This article is mentioned in:

Introduction

Medulloblastoma (MB) represents a genetically and epigenetically heterogeneous group of neuroepithelial primary tumors arising from the cerebellum (1). The World Health Organization (WHO) has classified four histological variants of MB which include classic, desmoplastic nodular (MB/N), MB with extensive nodularity (MBEN) and large cell/anaplastic type (LC/A) (2). Later, Garré et al proposed the classification of MB into two groups, standard-risk (SR) and high-risk (HR) groups (3,4) according to the risk-adapted treatments. In 2012, an international collaborative microarray study established four molecular subgroups in MB (WNT, SHH, group 3 and 4) (5,6). It is noteworthy that Gibson et al (7) found that WNT- and SHH-subtype medulloblastoma have different anatomical locations. For example, the WNT-Subtype is located within the IV ventricle and arises at the dorsal surface of the brain-stem, while the SHH-subtype arises within the cerebellar hemispheres. Current treatment protocols are dependent on the age of patients; for patients older than 36 months of age options include maximal safe tumor resection followed by craniospinal radiotherapy and chemotherapy. Infants younger than 36 months do not receive radiotherapy due the risk of severe side-effects (8). Cure rates are dependent on the MB molecular subtype. Patients with MB-WNT have a cure rate of over 90%, while in patients with MB-group 3 40–60% face chemo or radio-resistance or relapse (8,9). This variation in treatment response is a motivation to better understand the biology of these tumors, including gene expression patterns, in order to increase cure rates and/or to reduce treatment-related sequelae.

An earlier comprehensive sequence analysis study in 88 pediatric MB samples led to the discovery of a gene, histone-lysine N-methyltransferase MLL2, not previously known to be altered in MBs (10). In 14% of MB patients in the same study somatic mutations were observed in MLL2, which is known to be a member of the trithorax complex that induces epigenetic marks that correlate with accessible euchromatin (11). The connection between MLL2 alterations and tumorigenesis has not been investigated in medulloblastoma. One possibility is that, the MLL2 gene (new approved symbol KMT2D), which codifies the MLL2 protein and which plays a role in maintaining HOX gene expression (10,12) might deregulate HOX gene expression in cancer. HOX genes encode evolutionarily conserved homeodomain-containing transcription factors and in humans there are four clusters named HOXA, HOXB, HOXC and HOXD, which are located on chromosomes 7p14, 17a21, 12q13 and 2q31, respectively.

An additional motivation for studying HOX gene expression in MB arises from the role of HOX genes in cerebellum development. As described in Wang and Zoghbi (13), cerebellum development commences from the dorsal region of the posterior neural tube. Two of the regions of the posterior neural tube, the mesencephalon and the metencephalon depend on molecular signals received from the Isthmus Organizer (IO). The Otx-2 (MB oncogene) plays a role in IO development while various genes in HOX family affect metencephalon development (13).

Moreover, the expression of HOX genes is not limited to the embryonic phase of development (14). Takahashi et al (15) studied the expression profiles of 39 HOX genes in 20 different human adult normal organs by real-time PCR. They showed that in the cerebellum six HOX genes are expressed, among them: HOXB2, HOXB3, HOXB4, HOXC4, HOXD1 and HOXD3. Other studies have investigated the expression of HOX genes in cerebellum suggesting that the number of HOX genes transcribed in the cerebellum can be even higher. For example, recently, Hutlet et al (16) demonstrated low expression of 21 specific HOX genes in adult mice cerebella, all belonging to paralogy groups (PG) 1–8 (16). These data suggest complex patterns of HOX expression in the cerebellum, which in turn suggest the relevance of multiple HOX gene regulation mechanisms. These mechanisms include the use of multiple promoters, alternative splicing, antisense transcripts, long non-coding RNA, miRNA, as well as some translation controls (16,17).

In some cancers, certain HOX genes may function as oncogene or tumor suppressor genes (18). For example, in prostate, breast and colorectal cancers loss-of-function of HOXB13 mutations suggests that HOXB13 may function as a tumor suppressor gene (19). On the other hand, in breast and colorectal cancer, upregulation of HOXB7 promotes cell proliferation, which implies that it might function as an oncogene (20,21). Apart from loss of function, additional progress has been made in understanding changes in expression levels and the activation of certain HOX genes and cancer development (22,23). In brain tumors, the deregulation of HOX genes has been described in glioblastoma (24–27) neuroblastoma (28,29), tumor atypical teratoid rhabdoid, juvenile pilocytic astrocytoma and ependymomas (30). In medulloblastoma, Bodey et al (31) observed by immunocytochemistry techniques that HOXB3, HOXB4 and HOXC6 proteins were expressed in medulloblastoma tissues. In 2014, Chakravadhanula et al (30) reported that HOXC4-6, HOXC8-13, HOXD3-4, HOXD8-11 and HOXD13 genes were upregulated and HOXD1 gene was downregulated in medulloblastoma tissues when compared to control tissues.

Recently, in order to elucidate epigenetic changes that contribute to medulloblastoma progression, Vo et al (32) demonstrated through the use of engineered deletions of Ezh2 by gene editing nucleases, that inactivation of Ezh2 drives aggressive medulloblastoma. Ezh2 is a member of the polycomb complex known to induce marks that correlate with heterochromatinization and mediate repression of HOX genes (33). In addition to repression, the authors showed that loss of Ezh2 is also accompanied by upregulation of several HOX genes (32).

These studies suggest that several epigenetic mechanisms of HOX regulation might play a role in medulloblastoma. However, the role of HOX gene dysregulation in medulloblastoma has not been investigated. In 2011, a seminal study (34) suggested ten attributes which are acquired by tumor cells and which also sustain the malignant behavior in various types of cancer. In this study, we investigate the role of two specific HOX genes, HOXA10 and HOXB4, in two of these attributes in meduloblastoma: migration and proliferation potential.

Materials and methods

Cell lines

Four independent medulloblastoma cell lines (UW402, UW473, DAOY and ONS-76) were used in this study. UW402 and UW473 cell lines, have been previously characterized regarding chromosomal heterogeneity, instability profiles, cell morphology, cell population doubling time, colony-forming efficiency as well as chemo-sensitivity heterogeneity (35). Both DAOY and ONS-76 cell lines are long-established MB cell lines (36–38). The ONS-76, UW402 and UW473 cell lines were kindly provided by the Laboratory of Pediatric Oncology of the Clinical Hospital of the Medical School of Ribeirão Preto (CHMSRP), while the DAOY cell line was purchased from the American Type Culture Collection (ATCC: HTB-186; ATTC, Manassas, VA, USA).

The three cerebellum primary cultures (CP4, CP5 and CP6) were generated from cerebellum specimens that were harvested from infants at autopsy in the Pathology Service of the CHMSRP. CP4 was obtained from cerebellum of a full term newborn who died at two months old. CP5 and CP6 were collected from the cerebella of two infants born at 22 and 23 weeks of gestation, respectively, who died soon after birth. The causes of the babies' deaths were not brain related.

The primary cells were cultured in Dulbecco's modified Eagle's medium (DMEM)-F12 medium (Gibco, Grand Island, NY, USA) with 10% fetal bovine serum (FBS; HyClone Laboratories, Inc., Logan, UT, USA), the same medium in which UW402 and UW473 cell lines were cultured. The DAOY cell line was cultured in a-MEM (Gibco) 10% FBS and ONS-76 cell line was cultured in RPMI-1640 medium (Gibco) 10% FBS in an incubator at 37°C and 5% CO2. All culture media contained penicillin at 100 units/ml and streptomycin 100 μg/ml (Gibco). The study was approved by the Ethics Research Committee of the Clinical Hospital of Ribeirão Preto and Medical School of Ribeirão Preto of the University of São Paulo under protocol number 14920/2011. All the animal procedures of the present study were conducted in agreement with the ethical principles in animal research adopted by the Brazilian College of Animal Experimentation (COBEA) and this study, under protocol number 137/2011, was also approved by the Medical School of Ribeirão Preto of the University of São Paulo, Ethics Commission of Ethics in Animal Research (CETEA).

Preparation of total RNA and microarray experiments

Total RNA was isolated from cerebellum primary cultures and medulloblastoma cell lines using RNeasy Mini kit according to the manufacturer's instructions (Qiagen, Hilden, Germany). The quantity and purity of the total RNA was determined by spectrophotometer (NanoDrop; Thermo Fisher Scientific, Waltham, MA, USA) and samples with OD 260/289 between 1.8 and 2.0 were used. In order to check the integrity of the RNA 500 ng of RNA with loading-buffer (xylene cyanol containing gelred fluorescente dye from Biotium) was loaded in a 1% miniagarose gel in Tris-Acetate-EDTA buffer. After electrophoresis, the gel was analyzed using the ImageQuant 350 transilluminator (GE Healthcare) with short-wave-length UV illumination to visualize 28S and 18S bands. The integrity of the total RNA was also evaluated by microfluidic-based electrophoresis (Agilent 2100 Bioanalyzer; Agilent Technologies, Santa Clara, CA, USA) and samples with RIN >7.0 were used for microarray experiments.

A total of 200 ng of RNA of samples and reference were reverse transcribed by the Low RNA Input linear amplification kit and then transcribed to Cy3-labelled (samples) or Cy5-labelled (reference) according to the manufacturer's instructions (Agilent Technologies). The protocols used were described in de Oliveira et al (39). Briefly, the reverse transcription reaction was incubated at 40°C for 2 h and then incubated at 70°C for 15 min to degrade the excessive RNA. Next, in vitro transcription was carried out using the T7 RNA polymerase and the dyes Cy3 and Cy5. This reaction was incubated at 40°C for 2 h. After labeling and purification (Mini Spin kit; GE Healthcare), the efficiency of Cy3 and Cy5 labeling was quantified using the NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific). The numerical values for the labeling efficiency were determined by the base-to-dye ratios of Cy3-dCTP and Cy5-dCTP per μg of cRNA. The values obtained were consistent with those recommended for hybridization onto cDNA microaarrays. Next, the labeled samples and reference were fragmented according to the manufacturer's instructions and the hybridization buffer added. After the arrays were loaded, the arrays were incubated at 65°C for 17 h for hybridization.

Dye normalization was performed by the linear lowess method. For the spike in controls, the expected log ratio was plotted against the observed log ratio as a quality control check. For each lamina, the log ratio was plotted against the log processed signal to ensure that the signal strength distribution for upregulated and downregulated features were comparable.

Microarray data analysis

The raw data were obtained from a two-color Agilent platform (SurePrint G3 Human Gene Expression 8×60K Microarray kit; G4851A) using DAOY1 as a common reference. Quantile normalization was performed on the raw data after hybridization followed by normalization between arrays to obtain the gene expression levels. Expression levels for the various HOX genes were obtained by mapping probe names to the HOX genes using the NCBI GenBank. Whenever there were multiple probes present for the same HOX gene, the average over all probes was taken to determine the expression level for that particular gene.

All heatmaps were generated using the gplots package of Bioconductor. The identification of differentially expressed genes was performed using the Limma package in Bioconductor. A significance threshold of 0.05, along with P-values adjusted according to the Benjamini-Hochberg method, was used to determine which genes were differentially expressed. Anova contrasts were used to identify differentially expressed genes in the various comparisons shown in Fig. 1C and D. All P-values were converted to their log to the base 10 equivalents before use in Fig. 1C and D. The log fold change values used in the figures were those obtained from the Limma output.

Figure 1 Total mRNA analysis identifies the HOX genes that are differentially expressed in UW402, UW473, DAOY and ONS-76 MB cell lines compared with three primary cell cultures from the cerebellum (CP4, CP5 and CP6). (A) A heatmap showing the separation of UW473 and UW402 cell lines and primary cells from cerebellum based on the expression pattern of all 39 HOX genes. (B) A heatmap showing the separation of DAOY and ONS-76 cell lines and primary cells from cerebellum based on the expression pattern of all 39 HOX genes. (C, D) Volcano plots with differentially expressed HOX genes. The black lines indicate genes significant at adjusted P<0.05 and fold change >1.5.

The raw data from high-throughput microarray experiments are available on the Gene Expression Omnibus (GEO) website under accession number GSE95684.

Real-time PCR

The cDNA was directly synthesized from 1 μg total RNA from cerebellum primary cultures and medulloblastoma cell lines using the High Capacity cDNA Reverse Transcription kit (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's protocol. Each 25 μl cDNA synthesis reaction solution contained 1 μg of total RNA, 1X buffer, 1X random primers, dNTP (8 mM), RNase inhibitor (2 U) and 62 U of MultiScribe Reverse Transcriptase (Life Technologies). The cDNA synthesis was carried out with 1 cycle at 25°C for 10 min, 1 cycle at 37°C for 120 min followed by heating to 85°C for 5 min to terminate the reaction. Gene expression of the two HOX genes were quantified using the TaqMan Universal Master Mix as previously reported by Covas et al (40) and Fontes et al (41). Briefly, cDNA (2 μl of cDNA diluted 1:10) was applied in the real-time PCR assay with TaqMan probes (Applied Biosystems, Waltham, MA, USA) HOXA10 (Hs00172012_m1), HOXB4 (Hs00256884_m1) and the TaqMan Universal Master Mix according to the manufacturer's protocol. Beta actin (ACTB) gene (4310881E) was used as reference gene. Cycling was done for 2 min at 50°C, followed by denaturation at 95°C for 10 min. The amplification was carried out with 40 cycles of 15 sec at 95°C and 60 sec at 60°C. We obtained the number of amplification cycles needed for reaching the threshold fluorescence within the log-linear phase of the amplification curves after setting the appropriate baseline in each color channel of the GeneAmp® 7500 software using the manual baseline method. The relative expression level of target gene was shown as expression relative units (ERU) and was quantified according to the formula ERU = 10,000/2dCT following Albesiano et al (42), where ΔCT = CT target sample - CT reference sample. As in the study by Albesiano, the reference gene level (in our case ACTB level) is set equal 10,000. For each sample in Fig. 2A and B, cDNA was independently synthesized three times, and two technical replicates were carried out on each batch of independently synthesized cDNA. The same procedure was followed for the samples in Fig. 3C and E.

Figure 2 Quantitative validation of the microarray results by real-time PCR for HOXA10 and HOXB4 genes in MB cell lines and primary cells from cerebellum. (A) Quantitative RT-qPCR analysis was performed using total RNA from DAOY, ON-S76, UW402 and UW473 cell lines and primary cells from the cerebellum. The amount of HOXA10 transcript is relative to β-actin. (B) Quantitative RT-qPCR analysis was performed using total RNA from DAOY, ON-S76, UW402 and UW473 cell lines and primary cells from the cerebellum. The amount of HOXB4 transcript is relative to β-actin. (C) Heat Map showing the clustering between ON-S76, UW402, DAOY73 and UW473 generated by the 22 genes known from Northcott et al (6) to separate WNT SHH, group C and group D of medulloblastoma.

Figure 3 DAOY cells were transduced with shHOXA10-GFP-PUR and UW473 cells were transduced with HOXB4-GFP-PUR in order to investigate the effect of HOXA10 knockdown and HOXB4 overexpression in these cells. Control cells were transduced with empty vectors. (A, C and D) Generation and characterization of DAOY/shHOXA10+ cells and DAOY-control cells (with empty vector). (B, E and F) Generation and characterization of UW473-HOXB4+ cells and UW473-control cells (with empty vector). The transduction efficiency was evaluated by fluorescence microscopy and flow cytometry and >93% of the cells were GPF+ cells. Scale bar corresponds to 500 μm. The HOXA10 gene was silenced in DAOY/shHOXA10+ cells and the HOXB4 gene was overexpressed in UW473/HOXB4+ cells. The HOXA10 and HOXB4 expression levels were assessed by quantitative RT-qPCR analysis. The amount of HOXA10 and HOXB4 transcripts are relative to β-actin. The HOXA10 mRNA was reduced by 78% in comparison with the control cells. The HOXB4 mRNA was increased by 33.735±2.971 URE in comparison with control cells. Student's t-test was used with significance level of 0.05. The HOXA10 and HOXB4 proteins were assessed by western blot analysis. HOXA10 and HOXB4 proteins are present at the expected size. The HOXA10 protein level is decreased in DAOY/shHOXA10+ cells compared with DAOY-control cells and the HOXB4 protein level is increased in UW473/HOXB4+ cells compared with UW473-control cells.

Gene silencing

The lentiviral particles (Santa Cruz Biotechnology, Santa Cruz, CA, USA) used contained the sequences of short hairpin for human HOXA10 gene (shHOXA10), the green fluorescent protein (GFP) reporter gene and the resistance gene to the puromycin antibiotic (PUR), described as shHOXA10-GFP-PUR, and the empty vector control containing only the GFP and PUR gene sequences (sh-GFP-PUR). The DAOY cell line was transduced using hexadimethrine bromide (6 μg/ml; Sigma-Aldrich, St. Louis, MO, USA) at a multiplicity of infection (MOI) of 5. After 24 h, the cells were treated with puromycin antibiotic (1 μg/ml; Sigma-Aldrich, Hamburg, Germany) for 6 days. Cell lines were identified as DAOY/shHOXA10+ and DAOY/control. The cell transduction efficiency was evaluated by fluorescence microscopy and flow cytometry for GFP, and gene silencing efficiency was evaluated by real-time PCR and western blot analysis.

Gene overexpression

The human HOXB4 gene was over-expressed in UW473 the cell line by the cell transduction method. The vector used contained the sequences of human HOXB4 gene, the GFP reporter gene and the resistance gene to the antibiotic puromycin, described as HOXB4-PUR-GFP (VB150311-10017; Cyagen Biosciences, Santa Clara, CA, USA). As a control, we used an empty vector containing only the GFP and PUR gene sequences (VB150311-10018; Cyagen Biosciences). Vectors related to the viral capsid (8.91) and viral envelope (VSV-G) were used as accessory vectors for the production of lentiviruses by transfection of the 293FT cell line using Lipofectamine™ 2000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. The UW473 cell line was transduced with HOXB4-GFP-PUR and control lentiviral particles with a MOI of 2. After 24 h, the cells were treated with puromycin antibiotic (1 μg/ml; Sigma-Aldrich) for 6 days. Cell lines were identified as UW473/HOXB4+ and UW473/control. The cell transduction efficiency was evaluated by fluorescence microscopy and flow cytometry for GFP, and HOXB4 overexpression efficiency was evaluated by real-time PCR and western blot analysis.

Western blot analysis

Total protein was extracted using RIPA buffer (Sigma-Aldrich, St. Louis, MO, USA) supplemented with protease inhibitor (Complete Mini Protease Inhibitor Cocktail; Roche Diagnostics GmbH, Mannheim, Germany) and phosphatase (PhosSTOP phosphatase inhibitor cocktail; Roche Diagnostics), and quantified using the BCA protein assay kit (Pierce, Rockford, IL, USA) according to the manufacturer's protocol. Proteins were separated by SDS-PAGE using 10% polyacrylamide gels (Mini-PROTEAN TGX; Bio-Rad Laboratories, Hercules, CA, USA), transferred from the gel to a nitrocellulose membrane (40 μm, Hybond-C extra; Amersham Biosciences, Little Chalfont, UK) and incubated in a 5% skim milk blocking solution. The following primary antibodies were used: mouse anti-human HOXA10 (E-12) (1:250, sc-271139; Santa Cruz Biotechnology), mouse anti-human HOXB4 (D-1) (1:500, sc-365927; Santa Cruz Biotechnology), rabbit anti-human β-actin (1:4,000, #4967; Cell Signaling Technology, Danvers, MA, USA) and rabbit anti-human GAPDH (14C10) (1:4,000, #2118; Cell Signaling Technology). Mouse anti-sheep IgG (1:4,000, NA931V; GE Healthcare, Aurora, OH, USA) and goat anti-mouse IgG rabbit (1:4000, #7074; Cell Signaling Technology) HRP-conjugated secondary antibodies were used. The proteins bands on the membrane were developed using ECL Western Blotting Detection Reagent Prime kit (GE Healthcare) according to the manufacturer's instructions. As mentioned by Ferguson et al (43) β-actin and β-tubulin proteins (both being part of the cytoskeleton of eukaryotic cells), and GAPDH enzyme (glyceraldehyde-3-phosphate dehydrogenase), are classically used as internal standards for normalization of signals in western blot analysis in order to compare differences in the expression of the target proteins among samples and eliminating variations arising from technical reasons, such as differences in amount of total load protein (43). The protocols have been previously used by various co-authors of this study, as described in Thomé et al (44) (β-actin) and Palma et al (45) who have used both these control antibodies. As in this study, in Palma et al (45) both control antibodies, β-tubulin and GAPDH, were used.

In vitro proliferative and migration potential assay

The proliferative potential was assessed by cell counting in a Neubauer chamber. Initially, 8×104 viable cells were cultured in 75-cm2 flask for a period of 120 h. Next, the cells were trypsinized and viable cells were counted in the Neubauer Chamber with trypan blue in quadruplicate. The proliferative potential (PP) was calculated as PP = N - No, in which N is the final number of cells and No is the initial number of cells.

Scratch assay was used to evaluate the migration potential. Cells were cultured in 6-well plates (in triplicate for each cell line), and when they reached ~95% confluence a scratch was made. The wells were photodocumented by phase contrast microscopy (Olympus IX71 and PD controller software, USA) in two different points along the scratch in each well, and identified as time 0. After 16 or 20 h, the wells were again photodocumented at the same positions using the TScratch software (software developed by Tobias Gebäck and Schulz, ETH Zurich, Switzerland). The percentage of scratch area was measured in the images of initial and final time. The migration potential was calculated as percentage of migrated area (%MA), %MA = 1- (N/No), in which N is the final area (%) and No is the initial area (%).

In vivo tumorigenic potential assay

Male nude mice weighing 20–25 g were obtained from the Central Animal House of the University of São Paulo (Ribeirão Preto, Brazil), and were housed at 23±2C with 12-h light/dark cycle. This experimental protocol was approved by the Animal Research Ethical Committee of the Medical School of Ribeirão Preto, USP (protocol number 137/2011).

For the evaluation of the tumorigenic potential, 3×106 cells of medulloblastoma cell lines in 100 μl of Matrigel (Matrigel® Matrix; Corning, Inc., Corning, NY, USA) were infused subcutaneously into the backs of nude mice, which were previously anesthetized with a mixture of 2% isoflurane and oxygen. The mice were monitored in the periods 7, 15, 30 and 60 days after cell infusion, for the detection and quantification of GFP fluorescence using the in vivo imaging system IVIS Lumina and Living Image software (Perkin-Elmer, Waltham, MA, USA); and for measurement of the nodular Volume using a digital caliper. The nodular volume (NV) was calculated as: NV= L × W × H × 4/3 π, where L is the length, W is the width and H is the nodule height.

Statistical analysis

The Student's t-test and the one-way ANOVA test were used to evaluate the HOX expression level differences, as well as migration and proliferation potential, before and after genetic modification of the MB cell lines. For these assays, statistical tests were performed using the GraphPad Prism 5 for Mac OS X. All results were expressed as mean ± standard deviation. R was used for all other statistical analyses. Tests were declared statistically significant for P<0.05.

Results

Higher numbers of HOX genes are significantly deregulated in DAOY and ONS-76 cell lines compared to UW473 and UW402 cell lines

In Fig. 1 we show the results of hierarchical clustering of 39 HOX genes with tumor cell lines which are not characterized (UW473 and UW402 cell lines) and primary cell cultures from the cerebellum (CP4, CP5 and CP6), and also 39 HOX genes with tumor cell lines which are characterized as being from the SHH subgroup (ON-S76 and DAOY cell lines) (Fig. 1A and B) (35,46,47).

In both dendograms tumor cell lines and primary cell cultures fall into two distinct groups. For the DAOY and ONS-76 cell lines this distinction is even clearer producing two distinct clusters (Fig. 1B). We observed that in DAOY and ONS-76 cell lines several HOX genes are expressed, while these genes are not expressed in primary cell cultures. As regards the other two cell lines, we observed a distinction between HOX gene profiles in UW473 and UW402 cell lines. HOX genes are more expressed in UW402 cell line compared with UW473 cell line (Fig. 1A).

To visualize the significantly dysregulated HOX genes we constructed the volcano plot. We observed that in the UW402 and UW473 cell lines only 6 HOX genes are significantly upregulated (HOXB9, HOXB13, HOXC6, HOXC13, HOXD10 and HOXD11) compared with the primary cell lines from cerebellum (Fig. 1C). On the other hand, in the DAOY and ONS-76 cell lines 24 HOX genes are significantly upregulated: nine from HOXA group (HOXA2, HOXA3, HOXA4, HOXA5, HOXA6, HOXA7, HOXA9, HOXA10 and HOXA13), six from HOXB group (HOXB2, HOXB4, HOXB5, HOXB6, HOXB7 and HOXB9), six from HOXC group (HOXC4, HOXC5, HOXC6, HOXC9, HOXC10 and HOXC13) and three from HOXD group (HOXD3, HOXD10 and HOXD11) (Fig. 1D). Out of these 24 HOX genes, HOXA3 was the most dysregulated gene, however, upregulation of this gene has not been demonstrated in medulloblastoma or other tumor from the central nervous systems (CNS). These results are summarized in Table I, which contains all log fold change values. Gene names in bold face are those mentioned above as being statistically significantly upregulated.

Table I Comparison of fold change using microarray data set from HOX genes in UW402 and UW473 vs. cerebellum primary cells and DAOY and ONS-76 vs. cerebellum primary cells.

Table I

Comparison of fold change using microarray data set from HOX genes in UW402 and UW473 vs. cerebellum primary cells and DAOY and ONS-76 vs. cerebellum primary cells.

UW473 and UW402		DAOY and ONS-76
Genes	Log fold change	Genes	Log fold change
HOXA1	−0.569958863	HOXA1	−0.597828379
HOXA2	0.095698182	HOXA2	2.656257452
HOXA3	2.500799762	HOXA3	5.535882266
HOXA4	0.803518199	HOXA4	3.429616377
HOXA5	1.091665053	HOXA5	2.184209412
HOXA6	1.611731106	HOXA6	2.779732951
HOXA7	1.371254281	HOXA7	2.825761681
HOXA9	−0.262664551	HOXA9	1.664946587
HOXA10	0.131334357	HOXA10	1.866723269
HOXA11	0.143180662	HOXA11	1.846451081
HOXA13	2.670226837	HOXA13	3.755042843
HOXB1	0.946528446	HOXB1	0.324546332
HOXB2	−0.182254131	HOXB2	3.150953293
HOXB3	2.293320983	HOXB3	1.472715878
HOXB4	0.483773574	HOXB4	4.027400665
HOXB5	0.726902915	HOXB5	4.049698398
HOXB6	2.769873572	HOXB6	3.992872898
HOXB7	1.744203467	HOXB7	2.169081951
HOXB8	0.213373435	HOXB8	2.286976211
HOXB9	3.155648057	HOXB9	4.359008666
HOXB13	4.43490889	HOXB13	2.78047806
HOXC4	2.657559554	HOXC4	4.752524627
HOXC5	0.994376614	HOXC5	1.697798782
HOXC6	4.164237557	HOXC6	3.346096371
HOXC8	1.042329918	HOXC8	1.991863755
HOXC9	2.852715702	HOXC9	4.663234774
HOXC10	2.095372335	HOXC10	2.858686398
HOXC11	−0.378263228	HOXC11	−0.496091564
HOXC12	0.317817366	HOXC12	1.876619216
HOXC13	4.232848464	HOXC13	3.531212456
HOXD1	0.307750142	HOXD1	1.833684289
HOXD3	0.779984978	HOXD3	1.743693697
HOXD4	0.924886063	HOXD4	1.133382434
HOXD8	1.49312174	HOXD8	2.089211292
HOXD9	1.620166724	HOXD9	1.288963756
HOXD10	3.44561321	HOXD10	4.045820474
HOXD11	2.529309801	HOXD11	2.805492204
HOXD12	0.064926757	HOXD12	0.579149569
HOXD13	1.71540358	HOXD13	3.112325861

From the results of these analyses, we continued the present study with focus on two genes HOXA10 and HOXB4, which are significantly upregulated in DAOY and ONS-76 cell lines but not in UW402 and UW473. In mice MB-derived cancer stem cell (CSC), HOXA10 has been found to be one of the top-ranking genes highly expressed in highly tumorigenic (HT)-CSCs. Also, HOXA10 was found to have a high protein expression level in glioblastoma-derived neurospheres, and is associated with poor outcomes and treatment resistance in glioblastoma multiforme (GBM) (48). Moreover, the HOXA10 locus was associated with expression of a stem cell related HOX-signature in glioblastoma and with temozolomide resistance (49). HOXB4 was found to be expressed in medulloblastoma tissues (31) and recently, loss of Ezh2, an epigenetic regulator of HOX genes, drives aggressive medulloblastoma and was accompanied by upregulation of HOXB4 (32).

We validated the microarray data of HOXA10 and HOXB4 genes and compared gene expression levels by RT-qPCR in these four MB cell lines and control cells. As seen from Fig. 2A and B, HOXA10 and HOXB4 are upregulated in ONS-76 and DAOY, consistent with the results from the microarray analysis. The analysis showed that HOXA10 and HOXB4 transcripts are not expressed in primary cerebellum cells. Previous studies by Takahashi et al (15) and Hutlet et al (16) have demonstrated the absence of HOXA10 expression in human and mouse cerebellum tissues. These authors also showed the expression of HOX genes belonging to the 8 paralogy groups (PG1-8), which includes HOXB4, but not HOXA10. Therefore, one possible explanation for the absence of the HOXB4 expression in our primary cerebellum culture could be due to the differences between tissues and isolated cells. In cell therapy studies differences in gene expression between tissues and isolated cells have been previously observed (40,50). In our analysis of cell lines, HOXA10 transcripts are not expressed in two of the MB cell lines, UW402 and UW473. In the ONS-76 and DAOY cell lines, the observed expression levels of HOXA10 are 502.4±16.07 and 1305±22.42 URE, respectively (Fig. 2A). As regards the HOXB4 transcripts, they are expressed at low level in the UW402 and UW473 cell lines 1.05±0.16 URE and 1.072±0.33 URE, respectively. In the ONS-76 and DAOY cell lines, the expression levels of HOXB4 transcripts are 178.4±15.85 URE and 192.5±1.97 URE, respectively (Fig. 2B).

To better understand the classification of these four MB cell lines, we examine the expression levels of 22 signature genes proposed by Northcott et al (6) for molecular characterization of MB subgroups (Shh, Wnt, group C and group D). As shown Fig. 2C, we observed that under our culture conditions, in the DAOY cell line two Shh markers (EYA1 and PDL) but also the Wnt marker (WIF1) and the group C marker (GABRA5) are expressed. In the ONS-76 cell line the Shh marker (HHIP) and the group C markers (MAB21L2, EYS and NPR3) are expressed. On the other hand, in the UW473 cell line, genes from the four subgroups [GAD1 (from WNT), EYA1 [from Shh], EYS and NPR3 (from group C)] and OAS1 and RBM24 (from group D) are expressed and in the UW402 cell line genes from four subgroups [TNC and GAD1 (from Wnt), SFRP1 (from Shh), IMPG2 (from group C) and EOMES (from group D)] (Fig. 2C) are expressed.

Next, we investigated whether or not the HOXA10 gene has a direct effect on the in vivo tumorigenic potential and on the in vitro proliferative and migration potential of DAOY cell line by knockdown of this transcript. We also investigated whether or not HOXB4 transcripts have a direct effect on the in vivo and the in vitro properties in UW473 cell line after overexpression of this gene.

HOXA10 knockdown increases the migration potential but does not affect the proliferative potential of the DAOY medulloblastoma cell line

The HOXA10 gene was silenced in the DAOY cell line by shHOXA10 and an empty vector was used as control. The transduction efficiency was evaluated by fluorescence microscopy and flow cytometry, which showed that >95% of DAOY/shHOXA10+ cells were GFP+ (Fig. 3A). The HOXA10 knockdown was confirmed; the HOXA10 mRNA levels were reduced by 78% in comparison with the DAOY-control cells (268.1±16.25 vs. 59.20±2.1 URE) (P=0.0002) (Fig. 3C). The HOXA10 protein levels were also reduced in DAOY/shHOXA10+ cells (Fig. 3D). Analyzing the migration potential, the DAOY/shHOXA10+ cells displayed showed a migrated area 37±8% bigger when compared to DAOY-control cells (P=0.0046) (Fig. 4A and B). However, no difference was observed in the proliferative potential in which DAOY/shHOXA10+ and DAOY/control cells showed a similar amount of viable cells after 120 h of cell culture (P=0.868) (Fig. 4C). This result suggests that the low expression of HOXA10 gene is associated with an increase in the in vitro migration potential; however, there is no relationship with the in vitro proliferative potential of the DAOY medulloblastoma cell line.

Figure 4 Effect of HOXA10 knockdown on in vitro proliferative and migration potential of the DAOY cells. The DAOY/shHOXA10+ cells showed higher migration potential and similar proliferative potential compared to DAOY-control cells. The cell migration potential was evaluated by scratch assay (three replicates were performed) and the result was expressed as percentage of migrated area after 20 h. (A) Representative image from a scratch assay experiment of DAOY-control and DAOY/shHOXA10+ cell lines at time 0 and 20 h after the scratch. Contrast phase microscopy was used. (B) Quantification of the percentage of a migrated area after a period of 20 h using the Tscratch software. Results shown as mean ± standard deviation. The Student's t-test was used with 0.05 significance level (P=0.0046). (C) Proliferative potential as measured by counting the number of viable cells in the Neubauer chamber after 120 h of cell culture. Three replicates were performed.

HOXB4 overexpression decreases the proliferative and migration potential of the UW473 medulloblastoma cell line

The HOXB4 gene was overexpressed in the UW473 cell line and the transduction efficiency was evaluated by fluorescence microscopy and flow cytometry, which showed that >93% of cells were GFP+ in UW473/HOXB4+ and UW473-control cells (Fig. 3B). We also compared gene expression levels of HOXB4 by qPCR in UW473-control cells and in UW473/HOXB4+ cells. We observed that in UW473-control cells HOXB4 mRNA is present at a very low level (1.05±0.69 URE). However, in UW493/HOXB4+ cells the expression level of HOXB4 is 33.735±2.971 URE (P=0.0077) (Fig. 3E and F). Assessing the migration potential, the UW473/HOXB4+ cell line showed a migrated area 32±15% smaller when compared to the UW473-control cells (P=0.033) (Fig. 5A and B). The UW473/HOXB4+ cells line also showed a proliferative potential 34±6% lower compared to the UW473-control cells (P=0.032; Fig. 5C). These results suggest that the high expression of the HOXB4 gene is associated with a decrease in the in vitro proliferative and migration potential of the UW473 medulloblastoma cell line.

Figure 5 Effect of HOXB4 overexpression in the in vitro proliferative and migration potential of the UW473 cells. The UW473/HOXB4+ cells presented lower proliferative and migration potential compared to the UW473-control cells. The migration potential was evaluated by scratch assay and the result was expressed as the percentage of migrated area after 16 h. The proliferative potential was evaluated by counting viable cells in the Neubauer chamber after 120 h of cell culture. Three replicates were performed. The Student's t-test was used for statistical test with 0.05 significance level. (A) Representative image from a scratch assay experiment of UW473-control and UW473/HOXB4+ cells at time 0 and 16 h after the scratch. (B) Quantification of the percentage of migrated area after a period of 16 h using the Tscratch software. Results shown as mean ± standard deviation, P=0.032. (C) Proliferative potential was measured by counting the number of viable cells in the Neubauer chamber after 120 h (P= 0.033).

HOXA10 gene silencing and HOXB4 gene overexpression do not affect the tumorigenic potential of medulloblastoma cell lines

The effect of HOXA10 gene inhibition and HOXB4 gene overexpression on the tumorigenic potential of medulloblastoma cell lines was evaluated. DAOY/shHOXA10+, UW473/HOXB4+, DAOY-control and UW473-control cells were injected subcutaneously in mice and the GFP fluorescence intensity and tumor volume were monitored for a period of 7, 15, 30 and 60 days. After 60 days, the GFP fluorescence intensity (Fig. 6A and B) and tumor volume (Fig. 6C) were similar in DAOY-control and DAOY/shHOXA10+ cell lines with both cell groups giving rise to tumor nodules. No formation of tumor nodules was observed in either UW473-control or UW473/HOXA10+ cell lines, GFP fluorescence intensity (Fig. 6D and E) and tumor volume (Fig. 6F) were similar in both cell groups. These results demonstrate that the modulation of HOXA10 and HOXB4 gene expression did not change the tumorigenic potential of these medulloblastoma cell lines in nude mice.

Figure 6 Effect of HOXA10 knockdown and HOXB4 overexpression in the in vivo tumorigenic potential of medulloblastoma cell lines. There was no difference in tumor volume and GFP fluorescence intensity between tumor nodules generated from the injection of DAOY-control and DAOY/shHOXA10+ cells, and there was no formation of tumor nodules from the injection of UW473-control and UW473/HOXB4+ cells after 60 days of the subcutaneous cell injection in mice (n=3 in each DAOY group and n=5 in each UW473 group). The mice were monitored for a period of 7, 15, 30 and 60 days. (A and D) Representative images of the GFP fluorescence intensity emitted by GFP+ cells using the IVIS Lumina equipment. (B and E) Quantification of the GFP fluorescence intensity using the Living Image software. (C and F) Quantification of the nodular volume using a digital caliper. The Student's t-test was used at each time-point separately with a significance level of 0.05.

Discussion

This study evaluated the differences in aberrant upregulated HOX gene expression among the UW402, UW473, DAOY and ONS-76 human medulloblastoma cell lines and investigated the role of two HOX genes (HOXA10 and HOXB4) in the in vitro proliferative and migration potential and in vivo tumorigenic potential of these cell lines. Moreover, medulloblastoma is a heterogeneous disease and 22 transcripts have been used to stratify MB subgroups (6). Therefore, in order to understand these differences according to MB subtypes we compared the 22 transcript signatures in these MB-derived cell lines.

This study shows that different groups of HOX genes are differentially expressed according to MB-derived cell line. We identify 6 HOX genes (HOXB9, HOXB13, HOXC6, HOXC13, HOXD10 and HOXD11) whose expression levels are significant upregulated between UW402 and UW473 cell lines and normal cerebellum primary cells; and 24 HOX genes (HOXA2, HOXA3, HOXA4, HOXA5, HOXA6, HOXA7, HOXA9, HOXA10, HOXA13, HOXB2, HOXB4, HOXB5, HOXB6, HOXB7, HOXB9, HOXC4, HOXC5, HOXC6, HOXC9, HOXC10, HOXC13, HOXD3, HOXD10 and HOXD11) whose expression levels are significantly different between DAOY and ONS-76 cell lines and normal cerebellum primary cells.

To the best of our knowledge this is the first study that analyzes the expression pattern of all 39 HOX genes in MB-derived cell lines. HOX mRNA expression has been detected in MB tissues, Chakravadhanula et al (30) examined 10 MB tissues and studied the expression levels of 21 HOX genes. They showed that 10 HOX genes (HOXC4-6, HOXC8-10, HOXD3-4, HOXD8 and HOXD10) were significantly upregulated in medulloblastoma tissues when compared to control tissues. Also, Bodey et al (31) observed by immunocytochemistry techniques that HOXB3, HOXB4 and HOXC6 proteins were expressed in medulloblastoma tissues. In this study some of these HOX genes are also significantly upregulated. Two of them (HOXC6 and HOXD10) are significantly upregulated in these four MB cell lines (UW473, UW402, DAOY and ONS-76) and another five (HOXC4-5, HOXC9-10 and HOXD3) were identified as significantly upregulated in DAOY and ONS-76 cells.

The noticeable change in expression levels of most HOX genes in DAOY and ONS-76 cell lines compared with UW402 and UW473 merits attention. In order to determine whether differences in HOX deregulation are related to MB subtype we used the 22-transcript predictor in these 4 MB cell lines. We observed that UW473 and UW402 cell lines express genes from all four MB groups, and DAOY and ONS-76 previously classified as SHH subtype (38) also express genes from the other MB subtypes. DAOY expresses SHH, WNT and group C markers and ONS-76 expresses SHH and group C markers. These results suggest that under our cell culture conditions these cell lines cannot be classified as belonging to a unique molecular subtype and therefore the interesting differences in HOX pattern expression among these four MB cell lines cannot be directly associated with MB subtype. A similar effect has been observed in the D238-MB cell line, which has been classified as group 4 or group 3 (51,52).

Among the HOX genes significantly overexpressed in DAOY and ONS-76 cell lines compared to the UW402 and UW473 cell lines we selected HOXA10 and HOXB4 as mentioned earlier. In mouse medulloblastoma (MB)-derived cancer stem cell (CSC) the role of miR-135a, which binds to HOXA10 3'UTR in other cancers, was investigated (53,54). Hemmesi et al (54) showed that induced overexpression of miR-135a in mouse MB-derived CSC was responsible for a significant decrease in the expression of HOXA10 at mRNA level, but not at the protein level. Moreover, considering that miR-135a restoration inhibits tumor progression in CSC-derived MBs with no change in HOXA10 protein level, which suggests that in MB-CSCs HOXA10 can contribute to the malignant properties of MB cells depending on genetic background. HOXA10 is also one of the HOX genes, which shows a markedly higher expression in glioblastoma-derived neurospheres and was associated with worse outcome in patients assigned to TMZ/RT-TMZ therapy (48). From our initial analysis which showed that HOXA10 is differentially overexpressed in DAOY and ONS-76 cell lines but not in UW402 and UW473 cell lines, we decided to investigate whether knockdown of HOXA10 in DAOY cell line will result in changes in vitro migration and proliferation potential changes and in vivo tumor growth. We observe that knockdown of HOXA10 does in fact lead to an increase in the in vitro migration potential of the DAOY medulloblastoma cell line, however, there was no relationship between the HOXA10 expression level and the in vitro proliferative and the in vivo tumorigenic potential of that cell line. Altogether, although the exact mechanism that accounts for these effects in MB is unknown, these data might indicate that HOXA10 could contain different epigenetic regulators and has a role in several physiological processes.

Regarding the HOXB4 gene, we investigated whether or not overexpression of HOXB4 in the UW473 cell line promotes proliferation, migration and in vivo MB tumorigenesis. In the present study, we showed that the HOXB4 expression level is associated with a decrease in the in vitro proliferative and migration potential of the UW473/HOXB4+ cell line, however, it did not influence the in vivo tumorigenic potential of that cell line. It was previously known that HOXB4 is expressed in MB tissues (31) but the effect of this gene in medulloblastogenesis was not investigated. A recent study by Vo et al (32) measured RNA levels of HOXB4 gene in MB cells after gene-editing systems to induce loss-of-function mutation in Ezh2. In these Ezh2-mutated cells, HOXB4 was expressed more highly and the tumor became more aggressive. Ezh2 is an epigenetic regulator of HOX expression as a member of polycomb-repressive complex 2 (PRC2) together with EED, SUZ12 and other accessory proteins. Altogether, these results suggest that in UW473/HOXB4+ cells additional changes at the genetic or epigenetic levels are required to inactivate essential genes necessary for normal cellular development and function and to induce tumorigenesis.

To conclude, the present study was motivated by an earlier study of the importance of MLL2 in MB and the observation that MLL2 is an epigenetic regulator of the HOX gene. From the results of the microarray analysis presented (which were validated by RT-qPCR), we selected two HOX genes for functional analysis in vitro and in a mouse model. The in vitro analysis showed that these genes could be involved in migration and proliferative potential. The in vivo study with mice was not conclusive, which may be due to a variety of different causes. Other members of the HOX gene family could be involved. Another possibility is that MLL2 deletions have an effect on MB, but via other genes for example the MB oncogene named OTX2 or β-catenin known to regulate Wnt pathways, and which is implicated in one of the medulloblastoma subtype. Studies similar to the one we have undertaken should be extended to other genes and could help identify potential drug targets.

Acknowledgments

We thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) (Process no. 310619/2012-2) and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) (Process nos. 2011/20829-4 and 2011/18664-7) for the financial support for the present study. We also thank Sandra Navarro Bresciani of the Regional Blood Center of Ribeirão Preto-FMRP-USP, Brazil for help with the figures.

References