Signal regulatory protein $\alpha $ (SIRP$\alpha$) is a transmembrane protein that is commonly expressed in cells of the hematopoietic system and brain. Its function is not fully understood but it includes tumor suppressor properties and effects on differentiation. SIRP$\alpha $ may play a role in the development of medulloblastoma (MB), a WHO grade IV brain tumor, which is the most common malignant brain tumor in childhood. The aim of the current study was to determine the possible role of SIRP$\alpha $ in MB cells. Interestingly, in contrast to normal cerebellum, SIRP$\alpha $ mRNA was strongly downregulated in MB and its protein was not detectable in MB tissues. This down-regulation in MB cells was associated with transcriptional silencing of SIRP$\alpha $ via CpG island promoter hypermethylation. Furthermore, Oncomir cluster miR17-92 and miR-106a were correlated with SIRP$\alpha $ gene silencing in MB tumor specimens and cell lines. Histone modification and inhibition of DNA methylation using TSA (20 nM) for 24 hrs and 5-AZA (5 $\mu $M) and DZnep (2.5 $\mu $M) for 72 hrs, respectively, increased SIRP$\alpha $ expression 25-40 fold and resulted in 90% cytotoxicity of MB tumor cell lines D283-med and D458-med. Remarkably, forced upregulation of SIRP$\alpha $ by viral transduction in MB cell lines did not affect cell growth. In conclusion, SIRP$\alpha $ is epigenetically silenced in MB cells and tumor specimens by promoter hypermethylation and possibly by miRNA expression. SIRP$\alpha $ hypermethylation in MB might reflect the precursor cell state of these cells, rather than being a tumor-specific event, since SIRP$\alpha $ overexpression did not influence MB cell viability. The mechanism of the anti-MB action of epigenetic therapy requires further investigation since our findings indicate that this effect is independent of SIRP$\alpha $ upregulation.
Medulloblastoma (MB), a WHO grade IV tumor, constitutes 20% of all brain tumors in children and is the most common malignant brain tumor in childhood. MB arises in the cerebellar vermis, invading the cerebellum, occasionally metastasizing leptomeningeally via the cerebrospinal fluid. Aggressive therapy, which nowadays consists of chemoradiotherapy after initial surgery and maintenance chemotherapy has improved survival over the last three decades and has resulted in an 10-year overall survival of 42-70% in high-risk patients, depending on metastasis stage, and up to 91% in non-metastasized, standard-risk patients[1]. The downside to this improved survival is the therapy associated occurrence of late effects, such as endocrine and neurocognitive sequelae, mainly attributable to tumor localization and radiation therapy. There is an obvious need to define tumor specific targets, enabling tumor-targeted drug therapy to further ameliorate survival and increase the quality of life by reduction of late effects. In order to identify druggable targets, it is essential to decipher the molecular mechanism(s) on which MB cells rely to sustain their tumor growth. Hence, in our present study, we investigated signal regulatory protein $\alpha $ (SIRP$\alpha$) (PTPNS1, SHPS-1, CD172a, p84, gp93, MFR or BIT) in MB, as it has previously been associated with growth and differentiation of other cancers[2,3,4,5,6,7,8]. SIRP$\alpha $ is a transmembrane protein, which has three extracellular immunoglobulin superfamily (IgSF) domains and a cytoplasmic tail containing four immunoreceptor tyrosine-based inhibitory motifs (ITIMs). Ligation with CD47 and other ligands, such as thrombospondin-1 (THSP-1), leads to phosphorylation of these ITIMs, resulting in recruitment, phosphorylation and activation of tyrosine phosphatases SHP-1 and SHP-2, producing largely inhibitory and tumor-suppressive signals via SHP-2, mainly involving ERK and Akt signaling pathways[9,10,11].
SIRP$\alpha $ is an important protein in neural development, especially facilitating neuronal differentiation, maturation and synaptogenesis. Active neurons are able to shed a soluble fragment from the extracellular domain of the SIRP$\alpha $ protein to generate a synaptogenic signal, thereby governing presynaptic neurons. In addition, SIRP$\alpha $ is involved in inducing new synapses[12,13] and synapse maturation[14,15]. In cerebellar Purkinje cells, SIRP$\alpha $ associates with dendrites and growing parallel fibers. Upon interaction of SIRP$\alpha $ with its ligand CD47, abundantly expressed in brain, downstream signaling is mediated by the Src family of kinases, promoting the development of axons and dendrites in hippocampal neurons[12,13]. Thrombospondins interacting with SIRP$\alpha $ are also involved in this process[16]. SIRP$\alpha $ was found to induce apoptosis in acute myeloid leukemia (AML) cells upon ligation with an agonistic antibody[7]. In analogy to AML, and with its role in neuronal development, the aim of the current study was to investigate the expression and anti-tumor activity of SIRP$\alpha $ in MB.
mRNA expression of several genes involved in SIRP$\alpha $ signal transduction was determined using R2, a microarray analysis and visualization platform, provided by the Department of Human Genetics of the Academic Medical Center, Amsterdam, The Netherlands (http://r2.amc.nl). In a large MB dataset provided by the Division of Pediatric Neuro-oncology, German Cancer Research Center DKFZ, Heidelberg, Germany, we analyzed the mRNA expression of SIRP$\alpha $(SHPS1; 202897_at), CD47 (226016_at), SHP-1 (PTPN6; 206687_s_at) and SHP-2(PTPN11; 212610_at) using MAS5.0 normalized data of the previously defined four MB subgroups[17,18]. Whole genome bisulfite sequencing data were used to analyze methylation profiles of the SIRP$\alpha $ promoter region in a series of MB samples, comprising Wnt group MB (${\rm n}=5$), Shh group MB (${\rm n}=6$), Group 3 (${\rm n}=11$) and Group 4 (${\rm n}=12$) MB samples, with normal fetal (${\rm n}=4$) and adult (${\rm n}=4$) cerebellum samples as controls, collected within the International Cancer Genome Consortium (ICGC) PedBrain Tumor Project (Heidelberg cohort). Matching SIRP$\alpha $ mRNA expression array data (Affymetrix U133 plus 2.0) were used to correlate SIRP$\alpha $ methylation profiles with expression.
SIRP$\alpha $ immunohistochemistry was conducted on an independent MB tissue microarray (TMA) with tumors from 87 patients obtained from the bio-bank of the Department of Neuropathology of the Academic Medical Center. General informed consent was obtained for the use of brain tissue and for access to medical records for research purposes. Normal cerebellum, hippocampus and neocortex were used as control tissues. Tissue sections (5 $\mu $m), mounted on superfrost slides (Menzel, Braunschweig, Germany) had undergone dewaxing and rehydration, after which endogenous peroxidase activity was blocked for 30 min in methanol containing 0.3% hydrogen peroxide. Slides were then washed with distilled water and phosphate-buffered saline (PBS; 10 mM, pH 7.4). For antigen retrieval, the slides were placed in sodium citrate buffer (10 mM, pH 6.0) and heated in a microwave oven at 99$^{\circ}$C for 10 min, and cooled to room temperature. The sections were washed in PBS and pre-incubated with 10% normal goat serum (NGS) diluted in PBS 30 min prior to incubation with a SIRP$\alpha $-specific antibody (1:1000, ab8120, Abcam, Cambridge, UK) for 1 hr. The sections were washed with PBS and incubated at room temperature for 30 min with Dako REALTM EnVisionTM /HRP, Rabbit/Mouse (Dako, Glostrup, Denmark). Peroxidase activity was detected using Dako REALTM 3,3-diaminobenzidine-tetrachloride (DAB)$+$ Chromogen (Dako, Glostrup, Denmark), for 10 min in the dark. All sections were counterstained with hematoxylin, covered with a coverslip and examined under the microscope.
To determine protein (SIRP$\alpha $, SHP-1, SHP-2) levels in MB cell lines, Western blot analysis was used. Whole cell lysates were prepared using RIPA lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP40, 0.5% Na-deoxycholate, and 0.05% SDS) supplemented with 1mM pefabloc (Sigma-Aldrich, Zwijndrecht, The Netherlands). Protein concentration was determined using the Bio-Rad protein assay kit (Bio-Rad, Veenendaal, The Netherlands). Proteins (100 $\mu $g) were resolved by 7.5% SDS-PAGE and transferred to PVDF membrane (Millipore, Amsterdam, The Netherlands). Subsequently, the membrane was blocked with TBST (Tris-buffered saline with 0.1% Tween 20) $+$ 5% non-fat milk for 1 hr at room temperature (RT) and incubated with rabbit polyclonal anti-SIRP$\alpha $ (1:500; ab8120, Abcam, Cambridge, UK), rabbit SH-PTP2 (1:500, sc280, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) and $\beta $-actin monoclonal mouse antibody (1:2500; Millipore, Temecula, CA, USA). Protein lysates from tumor cell lines HL60 and CCRF-CEM were used as positive controls for immunoblots of SIRP$\alpha $ (HL60) and SHP-1 and SHP-2 (both HL60 and CEM), respectively. Thereafter, the membrane was incubated for 1 hr at RT with goat anti-mouse-HRP (Dako, Glostrup, Denmark) or goat anti-rabbit-HRP (Santa Cruz Biotechnology Inc, Santa Cruz, CA, USA), diluted 1:2500 in 1% blocking buffer. Subsequently, the membrane was washed three times with TBST $+$ 0.5% milk and SIRP$\alpha $, SHP-1, SHP-2 protein and $\beta $-actin was visualized on hyperfilm using an ECL plus system (Amersham Bioscience, England).
The human MB tumor cell lines D283-med, D425-med, D458-med and D556-med (kindly provided by Dr. Darrell Bigner, Duke University, NC, USA) were routinely cultured in suspension in Dulbecco's Modified Eagle Medium (DMEM; PAA Laboratories GmbH, Pasching, Austria) containing glutamine, supplemented with 1% penicillin/streptomycin (PAA Laboratories GmbH, Pasching, Austria) and 10% fetal bovine serum (FBS; PerBio Science Nederland B.V., Etten-Leur, The Netherlands). HL-60 (human promyelocytic leukemia) and CCRF-CEM cells (human pediatric acute lymphoblastic T-cell leukemia) were routinely cultured in RPMI-1640 medium (Gibco Laboratories, Irvine, UK) supplemented with 10% fetal calf serum (Integro BV, Dieren, the Netherlands).
Real-time Quantitative PCR (RT-qPCR) analysis was carried out on cDNA obtained from 1 $\mu $g total RNA, isolated using Trizol (Life Technologies) with the Omniscript RT kit (Qiagen, Venlo, The Netherlands) following the manufacturer's procedures. Validated Quantitect primers for SIRP$\alpha $ and the housekeeping gene GAPDH were purchased from Qiagen.
For the RT-qPCR analysis of mature microRNA clusters miR-17-92, miR-106a-25 and 106b-363, small RNAs were isolated from normal cerebellum and MB tissues, further characterized in Table 1, using the Mirvana kit, following the manufacturer's instructions. Validated primers for hsa-miR-17, hsa-miR-20a, hsa-miR-106a, hsa-miR-106b and hsa-miR-363, and control microRNA miR-20b and housekeeping gene RNU48 were purchased from Life Technologies. MiR-20b was employed as control micro-RNA, predicted not to interact with SIRP$\alpha $. Per 10 $\mu $l RT-qPCR reaction containing FastStart Universal SYBR Green Master reaction mixture (Roche), 0.25 $\mu $l cDNA were added. RT-qPCR was performed using the ABI7500 real time thermal cycler (Life Technologies). The delta-delta CT method[19] was applied to obtain relative SIRP$\alpha $ mRNA levels. Results are presented as fold change of SIRP$\alpha $ mRNA and these miRNAs to reference sample 9 (Table 1, 18-year old female, normal cerebellum).
| Sample# | Age | Sex | Histology |
|---|---|---|---|
| Normal cerebellum | |||
| 1 | 33GW | Female | |
| 2 | 37GW | Male | |
| 3 | 1y | Female | |
| 4 | 2y | Female | |
| 5 | 9y | Male | |
| 6 | 10y | Male | |
| 7 | 15 y | Male | |
| 8 | 17y | Female | |
| 9 | 18y | Female | |
| 10 | 31 y | Male | |
| Medulloblastoma | |||
| 1 | 1y | Male | MB NOS |
| 2 | 3y | Female | Classical MB |
| 3 | 4y | Male | Desmoplastic Nodular MB |
| 4 | 7y | Male | MB NOS |
| 5 | 8y | Male | Classical/Desmoplastic MB |
| 6 | 13y | Male | Desmoplastic MB |
| 7 | 15 y | Male | Large Cell Anaplastic MB |
| 8 | 31 y | Male | Desmoplastic MB |
| 9 | 6y | Female | Desmoplastic MB |
| 10 | 9y | Male | Desmoplastic MB |
| 11 | 5y | Female | MB NOS |
| 12 | 31 y | Female | Desmoplastic MB |
| 13 | 8y | Male | Desmoplastic MB |
| Medulloblastoma cell lines | |||
| A | D283-med | ||
| B | D458-med | ||
| C | D556-med | ||
For methylation specific PCR, DNA was isolated using Trizol. DNA was then subjected to bisulfite treatment using the EZ DNA Methylation Gold Kit (Zymo Research) following the manufacturer's instructions. DNA methylation patterns in the CpG island of the PTPNS1 gene (encoding SIRP$\alpha $; accession AL117335) were determined by PCR with specific primers designed for methylated, bisulfite-treated DNA. Two primer sets for MSP were designed to target two regions within the promoter CpG island (named R1 and R2), R1(F) 5'-AGGGTTAAAAGGTAGACGTTTTTTC-3', R1(R) 5'-ACGCTACGAATAACTACATCGTC-3' and R2(F) 5'-GTTCGTTTAGTTTCGAGTTTTAGC-3', R2(R) 5'-CACCTTCTCT ACGAACGAACG-3'. PCR conditions for R1 were: 94$^{\circ}$C 3 min (1 cycle); 94$^{\circ}$C 30 s, 60$^{\circ}$C 30 s (40 cycles); 72$^{\circ}$C (1 cycle), for R2: 94$^{\circ}$C 3 min (1 cycle); 94$^{\circ}$C 30 s, 56$^{\circ}$C 30 s (40 cycles); 72$^{\circ}$C (1 cycle). A PCR for the bisulfite converted housekeeping gene $\beta $-actin (ACTB-Fw: TGGTGATGGAGGAGGTTTAGTAAGT, ACTB-Rv: AACCAATAAAACCTACTCCTCCCTTAAA) was performed as a reference. For RT-qPCR mRNA expression experiments, snRNP was used as a housekeeping gene. MSP was performed on an ABI 7500 real-time PCR system (Applied Biosystems).
DNA methyltransferase inhibitor 5-aza-cytidine (5-Aza), lysine methyltranferase enhancer of zeste 2 (EZH2) inhibitor 3-deazaneplanocin A (DZN) and histone deacetylase (HDAC) inhibitor trichostatin A (TSA) were purchased from Sigma Aldrich (St Louis, MO, USA). For demethylation experiments, cells were seeded (0.5 $\times $ 10$^{5}$/ml) and incubated in 5% CO$_{2 }$ humidified atmosphere at 37$^{\circ}$C to obtain logarithmic growth. After 48 hr, cells were treated with 50 nM-5 $\mu $M AZA or 50 nM-2.5 $\mu $M DZnep or vehicle for 72 h, and 20 nM TSA or vehicle during the last 24 hr. Growth medium containing demethylating agents was renewed daily.
To ascertain the impact of SIRP$\alpha $ overexpression in MB cells, a sequence verified, lentiviral vector containing a full-length human SIRP$\alpha $ construct (pLenti6.3 - SirpBIT IRES NGFR) was kindly provided by Dr. T van den Berg (Department of Blood Cell Research, Sanquin Research and Landsteiner Laboratory, Academic Medical Centre, University of Amsterdam, Amsterdam, the Netherlands). The SirpBITconstruct was cloned into a pLenti4/TO/V5-DEST vector, using the ViraPowerTM HiPerformTM Gateway® Vector Kit (Life Technologies) according to manufacturer's instructions. D458-med cells were transduced to stably express a tetracycline (tet) repressor (Tet-on), using the ViraPowerTM HiPerformTM T-RExTM Gateway® Vector Kit. Subsequently, these cells were transduced with the vector containing the SirpBIT construct, allowing for inducible expression of SIRP$\alpha $ upon tetracycline treatment.
For the in silico analysis of SIRP$\alpha $ (SHPS1; 202897_at), CD47 (226016_at), SHP-1 (PTPN6; 206687_s_at) and SHP-2(PTPN11; 212610_at) in datasets of normal cerebellum and MB subgroups, one-way analysis of variance (ANOVA) was used to compare expression between these datasets. A $p$-value $<$ 0.0005 was considered statistically significant.To determine the correlation between SIRP$\alpha $ mRNA expression and CpG island methylation, and SIRP$\alpha $ mRNA and micro-RNA expression, Pearson's correlation coefficient (r) was calculated. For analysis of correlation of micro-RNA and SIRP$\alpha $ mRNA expression, Spearman's rank correlation coefficient($\rho$) calculation was used. Student's t-test was used to compare miRNA expression levels between normal cerebellum and MB or MB cell lines.
SIRP$\alpha $ mRNA expression in MB and non-malignant cerebellum was determined in silico in four subgroups of MB: Wnt group MB (${\rm n}=54$), Shh group MB (${\rm n}=116$), Group 3 MB (${\rm n}=94$) and Group 4 MB (${\rm n}=169$). SIRP$\alpha $ mRNA expression was found to be strongly downregulated in all of these four subgroups when compared to normal cerebellum (Fig. 1A).
Fig. 1.SIRP$\alpha $, CD47, SHP-1 and SHP-2 mRNA expression. In silico analysis using R2 analysis software on public datasets of non-neoplastic normal adult cerebellum (white) versus datasets of Wnt group, Shh group, Group 3 and Group 4 MB (grey), depicting (A) SIRP$\alpha $ (202897_at), (B) CD47 (226016_at), (C) SHP-1 (206687_s_at) and (D) SHP-2 (212610_at) mRNA expression. Outliers are indicated by ``o''. All MB subgroups showed significant down-regulation of SIRP$\alpha $ and SHP-1 mRNA expression relative to the normal cerebellum ($p <$ 0.00001), whereas SHP-2 was significantly overexpressed in MB. CD47 mRNA was not differentially expressed in MB.
To determine expression of SIRP$\alpha $-interacting genes, expression of the SIRP$\alpha $ ligand CD47 and SIRP$\alpha $ ITIM binding tyrosine phosphatases SHP-1 and SHP-2 were also determined in these normal and MB tissues. CD47 showed differential expression in MB, comparable to that observed in normal cerebellum (Fig. 1B). For the expression of tyrosine phosphatases SHP-1 and SHP-2 in MB, downregulation of the former (Fig. 1C) and upregulation of the latter ( Fig. 1D) was observed as compared to normal cerebellum.
SIRP$\alpha $ protein levels were evaluated by immunohistochemistry in non-malignant cortex (Fig. 2A), hippocampus (Fig. 2B) and cerebellum (Fig. 2C) and compared to a tissue array of MB, composed of 276 pediatric MB tissue specimens obtained from 87 patients (example pictures depicted in Fig. 2D-fig2F). Subgroup information for these tumors was obtained previously by immunohistochemistry using antibodies for the subgroup-specific protein markers $\beta $-catenin (WNT), DKK1 (WNT), SFRP1 (SHH), NPR3 (Group 3), and KCNA1 (Group 4) as previously described[17]. The patients' clinical and tumor characteristics are summarized in Table 2.
Fig. 2.SIRP$\alpha $ immunohistochemistry on paraffin-embedded sections. (A) Representative images of cytoplasmatic SIRP$\alpha $ staining in neuronal synapses in non-malignant cortex, (B) dentate gyrus of hippocampus and (C) molecular layer of cerebellum. (D,E) No SIRP$\alpha $ protein expression was observed in MB tissues. (F) Positive SIRP$\alpha $ staining in normal cerebellar tissue adjacent to MB tumor tissue. Arrows indicate positive staining.
| Characteristics (N=87) | Number (%) |
|---|---|
| Gender |
27 (31) |
| Female | 12(14) |
| Unknown | 48 (55) |
| Age at diagnosis (median 7.0, range (1-53) | |
| Infants (<4) | 7(8) |
| Children (4-16) | 19 (22) |
| Adults (> 16) | 15 (17) |
| Unknown | 46 (53) |
| Histology |
23 (26) |
| Large cell/anaplastic | 3(4) |
| Nodular/desmoplastic | 8 (9) |
| Unknown | 53 (61) |
| Molecular subgroups |
7 (8) |
| SHH | 18 (21) |
| Group 3 | 14(16) |
| Group 4 | 33 (38) |
| Unknown Metastatic stage | 15 (17) |
| M0 |
35 (40) 6(7) |
| Unknown | 46 (53) |
| Survival |
61 (70) |
| Diseased | 26 (30) |
A strong SIRP$\alpha $ protein signal was detected in neuronal synapses in the cortex, hippocampal dentate gyrus and the molecular layer of the cerebellum. Most strikingly, immunohistochemical analysis of the MB tissue arrays, showed no detectable levels of SIRP$\alpha $ protein in any of these MB tissues (Fig. 2D and fig 2E), in agreement with the downregulation of SIRP$\alpha $ mRNA in silico. In one patient sample that included adjacent normal cerebellar tissue (molecular layer), the cerebellar tissue stained positively for SIRP$\alpha $, whereas tumor cells were negative (Fig. fig2F). Consistent with the MB tissue staining, no SIRP$\alpha $ protein was detected in the MB tumor cell lines D283-med, D458-med and D556-med, whereas its downstream ligand tyrosine phosphatase SHP-2 was expressed. In addition, SHP-1 was also not expressed in these MB tumor cells (Fig. 3).
Fig. 3.Immunoblot of baseline SIRP$\alpha $ (75-110 kD), SHP-1 (57 kD), SHP-2 (80 kD) and $\beta $-actin (42 kD) in MB cell lines and control cell lines (CEM and HL60 cells respectively).
To determine the molecular mechanism underlying SIRP$\alpha $ silencing in MB, whole genome bisulfite sequencing analysis was used to investigate in silico whether CpG island hypermethylation contributes to the observed downregulation; the results of this analysis are depicted in Fig. 4A. Whole genome bisulfite sequencing data from the four MB subgroups versus adult and fetal normal cerebellum revealed hypermethylation (red) of the promoter region (indicated by red bar*) in most MB samples and, partially, in normal fetal cerebellum, whereas in normal adult cerebellum hypomethylation (blue) was observed. CpG island hypermethylation of SIRP$\alpha $ was irrespective of MB subtype. Interestingly, a similar pattern of hypermethylation in MB and fetal cerebellum, and hypomethylation in adult cerebellum was observed in the region just downstream of the transcription start site (green bar**).
Fig. 4.Epigenetic silencing of SIRP$\alpha $ in MB. (A) SIRP$\alpha $ hyper- (red), and hypomethylation (blue) in MB, normal fetal cerebellum, and normal adult cerebellum. Depicted are the CpG island promoter (red bar*), and the region downstream of the transcription start site (green bar**). (B) Strong negative correlation of SIRP$\alpha $ CpG island methylation with SIRP$\alpha $ mRNA expression in normal adult and fetal brain and MB subgroups (${\rm r}=-0.76$). (C) Methylation-specific PCR (MSP) on two regions (R1/R2) in SIRP$\alpha $ promoter CpG island on an independent set of normal cerebellum tissues, MB and MB cell lines, showing methylation in 7/13 MB tissues, versus 2/10 normal cerebellum tissues in R1 and methylation in 5/13 MBs and in 0/10 normal cerebellar tissues in R2. Hypermethylation of both regions was detected in all three MB cell lines.
To validate the observed in silico SIRP$\alpha $ promoter hypermethylation in MB, methylation specific PCR was performed on bisulphite treated DNA of an independent set of normal cerebellum tissues and MB (characteristics in Table 2). Two regions (R1 and R2) in the CpG island within the SIRP$\alpha $ promoter were investigated and the results are depicted in Fig. 4B. A similar pattern was observed as was found in silico. For R1, methylation was observed in 7/13 MB tissues, versus 2/10 in normal cerebellum tissues (one fetal and one adult tissue sample). R2 was methylated in 5/13 MBs and in 0/10 normal cerebellar tissues. For MB cell lines D283-med (A), D458-med (B) and D556-med (C), hypermethylation of both regions was observed.
To ascertain the correlation between SIRP$\alpha $ methylation and expression in MB and cerebellum, the extent of CpG island methylation of the SIRP$\alpha $ gene was plotted against SIRP$\alpha $ mRNA expression, using in silico analysis of expression and methylation data (Fig. 4C). SIRP$\alpha $ CpG island methylation was found to be in a strong negative correlation with SIRP$\alpha $ mRNA levels in normal adult and fetal brain and the four MB subgroups (Wnt, Shh, Group 3 and 4), with a Pearson's correlation coefficient (r) of -0.76 ($p < 0.000005$).
To overcome the CpG island hypermethylation-based SIRP$\alpha $ gene silencing, MB cell lines D458-med and D556-med were treated for 72 hr with epigenetic modulators as single drugs or as a combination of drugs, and the results are depicted in Fig. 5. All drugs were able to induce SIRP$\alpha $ expression, and a synergistic effect was observed when cells were treated with a combination of the three drugs. In particular, treatment with 5-AZA and DZnep resulted in severe cytotoxicity, even up to 95% when these agents were combined (data not shown). To investigate mRNA expression without extensive cell death, the optimal epigenetic modulator drug exposures were 50 nM, 50 nM, 20 nM for 5-AZA, DZnep and TSA, respectively for 24 hrs. We found SIRP$\alpha $ mRNA upregulation although it did not affect cell viability (shown in Supplementary Fig. S1).
Fig. 5.Restoration of SIRP$\alpha $ mRNA expression upon epigenetic modulation. Seventy two hours of treatment of the MB cell lines D458-med and D556-med with 5-aza-cytidine (5-Aza), 3-deazaneplanocin A (DZNep) and Trichostatin A (TSA) increased SIRP$\alpha $ expression up to 20-40-old with the combination of the drugs.
These data suggest that the antitumor activity of these epigenetic modulating drugs is independent of SIRP$\alpha $ expression; low concentrations of epigenetic modulating drugs does not affect cell viability, while SIRP$\alpha $ expression is induced. Therefore, we achieved forced SIRP$\alpha $ expression in MB cells by transduction of an inducible, human SIRP$\alpha $ expression construct. Upon tetracycline treatment, a strong induction of both SIRP$\alpha $ mRNA (Fig. 6A) and protein (Fig. 6B) expression was observed, whereas in the negative controls, D458-med cells expressing an empty TET-repressor, no induction of SIRP$\alpha $ was observed. However, no differences in tumor cell viability between SIRP$\alpha $ expressing and non-expressing D458-med cells were observed (Fig. 6C). In addition, restoration of SIRP$\alpha $ expression did not influence protein expression of its downstream ligand phosphatase, SHP-2 (Fig. 6B).
Fig. 6.Overexpression of SIRP$\alpha $ overexpression in MB cell line D458-med by using an inducible lentiviral construct. Strong induction of (A) SIRP$\alpha $ mRNA and (B) protein was observed in D458-med cells with TET-on inducible SIRP$\alpha $ overexpression construct (black bars), versus no induction of expression in control cells (grey bars). (C) Cell viability in D458-med cells is independent of the expression of SIRP$\alpha $. No difference in viability between D458-TET-SIRP$\alpha $ cells (black line) versus D458-TET-control cells (grey line).
As SIRP$\alpha $ expression was recently described to be post-transcriptionally regulated in macrophages by microRNA cluster miR-17-92[20], which is overexpressed in MB[21,22,23], we investigated the correlation of microRNA cluster miR-17-92, and its paralogous clusters miR-106a-363 and miR-106b-25 in normal cerebellum and MB tumor specimens; the results are depicted in Fig. 5. A trend to 3.4-fold upregulation of miR-17 ($p = 0.079$) and miR-20a (2.9-fold, $p = 0.058$) and a significant 4.0-fold overexpression of miR-106a ($p = 0.038$) was observed in MB specimens, when comparing normal cerebellum with MB. Significant differences ($p < 0.0001$) were observed when comparing normal cerebellum with MB tumor cell lines D283-med, D458-med and D556-med, with 11-, 10- and 18-fold upregulation of miR-17, miR-20a and miR-106a, respectively (Fig. 7A). No differences between normal cerebellum and MB specimens or MB tumor cell lines were observed for MiR-106b and miR-20b. MiR-363 was significantly, 5.4-fold, down-regulated in MB, compared to normal cerebellum ($p = 0.01$).
Fig. 7.Expression of oncomiR cluster 17-92 correlates with SIRP$\alpha $ mRNA expression. (A) As compared to normal cerebellum, miR-106a is significantly overexpressed in MB ($p = 0.038$), while miR-17 and miR-20a show a trend to upregulation ($p = 0.076$ and $p = 0.058$ respectively). miR-17, miR-20a and miR-106a are overexpressed in MB cell lines D283-med, D458-med and D556-med ($p < 0.0001$). miR-363 is significantly downregulated in MB ($p = 0.01$). (B) miR-17, miR-20a and miR-106a Are negatively correlated with SIRP$\alpha $ mRNA expression (Spearman's $\rho = -0.68$, $-$0.66 and $-$0.68 respectively). miR-106b, miR-363 and MiR20b do not significantly correlate.
Using sample 9 as a reference, SIRP$\alpha $ mRNA expression was plotted against miR-17, miR-20a, miR-106a, miR-363, miR106b and miR20b (Fig. 7B), and Spearman's rank correlation coefficients ($\rho )$were calculated for each micro-RNA. For miR-17, miR-20a and miR-106a, a significant correlation was found with SIRP$\alpha $ mRNA, with $\rho $-values of $-$0.68, $-$0.66 and $-$0.68 respectively, ($p < 0.005$). MiR-106b and MiR-20b expression was not significantly correlated with SIRP$\alpha $ expression ($\rho = -0.21$, $p = 0.37$ and $\rho = 0.40$, $p = 0.08$ respectively).
Based on its established role in other cancers, we investigated the neuronal synaptic protein SIRP$\alpha $ in MB. We found that SIRP$\alpha $ was not expressed in MB, in striking contrast with the normal cerebellum, in which strong expression was observed. In MB tumor cell lines D283-med, D458-med and D556-med, SIRP$\alpha $ was also not detectable. Remarkably, these cells did express SHP-2, which is a phosphatase known to elicit, mostly inhibitory, signaling by SIRP$\alpha $. This raised the question of how SIRP$\alpha $ re-expression would affect the behavior of MB tumor cells. In elucidating the mechanisms underlying loss of SIRP$\alpha $ expression, we found SIRP$\alpha $ to be transcriptionally downregulated in MB patient specimens and MB tumor cell lines by epigenetic repression via CpG island promoter hypermethylation. This hypermethylation was also observed in fetal cerebellum, whereas in adult cerebellum it was absent. This supports the idea that MB is a result of impairment of normal cerebellar precursor cell development. Epigenetic mechanisms seem to play an important role in gene expression in MB. In the latter brain tumor, multiple developmental pathways are epigenetically silenced both through CpG island promoter hypermethylation[24,25,26,27] as well as methylation of histones[28]. In recent years, comparative genomic hybridization (CGH) and gene expression profiling studies have made a revolutionary increase in the insight in MB biology and revealed that MB actually consists of four distinct disease entities, with sonic-hedgehog (Shh), wingless (Wnt), group 3 and group 4 tumors, each with its own unique molecular background, prognosis and potential therapeutic targets. Interestingly, these four MB subgroups were recently more robustly defined by 450k DNA methylation analysis, which re-emphasized that MB is actually a group of brain cancers characterized by epigenetic dysregulation of normal developmental processes[29,30,31,32,33,34,35]. Notably, with in silico analysis, non-promoter CpG island hypermethylation was also observed in a CpG island downstream from the SIRP$\alpha $ gene transcription start site. Non-promoter CpG island hypermethylation, has been described to both positively and negatively influence gene expression in other cancers[36,37].
We employed epigenetic modulators in MB cell lines D283-med and D458-med, aiming at re-expression of SIRP$\alpha $, and subsequent activation of SHP-2, to decrease cell viability, as postulated from the tumor-suppressive effects seen in other cancers. We therefore treated these MB tumor cells with a combination of DNA methyltransferase inhibitor 5-Aza-Cytidine (AZA), HDAC inhibitor trichostatin A (TSA) and histone methyltransferase EZH2 inhibitor 3-deazaneplanocin A (DZnep) leading to restored expression of SIRP$\alpha $ (Fig. 5), which concurrently resulted in a strong reduction in cell viability. To ascertain whether the strong effects of epigenetic therapy on MB cell survival are attributable to SIRP$\alpha $ upregulation or other `un-silenced' genes, we transduced D458-med cells with a TET-repressed, inducible SIRP$\alpha $ overexpression construct. Induction with tetracycline resulted in a 40-fold upregulation of SIRP$\alpha $ mRNA in these cells, after 72 hr of treatment. Furthermore, strong induction of SIRP$\alpha $ protein expression was observed. However, no differences in cell viability were observed when compared to control D458-med TET-repressor cells. This finding is in sharp contrast with other tumors, in which signaling via SHP-2, negatively influences tumor cell survival[3,38]. The absence of an effect on cell viability could not be explained by differences in expression of downstream effector phosphatase SHP-2, since this was indifferently expressed, irrespective of epigenetic therapy (Fig. 6B). The observed SIRP$\alpha $ hypermethylation and silencing in MB tissues and in fetal brain, together with an absence of effect upon SIRP$\alpha $ overexpression in MB cells, might suggest that SIRP$\alpha $ silencing is not a tumor specific event, but rather reflects a developmental state of cerebellar precursor cells that give rise to MB. Other factors such as SIRP$\alpha $ protein mislocalization, or absent ligand (CD47, thrombospondins) or downstream effector expression might explain that no effect of SIRP$\alpha $ expression in MB was seen. These surprising findings require further investigation. In this respect, epigenetic cancer therapeutics are of increasing interest in MB[39]. Short exposures to lower doses of these drugs were also able to induce SIRP$\alpha $ in MB cells (Supplementary Figure S1), albeit to a lesser extent. More research is therefore needed to optimize the dose and schedules when combining these epigenetic modulator agents, since these are likely to influence their demethylating and/or cytotoxic properties[40].
Besides epigenetic changes, increasing evidence suggests that microRNAs[28,41,42] post-translationally regulate gene expression in MB. MicroRNA clusters miR-17/92 (OncomiR) and paralogous clusters miR-106a/25 and miR-106b/363 were previously described to be upregulated during neuronal lineage differentiation in stem cells, and in MB, mainly in Shh group, which are MYC/MYCN-driven tumors[21,28,43]. These microRNAs have been described to target and silence SIRP$\alpha $ mRNA[20] in myeloid cells. Interestingly, we found miR-17, miR-20a and miR-106a expression to be negatively correlated with SIRP$\alpha $ mRNA expression (Spearman's rank correlation coefficients ($\rho )$ of $-$0.68, $-$0.66 and $-$0.68, respectively, $p < 0.005$), in samples of MB tumors and normal cerebellum from infants, older children and adults, suggestive of post-translational silencing of target gene SIRP$\alpha $ by these miRNAs. For miR-363, and, as expected, for miR-20b, no significant correlation was found. Recently, miR-520d-5p and miR-520d-3p have been found to be associated with SIRP$\alpha $ downregulation in astrocytes[44] and might be interesting to investigate further in MB.
In conclusion, we found that SIRP$\alpha $ is transcriptionally and, possibly, post-transcriptionally silenced in MB as well as during human fetal brain development. Histone modification and inhibition of DNA methylation resulted in markedly increased SIRP$\alpha $ gene expression. This epigenetic modulation strongly reduced viability in MB cells, but in a SIRP$\alpha $-independent manner, since inducible SIRP$\alpha $ overexpression did not influence cell viability. This suggests that SIRP$\alpha $ hypermethylation in MB reflects the precursor cell state of these cells, rather than being a tumor-specific event. Epigenetic therapies and, possibly future micro-RNA targeted therapies could be of therapeutic benefit in MB, warranting further research to elucidate which genes are involved in these observed strong anti-tumor effects.
The authors declare no conflict of interest related to the work described in this article.