Decitabine is a cytotoxic agent in addition to being a demethylation agent. U2OS growth rate was measured in the absence, or presence of 1 μM decitabine. This relatively low-dose was chosen to minimize cytotoxicity. The doubling time for U2OS in the absence of 1 μM decitabine was close to 2.75 days (66 hours) while the decitabine treated U2OS cells doubled at 3.5 days (84 hours). This was accompanied by a minor reduction in viable cell numbers as a result of decitabine treatment (Figure 1A). To confirm that decitabine was also causing DNA demethylation of specific loci we examined the extent of loss of the methylation known to characterize the small nuclear ribonucleoprotein polypeptide N (SNRPN) gene. This heavily methylated gene has previously been shown to undergo reduction in methylation following treatment with decitabine 16. After 72 hours treatment of U2OS with 1 μM decitabine, ~60 % more DNA could be digested with a methylation-sensitive restriction endonuclease [see Additional file 1]. To determine whether a slower growth rate was accompanied by higher level of cell death following decitabine treatment, death assessment was studied by flow cytometry after PI staining. The control (no-treatment) cells showed a 5% death rate whereas the decitabine treatment significantly increased the death rate of 10% more than the control at day 3 (p < 0.05) (Figure 1B).

U2OS has been shown to be generally non-tumorigenic using heterotopic grafts under the mouse skin 17. In contrast the sub-renal capsule site has proved to be an excellent grafting site for xenografts in general 181920. Therefore U2OS xenografts were obtained from mice and cut into pieces of ~4 mm³ and 24 tumor tissue pieces were regrafted under the renal capsules of 6 SCID mice (4 grafts per mouse). Four weeks after grafting the host mice were divided into two groups. One group of three mice was treated with Decitabine (2.5 mg/kg) intraperitoneally on Days 29, 31 and 33. The other group of three mice was given saline as control at the same schedule. On Day 37, all 6 mice were sacrificed and xenografts were dissected from the mice, and the tumor volumes were measured and compared in both treatment arms. The effectiveness of demethylation in vivo was determined by analyzing the relative cellular distribution of 5-methylcytidine levels using a specific 5-methylcytidine antibody and 6 tissue sections from 6 xenografts (3 control and 3 decitabine treated). In these analyses, nuclei in host kidney stromal cells were used as internal positive controls (see dark staining nuclei Figure 2). Negative controls obtained by omitting the primary antibody were also performed. Nuclear xenograft tumor staining was recorded by a semiquantitative and visual grading, considering both the intensity of staining and the proportion of positive tumor nuclei in the selected sections. Xenograft derived from the untreated control mice exhibited stronger nuclear staining than the staining present in nuclei from xenografts in decitabine treated mice (p < 0.05). These results were consistent with a widespread loss of methylation in U2OS xenografts derived from the treatment group.

Tumor volumes were determined for all 24 xenografts [see Additional file 2]. The average volume size for the xenografted tumors in the control group was 49 mm³ (± 25SD). The xenograft tumors from the decitabine treated group had an average volume size of 27 mm³ (± 15.9SD) indicating a significant decrease (p = 0.0096) of tumor volume as a result of the decitabine treatment (Figure 3A). All U2OS xenografts were whitish in color and the texture of tissues from both treatment groups had a moderate-hard consistency, with no apparent kidney tissue invasion.

Comparison between the U2OS xenograft histology from the decitabine-treated mice and the control mice identified differences in bone matrix (osteoid) content. In these studies relative levels of osteoid were determined in 9 sections each from the control and decitabine treatment groups. As shown in Figure 3B, histological analysis of the stained xenograft sections identified an overall average osteoid area of 42% in the control group. The decitabine treated group, however, had a significant increase (p < 0.0001) in osteoid formation reaching up to an average of 86%. With respect to tissue morphology, the tumors from the control mice showed solid sheets of poorly differentiated cells associated with a small amount of osteoid (Figure 3C left panel). In contrast, the xenograft tumors from the decitabine-treated mice showed a less dense cell population with increased areas of osteoid, seen as light pink and lacy matrix with the nuclei of osteoblasts sitting close to the matrix, a feature of differentiated osteoblasts in normal bone (Figure 3C right panel). The mitotic count was assessed in the same sections (Figure 3D). The mitotic index showed an average of 11.9 mitotic figures in the control group and an average of 3.1 mitotic figures in the decitabine treated group (p < 0.0001).

Since decitabine induced a higher level of cell death in vitro and there was a marked decrease in the size of U2OS derived tumors in vivo, the level of apoptotic cells in the xenograft tumors was analyzed using the TUNEL assay. Scoring was performed on digital images from 9 control sections and 9 decitabine treatment sections. The average apoptotic indices were 3.2% in the control group and 6.2% in the decitabine treatment group (p = 0.0329) (Figure 3E and 3F).

Expression profiling of 14,500 transcripts with known gene function was performed using the Affymetrix U133A microarrays following treatment of U2OS cells with 1 μM decitabine. Eighty-eight (88) genes [see Additional file 3] exhibited significant (p < 0.0025) up-regulation after two independent treatments of U2OS with 1 μM decitabine. Within this large group of 88 genes there were 13 with a ≥2-fold change in both experiments (Table 1). The expression of subset of 7 of these robustly reactivated genes was examined by real-time PCR and increased gene expression was consistent with the levels detected by microarray analysis (Table 1).

An in silico analysis of CpG islands associated with the 88 up-regulated genes was then performed. 63 genes (71%) were found to contain CpG-islands within their 5'-regulatory region. Of the 13 genes with a ≥2fold-change, 11 genes (84%) had a CpG-island within their 5' region (Table 1). Pathway enrichment analysis was performed on the 88 up-regulated genes, and the three top enriched pathways included: "negative regulation of cellular processes" (p = 0.007), "positive regulation of programmed cell death" (p = 0.01) and, "organelle organization and biogenesis" (p = 0.03). Interestingly, when the 13 genes with a ≥2-fold change were analyzed in the same manner, six genes were assigned to the "apoptosis pathway" with high significance (p < 0.0001). These genes include GADD45A, HSPA9B, PAWR, PDCD5, NFKBIA, and TNFAIP3, which were selected for expression validation in the xenografts and in normal human osteoblasts (NHOst).

Real-time PCR (TaqMan method) was used to validate expression of the apoptotic genes shown in Table 1 in the xenografts and NHOst. As illustrated in Figure 4, the change of expression was expressed as fold change relative to the control (no treatment), using beta-actin (ACTB) as a reference. Up-regulation, as a result of decitabine treatment was consistent in U2OS cells in vitro and in vivo. Gene expression of the apoptotic genes was up-regulated up to 8-fold for GADD45A, 8-fold for HSPA9B, 12-fold for PAWR, 8-fold for PDCD5, 9-fold for NFKBIA, and 5-fold for TNFAIP3. In addition, real-time expression for the six genes was examined in normal osteoblasts before and after treatment with decitabine. There was marginal increase of expression after treatment (<2-fold change) in normal osteoblast (Figure 4).

Methylation Pyro Q-CpG sequencing analysis was performed on six genes with expression up-regulation after decitabine treatment including GADD45A, HSPA9B, PAWR, PDCD5, NFKBIA, and TNFAIP3. GADD45A, HSPA9B, PAWR, and PDCD5, but not NFKBIA, and TNFAIP3, showed reliable and reproducible results for the tested amplicons. GADD45A, PAWR, and PDCD5 had a high methylation content (>70%) without decitabine treatment, while HSPA9B had an intermediate (50% – 70%) methylation content. Decitabine treatment lowered the methylation content in all CpG sites tested for the four genes in vitro and in vivo (Figure 5). Induction of demethylation was most marked for PAWR, and PDCD5. GADD45A had intermediate loss of methylation and HSPA9B had the least change but overall methylation loss remained significant. Analysis of relative change in methylation was also performed on normal osteoblasts, which had an initial low methylation content comparable to the negative control (DNA from early embryos), and no change in this basal content was apparent following decitabine treatment.

More detailed results on the methylation status for the same four genes for all tested CpG positions is provided in Figure 6 [also see Additional file 4]. For GADD45A, a CpG rich sequence at the promoter region containing a cluster of eight CpG positions was tested by Pyro Q-CpG. In U2OS cells, the methylation percentage on the eight CpG positions had an average of 83% in the no-treatment (control) cells. This was reduced to an average of 44% after treatment with 1 μM decitabine (p < 0.001). Similarly, the no-treatment xenograft tumors had averages of 84% in Xeno-1, 81% in Xeno-2, and 85% in Xeno-3, which was reduced after three doses of 2.5 mg/kg decitabine to 46%, 43% and 37% in Xeno-4, Xeno-5 and Xeno-6 respectively (p < 0.001).

A CpG rich sequence at the promoter region of HSPA9B had a cluster of 13 CpG positions which were tested by Pyro Q-CpG. The overall methylation quantity seen in this sequence was lower than what was observed in GADD45A, none-the-less the difference between the no-treatment (control) samples and decitabine-treated samples was significant (p < 0.001). U2OS cells (in vitro) had an average of 59% before treatment, and 42% after treatment (p < 0.001). Before treatment, the xenograft tumors had averages of 61% in Xeno-1, Xeno-2, and Xeno-3 which was reduced after treatment to an average of 49% in Xeno-4 and Xeno-5 and 43% in Xeno-6 (p < 0.001).

The analysis was done in the same manner for 13 and 12 CpG positions related to the CpG-island associated with PAWR and PDCD5 respectively. These two genes had very high methylation percentage before decitabine treatment in U2OS, Xeno-1, 2, and 3 for the tested CpG positions. In the case of the 13 CpG positions tested in PAWR the average of methylation percentage before decitabine treatment was 93% in U2OS cells, 93% in Xeno-1 and Xeno-2, and 94% in Xeno-3. After decitabine treatment, this was reduced significantly (p < 0.001) to an average of 26% in U2OS cells, 29% in Xeno-4 and 28% in Xeno-5 and 6. In the 12 CpG positions tested in PDCD5, U2OS cells had an average of 93%, Xeno-1 had an average of 94, Xeno-2 had an average of 93%, and Xeno-3 had an average of 92% before decitabine treatment. The methylation was reduced significantly after decitabine treatment (p < 0.001) to 26% in U2OS cells, Xeno-4, and Xeno-5, and 27% in Xeno-6.

The normal low-passage osteoblast had a very low methylation percentage in all 4 genes. The range of methylation percentage had averages from 2.1% to 3.6% before treatment, and 1.4% to 2.7% after treatment with 1 μM decitabine across all the tested CpG positions in the 4 genes (Figure 5, Figure 6 and [Additional file 4]). Importantly, the methylation status of the tested CpG sequences reflects the patterns of expression seen in all four genes after decitabine treatment in U2OS cells and xenografts. This is consistent with the possibility that decitabine treatment modulated the expression through reducing the amount of methylation on CpG-dinucleotides. To determine whether decitabine activated GADD45A methylation status and expression was also tested in two other OS cell lines, MG63 and HOS using identical decitabine treatment conditions. Similar to U2OS, induction of GADD45A gene expression was associated with significant loss of methylation and increased transcript expression in MG63 cells. However in HOS cells, treatment had no significant demethylation effect on the GADD45A promoter regions and also failed to activate expression of the gene (data not shown).

The human OS cell line U2OS was obtained from the American Type Culture Collection (ATCC) (Rockville, MD) and maintained in alpha-Minimum Essential Medium (alpha-MEM) supplemented with 10% heat inactivated Fetal Bovine Serum and 2 mM L-Glutamine. Treatment with decitabine was performed as described by Liang et al 39. Briefly, 5 × 10⁵ cells were plated in 56 cm² culture plates with 10 ml growth medium. 12 hours after plating they were treated with freshly prepared decitabine (Sigma Chemical Co., St Louis, MO) to a final concentration of 1 μM without changing the medium. The cells were harvested by trypsinization after 3 days of treatment, where the cells were portioned and used for total RNA extraction, DNA extraction, or Propidium Iodide (PI) staining for cell death by flow cytometery. A control (medium only) culture was maintained and processed over the same period of time under the same condition as the treated cells. To establish growth curves for U2OS cells with or without 1 μM decitabine, cells were plated at 5 × 10⁵ cells/56 cm² culture plates with 4 mm² grids. The cells were allowed to attach to the surface of the plates for 12 hours before the start of the treatment. Adherent cells were counted in 2 independent cultures in multiple 4 mm² grids every 12 hours after plating and the experiment was repeated after culturing the cells for 5 passages. When cell growth was near confluent, the cells were trypsinized, re-suspended in growth medium (10% serum) and cell viability was determined using Vi-CELL™ XR (Beckman Coulter, Fullerton, CA) after Trypan Blue staining.

Xenograft tissue sections were de-paraffinized using xylene and re-hydrated in a series of alcohols. The tissue sections were then incubated at room temperature (RT) in 3% H₂0₂ in PBS for 10 minutes to inactivate endogenous peroxidase. Following incubation the slides were washed 3 times in PBS for 3 minutes each. Antigen retrieval was obtained by heating in a microwave at maximum heat for 20 minutes in Tris-EDTA buffer (10 mM Tris Base, 1 mM EDTA solution, 0.05% Tween 20, pH 9.0) and cooling for 20 minutes at RT. Slides were again washed 3 times in PBS for 3 minutes each. The slides were blocked (30 minutes in a humid chamber at RT) with serum to reduce non-specific binding. Serum was removed from the slides and the slides were then incubated with the primary antibody 5-methylcytidine (5-mc-Ab) (Eurogentec, San Diego, CA) at 1:500 dilution at 4°C overnight.

Following incubation the slides were washed 3 times in PBS for 3 minutes each. The slides were then incubated with a secondary antibody [Polyclonal rabbit anti-mouse immunoglobulins/biotinylated rabbit F(ab')2; Dako] for 30 minutes at RT, followed by 3 washes in PBS for 3 minutes each. The slides were then incubated with StreptABCComplex/HRP (Dako, Glostrup, Denmark) for 30 minutes in a humid chamber at RT, followed by 3 washes in PBS for 3 minutes each. A 3,3'diaminobenzidine (DAB) substrate (Vector Laboratories, Burlingame, CA) was used for detection and hematoxylin was used for counterstain. The slides were then dehydrated and mounted.

Whole sections from xenografts were scanned by ScanScope CS (Aperio technologies, Vista, CA). The slides were digitized to 20× magnification (~0.5 microns/pixel). Images were then viewed with Aperio's image viewer software (ImageScope), which allows performing quantitative analysis of stain intensity on snapshots from the sections. Five to ten ~0.3 mm² snapshots (each containing 3,000 to 5,000 cells) were analyzed per section using the following parameters: compression quality = 30, and color saturation threshold = 0.04. Positivity thresholds were150 to 220 = high positive, 100 to 150 = low positive, and 0 to 100 = negative. Descriptive analysis such as mean and standard deviation for 5-mc immunostaining intensity were calculated based on the percentage of positivity (total positivity/total negativity per snapshot). Comparison between control and decitabine-treated sections was done using the student t-test and p < 0.05 was considered significant.

Six- to eight-week old male immune-deficient NOD-SCID and Rag-2M mice were bred and maintained by the Animal Resource Centre at the British Colombia Cancer agency, Vancouver, Canada. U2OS cells, in general, were considered as non-tumorigenic in mice while grafting the cells subcutaneously or orthotopically 17. As such, in order to establish their xenografts, U2OS cells were grafted under the renal capsule, a site proven to be an excellent site for tumor engraftment 181920. Briefly U2OS cells were cultured in alpha-MEM and washed in growth medium containing 20% FBS. The viable cells were counted after trypan blue staining. 2 × 10⁶ cells were pelleted, re-suspended and grafted beneath the renal capsule of adult male SCID mice as previously described 18. After 5 months, a well-grown xenograft was selected for re-grafting to establish multiple stable U2OS xenografts under the kidney capsules of NOD-SCID mice (two per kidney per mouse). The re-grafted U2OS xenograft had a 100% take rate, with a doubling time of ~10 days and exponential growth phase starting after ~2 of tumor growth doubling. After 5 generations of tumor growth doubling the xenografts were surgically removed from the mice and cut into approximately 4 mm³ portions then were re-grafted under the renal capsules of 6 male Rag-2M mice (4 grafts per mouse, 2 per kidney). Four weeks after grafting (~2–3 doubling of tumor growth), the host mice were divided into two groups. One group (3 mice) was given decitabine (2.5 mg/kg body weight) dissolved in saline (0.9% w/v NaCl), intraperitoneally 8 on days 29, 31 and 33. The other group (3 mice) was given saline alone as a treatment control over the same schedule. On day 37, mice from both groups were sacrificed. Tumor volumes were measured using a digital caliper, recorded and expressed in mm³, using the formula: volume (mm³) = (0.52) × length (mm) × width (mm) × height (mm). Data were presented as means ± Standard Deviation (SD) and student t-test was used to analyze the difference between the two treatment groups. The xenograft tissues were then snap frozen, or prepared in paraffin and sectioned according to standard procedures 18.

The tissues from the control (no treatment) and decitabine treated groups were paraffin embedded using routine protocols 18 and stained with hematoxylin and eosin. Sections were assessed blindly. Extracellular matrix was defined as eosinophilic osteoid-like material surrounding individual cells and small clusters of 3–5 cells, and the percentage of tumor with osteoid was then calculated. Mitotic counts were performed in areas with the highest mitotic rate, and ten high-powered fields (× 400) were counted per section.

In situ hybridization for terminal deoxynucleotidyl transferase-mediated nick end labeling (TUNEL) was performed on paraffin sections as recommended by the manufacturer (Ventana Medical Systems, Tucson, AZ). Scoring of the sections was performed using Simple PCI analytical software (Nikon, Tokyo, Japan). Sections were examined and the most intense areas of staining were photographed using a DXM1200 digital camera (Nikon) at a power of x200. The digital image was then scanned using the Simple PCI program and the numbers of positive and negative nuclei were obtained. Control and treatment images were all photographed at a uniform brightness, and all images were subjected to uniform binary image modification and size calibration prior to counting by Simple PCI. The positivity index was obtained by dividing the number of positive nuclei by the total number of nuclei (positive + negative). The number of nuclei counted was always over 1000, and ranged from 1100 to 2000. Positivity indices were compared by Student's t-test and a p-value of <0.05 was considered significant.

Total RNA was extracted using the RNeasy kit (Qiagen, Germany) from duplicate experiments of U2OS cells at day 3 after treatment with 1 μM decitabine or medium alone (control). In each experiment RNA yields were pooled from two independent cultures per treatment arm to minimize experimental noise. For each case, 10 μg of RNA was labeled and hybridized to the Affymetrix HG-U133A GeneChips using the manufacturer's protocol (Affymetrix, Santa Clara, CA) by the Centre of Applied Genomics at the Hospital for Sick Children (Toronto, Canada). Data were extracted using the Microarray Suite (MAS) version 5.0 (Affymetrix) and linearly scaled to achieve an average intensity of 150 across each chip. The candidate gene list obtained from the MAS 5.0-extracted data was selected by eliminating genes that were not present in at least one experiment. The arrays were subjected to a pair wise comparison using MAS 5.0, with signal intensities from the no-treatment cells as the baseline. The statistical significance for the change of expression for each probe set between the decitabine treated and control was calculated by the MAS 5.0 software. The criteria for gene selection for real-time expression validation analysis was based on the statistically significant up-regulation (p < 0.0025) and fold change of ≥2 for expression after decitabine treatment. The gene list was annotated based on the NetAffx data-base 58 and further verified using the Human Genome Browser data base 59. All the raw data for expression arrays is available in 60 under the series record number (GSE7454).

The criteria for a CpG-island was based on those outlined by Takai and Jones 61, where the GC ≥ 55%, Obs/Exp ≥ 0.65, and length > 300 bp which was reported to exclude most Alu-repetitive elements. We identified the genes that harbored CpG-island within a 2000 bp window upstream or downstream from the transcription start site based Human Genome Browser data base 59. To be certain that there were no CpG island closer to the TSS and gene promoter regions, we submitted the sequences of interest (including a 2000 bp window upstream and downstream from TSS) to the CpG search engine available in reference 61 and verified that there was no CpG islands that are closer to TSS for the genes we tested. Up-regulated genes with CpG-island associations were further analyzed through the Microarray Literature-based Annotation tool MILANO 62 to look for evidence of epigenetic modifications in the literature. MILANO is a web-based tool that allows annotation of lists of genes derived from microarray results by user defined terms 62. Using MILANO we searched for literature associations between our list of genes and the terms 'epigenetics', 'methylation' 'chromatin modification' 'cancer', and 'disease'. To identify the putative functional pathways for each gene list, we used the functional annotation enrichment tool. This tool utilizes the Gene Ontology database and uses GO Terms to identify enriched biological themes in the gene lists 6363. The Fisher Exact test was applied to determine the significance in the proportions of genes falling into a certain pathway in each gene list. We used this tool to look for enriched pathways of up- or down- regulated genes with CpG-island associations from the gene lists from the cell lines.

Total RNA from xenografts was extracted using the TRIzol reagent method. 1 ml of TRIzol (Invitrogen, Osaka, Japan) was used for every 50–100 mg of tumor tissue and homogenized in an RNase free environment. Chloroform was then added (200 μl for each 1 ml TRIzol) and the samples were centrifuged at high speed for 15 minutes at 4°C. The aqueous layer was then transferred into a new tube and RNA was precipitated with iso-propanol followed by one wash using 70% ethanol. The RNA precipitate was then dissolved in 10–15 μl of RNAse free water and analyzed for quantity and quality using a spectrophotometer. A two-step reverse transcription-PCR procedure was performed. Total RNA was reverse transcribed using the GeneAmp kit (Applied Biosystems; ABI, Foster City, CA). 20 ng of the resulting cDNA was then used in the real-time PCR step. Six genes were tested by real-time PCR including: growth arrest and DNA-Damage inducible, alpha (GADD45A), heat chock70KDA protein 9b (HSPA9B), parkc apoptosis wt1-regulatort (PAWR), programmed cell death 5 gene (PDCD5), nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (NFKBIA), tumor necrosis factor, alpha-induced protein 3 (TNFAIP3). We used the TaqMan primers (ABI) for all the genes that we tested (primer information is provided the in [Additional file 5]). All real-time PCR assays were performed in triplicate in a 96-well plate using the 7900 Sequence Detector System (ABI) according to the manufacturer's protocol. Data analysis was performed using the Sequence Detector System (SDS) software (ABI) and the results were expressed as fold-change in relative mRNA expression level, calculated using the ΔΔCt method with β-actin (ACTB) as the reference gene and the non-treated cells as baseline. The validation was carried out on RNA from three replicate experiments of U2OS cells, three decitabine-treated U2OS xenograft tumors (Xeno- 1, 2 and 3), three no-treatment (control) U2OS xenograft tumors (Xeno- 4, 5, and 6) and three replicate experiments of NHOst.

Quantitative Bisulfite Pyrosequencing for CpG islands (Pyro Q-CpG) is a sequencing-based analysis of DNA methylation that quantifies multiple CpG sites per amplicon using Pyro Q-CpG software. 2 μg of DNA from the control and decitabine treatment were bisulfite-treated using the Zymo DNA Methylation Kit (Zymo Research, Orange, CA). Bisulfite-treated DNA was amplified by PCR then sequenced according to the manufacturer's protocol (Biotage, Kungsgatan, Sweden). The target sequences inside the CpG-islands of the candidate genes and the primer sequences are shown in [Additional file 5]. The percentage of C content (methylated alleles) versus T content (unmethylated alleles) is calculated by the Pyro-Q-CpG software for each CpG position in each sample. Analysis was performed on DNA samples from 3 replicate experiments of U2OS cells in vitro and six U2OS xenograft tumors; three decitabine treated (Xeno-1, Xeno-2 and Xeno-3), and three saline (control) treated (Xeno-4, Xeno-5 and Xeno-6). Universally methylated DNA was used as a methylation positive control. DNA isolated from early embryos (Biotage, Kungsgatan, Sweden) was used for methylation negative control. DNA from low-passage normal human osteoblasts (PromoCell, Germany) was used for experiment control.