACAGCAGGTTCA-3′), hSphK2 (forward primer 5′-CCAGTGTTGGAGAGCTGAAGGT-3′; reverse primer, 5′-GTCCATTCATCTGCTGGTCCTC-3′), hSPL (forward primer 5′-GAACCAAGTTGCAGTTCCCACCA-3′; reverse primer, 5′-ACAGTTGTCTGGGCCATGCCATAGA-3′), β-actin-specific primers (forward primer, 5′-GCAAAGACCTGTACGCCAAC-3′; reverse primer, 5′-AGT ACTTGCGCTCAGGAGGA-3′), TBP-specific primers (forward primer, 5′- TGACCTAAAGACCATTGCACTTCG-3′; reverse primer, 5′- CGTGGTTCGTGGCTCTCTTATC -3′). For analysis, all genes were normalized to expression of zeta polypeptide (YWHAZ) gene as endogenous control.
2.8. Cell Viability Assay
The MTT reagent (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) was used to determine cell death as previously described [
46]. Briefly, cells were seeded at 5000 cells/well in 24-well plates and allowed to attach overnight. After 72 h of treatment, cells were incubated at 37 °C and 5% CO
2 with 25 μL MTT solution (5 mg/mL; Sigma-Aldrich) for 4 h. After solubilization with 500 μL of lysis buffer (DMSO), formazan was quantified by spectrophotometry with a microplate reader at 570 nm absorbance. The GI
50 values corresponding to the concentration that caused 50% inhibition of cell proliferation were calculated from dose–response curves obtained by nonlinear regression analysis. All the results were calculated from data obtained in three independent experiments.
2.9. RNA Interference Experiments
Transient interference was achieved by double-stranded human siRNAs 5′-GGGCAAGGCCUUGCAGCUCdTdT-3′ (siSphK1) as previously reported [
25]. Aleatory sequence siScr was from Eurogentec (Angers, France). Transfections were carried out using Lipofectamine 2000 in OPTI-MEM medium according to the manufacturer’s instructions (Invitrogen, Villebon-sur-Yvette, France).
2.10. Statistical Analysis
The statistical significance of differences between the means of two groups was evaluated by unpaired Student’s t test. All statistical tests were two-sided, and the level of significance was set at p < 0.05. Calculations were performed using Prism 9 (GraphPad Software, San Diego, CA). For the overall survival analysis, the outcome was the occurrence of death, and the time was defined as the time between the date of diagnosis and the date of death (event) or the date of last follow-up (censorship). For the analysis of metastasis progression-free survival (MPFS), the outcome measure was the occurrence of metastasis and/or the occurrence of death. The time was defined as the time between the date of diagnosis and the date of metastatic progression, if there was presence of metastases, or the date of death (event) or the date of last follow-up (censorship). Patients who locally relapsed as their first event were considered to be censored data, in order to avoid the bias related to the quality of the surgical resection margins. All survival rates were estimated by the Kaplan–Meier method with 95% confidence intervals (CI). The association between the presence of markers expressed in percentage of cells (for a 10% increase in labeled cells) or using the IRS score (for an increase in the score of 1) and the occurrence of death or metastases/death was estimated using a Cox proportional hazards model with an adjustment for age. The association between the presence of markers and response to chemotherapy was estimated using logistic regression with adjustment for age. These multivariate analyzes were performed using SAS 9.4 software.
4. Discussion
Hypoxia is a characteristic of solid tumors, and the adaptation of cancer cells to hypoxia is instrumental in the development of aggressive phenotype and associated with a poor prognostic in patients [
55]. At the cellular level, the adaptation to hypoxia is predominantly mediated by the hypoxia-inducible factors (HIFs), consisting of an oxygen-sensitive α-subunit and a constitutively expressed β-subunit, that regulate the expression of target genes promoting angiogenesis such as VEGF, glycolysis such as GLUT-1, acidosis such as CA-IX (carbonic anhydrase IX), metastasis, increased tumor growth and resistance to treatments [
16].
Immunohistochemical-based analyses (HIF-1α, GLUT-1, VEGF, CA-IX…) have shown a connection between hypoxia and outcome of tumor therapy. In osteosarcoma, initial studies examined VEGF expression, which was found to be associated with poor outcome (DFS, OS) and metastasis [
56,
57,
58]. As a target gene of HIF transcription factors, VEGF was also correlated with HIF-1α expression [
57,
58]. Most studies have examined the expression of HIF-1α suggesting that it would be associated with higher pathologic grade and tumor stage, higher MVD (microvascular density), higher rate of metastasis, poorer overall survival, poorer disease-free survival and poor response to chemotherapy [
48,
50,
51,
58,
59,
60,
61,
62,
63]. Three meta-analyses recently concluded that HIF-1α expression might be an effective predictive factor of poor prognosis in osteosarcoma [
17,
18,
19]. As pointed out by the authors, the results of these meta-analyses should be interpreted with caution as some major biases could not be excluded such as the limited sample size of nearly all studies, the fact that most of the selected studies have been carried out in China, proposing that similar studies have to be conducted in populations of other origins, the heterogeneity across the assessment of HIF-1α expression (different antibodies or dilutions that could affect the sensitivity) as well as nonuniform criteria including different cut-off values defining high HIF-1α level of expression. Fewer studies have examined GLUT-1 expression in canine and human osteosarcomas leading to conflicting results [
64,
65,
66] ranging from no statistical correlation between GLUT-1 expression and DFS, survival time or percentage of necrosis in dogs [
64], to strong GLUT-1 staining in both primary sites and metastases [
65] and association with shorter DFS [
66]. In addition, these studies of GLUT-1 expression are also limited by the size of the samples (
n = 10 to 44). More recently, a study examining the expression of plasma GLUT-1 established that GLUT-1 was overexpressed in patients with osteosarcoma (
n = 42) compared to healthy volunteers (
n = 38). In addition, levels of GLUT-1 mRNA were significantly higher in tumor tissues than in adjacent healthy tissues of osteosarcoma patients [
67].
In this work, we have used TMAs from different origins, which is therefore a material, obtained with different fixation and decalcification conditions, that does not allow for reliable evaluation of HIF-1α expression, with an antibody known to be difficult to use in routine. For this reason, we relied on GLUT-1 staining using a more robust antibody used in clinical practice. We found that GLUT-1 was expressed in both long and flat bones with a higher expression of GLUT-1 in long bones. Although no correlation was found with response to chemotherapy, GLUT-1 expression was associated with a higher risk of death (OS) in flat bones and a tendency for death or metastatic progression (MPFS). These data thus suggest that hypoxia might indeed be an effective predictive factor of poor prognosis in osteosarcoma, in line with the previously reported studies with HIF-1α [
48,
50,
51,
58,
59,
60,
61,
62,
63], the transcription factor that regulates GLUT-1 under hypoxia. We also examined the relationship between hypoxia (GLUT-1) and the SphK1/S1P signaling, which has been previously documented by our team to be central for the adaptation to hypoxia in a wide array of cellular (prostate, glioblastoma, lung, breast, renal cell carcinoma) and animal (prostate, renal cell carcinoma) models [
25,
26,
27,
38].
SphK1/S1P signaling is commonly upregulated in cancer cells [
24] and correlates with poor patient prognosis in a number of tumor types (reviewed in [
68]). The expression of SphK1 in osteosarcoma had never been reported until this work. A recent study found that S1P levels in serum samples from osteosarcoma patients (
n = 17) are decreased after chemotherapy [
69], suggesting that S1P could represent a potential biomarker of chemotherapeutic efficacy. As the authors did not assess serum S1P levels in healthy controls, they could not demonstrate that S1P content is in fact increased in osteosarcoma patients.
In our tumor collection, both SphK1 and S1P
1 were expressed in long and flat bones together with GLUT-1. A strong correlation was essentially seen between GLUT-1 and SphK1, regardless of the location (flat versus long bones), suggesting that SphK1 activation is associated with hypoxia in vivo. No association between the response to treatment and the expression of GLUT-1, SphK1 or S1P
1 was found, with the exception of long-bone osteosarcomas where statistical differences were found with an increased expression of S1P
1 in poor-responder patients. These data regarding response to chemotherapy should be interpreted with caution, as the different groups of patients (long- versus flat-bone osteosarcomas) are not always treated with the same chemotherapy regimen. Importantly, age-adjusted multivariate analyses showed that GLUT-1, SphK1 and S1P
1 were associated with an increased risk of death (OS) depending on the osteosarcoma location (long versus flat bones). Interestingly, S1P
1 was the only biomarker associated with increased risk of death (OS), increased risk of death or metastasis (MPFS) and predictive of poorer response to chemotherapy in long bones. However, one should be cautious in interpreting these findings on bone tissues as the staining intensity of S1P
1 was weaker when compared to the robust GLUT-1 and SphK1 antibodies. The strong expression of SphK1 in tissues was confirmed by the quantification of SphK1 enzymatic activity, which we originally reported in prostate cancer patients [
22], using 12 osteosarcoma compared to 10 nontumoral bone samples. We found that SphK1 activity was dramatically augmented in osteosarcoma with over a 50-fold increase in S1P production (431 pmoles/min/mg versus 6.6,
p = 0.023). Measuring enzymatic activity in human biopsies is technically demanding and hardly amenable to routine use as a prognostic factor, yet it provides the most accurate information by directly quantifying the activity of an enzyme when compared to mRNA or protein assessment. The relationship between SphK1/S1P signaling and hypoxia was further validated in vitro in various osteosarcoma cell lines exposed to hypoxic conditions. Our data suggesting that SphK1/S1P signaling activation is associated with HIF-1α expression in osteosarcoma are in line with our previous findings in prostate and renal cell carcinoma cell models [
25,
26,
27,
38]. Interestingly, in contrast to classical chemotherapy, the dual SphK1 and S1P
1 inhibitor FTY720 remarkably reduced cell proliferation to the same extent in normoxic and hypoxic conditions in all osteosarcoma cell lines, suggesting that efficacy of FTY720 was related to its anti-HIF-1α effect.