Epigenetic regulation of CpG promoter methylation in invasive prostate cancer cells

LNCaP and DU145 human prostate cancer cell lines were obtained from ATCC and cultured accordingly (Manassas, VA). Primary human prostate cancer cells (PCSC1-2) [7] were acquired from Celprogen (San Pedro, CA) and maintained as recommended using specific coated culture plates and defined media. Human bone marrow derived mesenchymal stem cells (hMSCs) were obtained from Lonza (Gaithersburg, MD) and maintained using their recommended conditions. The cultures were maintained in 5% CO₂ air at 37°C. Human serum was obtained from Gemini Bioproducts (West Sacramento, CA). The following inhibitors were also used: Anti-human IL-6 antibody (R&D Systems, Minneapolis, MN), PI3K inhibitor LY294002 (Cell Signaling, Danvers, MA), Tec Kinase inhibitor LFM-A13 (Tocris, Ellisville, MO), MEK inhibitor PD98059 (Gibco, Carlsbad, CA), JAK inhibitor AG490 (Gibco), and STAT3 inhibitor Stattic (Sigma Aldrich, St. Louis, MO).

Matrigel-coated 24-well inserts (8 μM pore size) and non-coated control inserts purchased from BD Biosciences (Palo Alto, CA) were used according to manufacturer's instructions. A range of 20,000-100,000 cells were seeded for the invasion (higher for less invasive LNCaP cells). Cells were seeded in serum-free RPMI and migrated toward media specific for stem cells (SCM) containing DMEM/F12 with human supplementation of 10 ng/mL bFGF, 20 ng/mL EGF and 5 μg/mL insulin along with 0.4% BSA (each from Sigma, St. Louis, MO). Routine invasion assays were performed for 24 hours and then stained with the Diffi-Quick Staining kit (Dade Behring, Deerfield, IL). Three to five microscopic fields (20×) were photographed and counted for each sample. Percent invasion was calculated as average number of cells/field (Matrigel) divided by average number of cells/field (control insert). Values were averaged from 2-5 independent experiments. For the isolation of cells from top 'non-invading' and bottom 'invading' cells, parallel invasion chambers were setup. For non-invading cells, the bottom of the membrane was scrubbed with a cotton swab and cells on top were harvested using 500 μL of Accutase (eBioscience, San Diego, CA) incubated at 37°C for 5 minutes. To obtain the invading cells, the top of the membrane was scrubbed with a cotton swab and the chambers were placed into another 24- well plate containing 500 μL of Accutase incubated at 37°C for 5 minutes.

Matrigel invasion assays were carried out as previously described. For the isolation of DNA from both non-invasive and invasive cells the DNeasy kit from Qiagen (Valencia, CA) was used and parallel invasion chambers were setup. For non-invading cells, the bottom of the membrane was scrubbed with a cotton swab and cells on top were trypsinized and harvested in 200 μL of PBS followed by the direct addition of lysis buffer or stored at -80°C. For bottom 'invading cells' the top of the membrane was scrubbed with a cotton swab and the membrane was removed and placed directly into lysis buffer or stored at -80°C until needed. A modified version of Agilent's (Santa Clara, CA) protocol for Mammalian ChIP on ChIP was used to capture methylated DNA with immunoprecipitation (MeDIP). DNA was quantified and 2 μg (or total yield if less) was digested with MseI overnight at 37°C. Linkers (JW102-5'-GCGGTGACCCGG-GAGATCTGAATTC-3' and JW103-5'-TAGAATTCAGATC-3') were ligated at 16°C using T4 ligase overnight and the next day used as input for the MethylCollector (Active Motif, Carlsbard, CA) assay to isolate methylated and non-methylated fractions of DNA. The kit utilizes histidine-tagged MeBP2 (methyl-binding protein 2) and magnetic bead separation. The isolated methylated and non-methylated DNA from each sample (as little as 5 ng) was then amplified in a series of PCR reactions following the mammalian ChIP on ChIP protocol. The input DNA was labeled with Cy3-dUTP and the methylated DNA with Cy5-dUTP and then immediately applied to Agilent's 2 × 244 K Human Promoter Tiling Arrays for 40 hours at 65°C. The arrays were scanned using a Gene Pix 4000B scanner (Molecular Devices, Sunnyvale, CA) with GenePix Pro software version 6.1 and extracted using Agilent's Feature Extraction software version 9.5.3.1. The data was annotated using Agilent's ChIP Analytics software version 4.0. Normalization was carried out using a blank subtraction model and statistical stringency (p-value) between 0.01-0.05 was applied using a Whitehead Per-Array Neighbourhood Analysis. This analysis allowed for the determination of differentially methylated genes between non-invasive and invasive cells. Ingenuity core analysis was carried out to determine which pathways are of functional significance based on the gene lists identified http://​www.​ingenuity.​com/​. Genomatix software was used to determine transcription factor binding sites (matrix). A perfect match to the matrix gets a score of 1.00 (each sequence position corresponds to the highest conserved nucleotide at that position in the matrix), a "good" match to the matrix usually has a similarity of >0.80.

Mismatches in highly conserved positions of the matrix decrease the matrix similarity more than mismatches in less conserved regions.

A total of 1 μg of DNA extracted from total (parental) DU145 and LNCaP cells was bisulfite modified using the EpiTect Bisulfite kit from Qiagen. PCR was performed using Platinum Taq Polymerase (Invitrogen) and 200 ng of either genomic or bisulfite treated DNA. The PCR method utilized was 94°C for 2 minutes, then 35 cycles (94°C for 30 seconds, 55°C for 30 seconds and 72°C for 1 minute) with a final extension of 10 minutes at 72°C. The unmethylated primers however were run with an annealing temperature of 42°C since their melting temperature values were drastically different from their methylated counter part. A portion of the PCR product was run on a 1% agarose gel containing ethidum bromide.

hBMX-Forward 5'- TGGTGAGACATCATGTGTTCCATT-3';

hBMX-Reverse 5'- ATGCCCTCAGTTGAGAACCACTGT-3';

hSOX1-Forward 5'-ATGATCAGCATGTACTTGCCCGC-3';

hSOX1- Reverse 5'-TCCGCTTCCTCCGTAGGTGATAAA-3'

hBMX-Forward 5'- TGGTGAGATATTATGTGTTTTATT-3';

hBMX-Reverse 5'- ATGTTTTTAGTTGAGAATTATTGT-3';

hSOX1-Forward 5'-ATGATTAGTATGTATTTGTTTGT-3';

hSOX1-Reverse 5'-TTTGTTTTTTTTGTAGGTGATAAA-3'

Total RNA was isolated using TRIzol (Invitrogen Corporation, Carlsbad, CA). RNA from 'top' cells was isolated using a cell pellet acquired from trypsinizing cells from one membrane after bottom cells were removed with a cotton swab. Conversely, RNA from the bottom cells was isolated by combining three membranes where the top cells were removed using a cotton swab. The membranes were pooled and placed in TRIzol for 10 minutes at room temperature, and the conventional procedure for isolation of RNA was then followed. To increase the yield of RNA, 5 μg of linear acrylamide (Ambion, Austin, TX) was added prior to precipitation of RNA with isopropanol. Additionally to increase overall yield, 100 ng of RNA was amplified using the MessageAmp aRNA Amplification Kit (Ambion, Austin, TX). cDNA was prepared using the SuperScript^®III First-Strand Synthesis System (Invitrogen Corporation, Carlsbad, CA). Quantitative real time polymerase chain reaction (qRT-PCR) analysis was performed using a StepOne Real-time PCR machine (Applied Biosystems, Foster City, CA) with TaqMan Gene Expression Assay reagents and probes (Applied Biosystems). A total of 4 μL of cDNA was used in a 20 μL reaction resulting in a 1:5 dilution. The following FAM labeld human probes were used: BMX (Hs00174139_m1), IRX3 (Hs00273561_s1), SOX1 (Hs01023894_m1), MCL-1 (Hs00172036_m1), MYC (Hs00153408_m1), STAT3 (Hs01047580_m1), SURVIVIN (Hs00977611_g1) and 18S rRNA (Hs99999901_s1). Relative fold induction of mRNA was compared between non-invasive and invasive cells using the Delta-Delta CT method of quantitation, and 18S rRNA was used as a loading control.

The Trans-Lentiviral pTRIPZ system from Open Biosystems (Huntsville, AL) was used to introduce shRNA against BMX (Clone ID: V2THS_150067) and SOX1 (V2THS_197330) along with a non-silencing control vector. The vectors were transfected into HEK239T cells which were seeded in serum-free media at 60% confluency in 10 cm² dishes using the Arrest-In reagent provided in the kit. The cells were transfected for 6 hours and then replaced with complete media. After 24 and 48 hours lentiviral supernatants were harvested, spun at 1500 rpms, and filtered using a 0.45 μM filter to clear them. The viral titer was mixed 1:1 with DU145 media and placed on sub-confluent DU145 cells for 4-6 hours and changed to complete media. The next day media containing 1 μg/mL of doxycycline (Sigma) was added to ensure efficient transfection/infection has occurred. Efficient transfection was observed using a TET inducible TurboRFP (red fluorescent protein) upstream of the shRNA that appears red upon successful infection. The cells were selected for 2 weeks in 1 μg/mL of puromycin (Sigma). Single cell clones were then generated and lowered expression was confirmed using Western blotting.

Total cell lysates were prepared using RIPA buffer (Sigma) and sub-cellular fractions using the NE-PER Nuclear Protein Extraction Kit (Thermo Scientific, Rockford,IL). Samples were loaded onto a 4-20% Tris-glycine gel and transferred to a PVDF membrane. The membranes were blocked at room temperature for 45 minutes in 5% non-fat milk in TBS-Tween (0.05%). Primary antibodies were as follows: BMX (Abcam-59360), pBMX (Cell Signaling-3211S), STAT3 (Santa Cruz-SC482), pSTAT3/Tyr705 (Cell Signaling-9131S), SOX1 (Cell Signaling 4194S) and Actin (Abcam-8227-50) and incubated overnight at 4°C. The membrane was washed 3× for 10 minutes each using TBS-T (0.1%). Secondary antibody was applied for 1 hour at room temperature (infrared goat-anti rabbit or mouse in the 800 channel) and washed. The membrane was developed using the Odyssey from Licor (Lincoln, NE). Protein loading was normalized using actin as a control. Densitometry analysis was performed using ImageJ (NIH, Bethesda, MD).

Cells were seeded overnight in a 96 well plate in 100 μL of regular media at a density of 2000 cells per well. Cell proliferation was measured using the CellTiter-Glo assay from Promega on Day 1, 3, 5 and 7 using 100 μL of reagent and an incubation time of 20 minutes. The relative luciferase units (RLU) were quantified using a Tecan Infinite 200 plate reader.

LNCaP and DU145 cells were seeded at 1000 cells per mL in replacement media SCM supplemented with KO Serum Replacement (Invitrogen) for LNCaP or B27 (Invitrogen) for DU145 cells in non-adherent 6 well plates coated with Hydrogel (Corning Life Sciences, Chemlsford, MA). The prostatospheres were generated for 5-7 days and then quantified or RNA extracted.

Staining of invasive or non-invasive cells was performed directly on the Matrigel membrane. Duplicate invasion chambers were used for each antibody; one each for staining invasive cells or non-invasive cells. Cells not being stained were removed from each insert, and cells of interest were fixed to the membrane in 4% para-formaldehyde for 15 minutes at 25°C and permeabilized with 0.5% saponin in PBS for 15 minutes at 25°C followed by a series of washes with PBS. Non-specific antibody binding sites were blocked for 15 minutes with 1% BSA in PBS containing 0.1% Tween-20 (PBS-T). Cells were incubated with either anti-pBMX antibody in PBS-T, SOX1 (Cell Signaling, Danvers, MA), or pSTAT3 (Millipore/Upstate Technologies, Billerica, MA (4°C, overnight). Following 3× PBS-T washes, infrared goat anti-rabbit Alexa 488 (Molecular Probes, Carlsbad, CA) was added for 1 hour at 25°C using a 1:500 dilution in PBS-T and again washed, then air-dried. Membranes were mounted on glass slides with Vectashield containing DAPI (Vector Laboratories, Burlingame, CA). Cells were visualized with a Zeiss-510 L5 confocal microscope where separate images were obtained for Alexa-488 and DAPI fluorescence, as well as overlays and 10 slice Z-stacks. Images were analyzed using the Zeiss LSM5 Image Browser (version 3.2.0.115) and further prepared in Adobe Photoshop CS. "Non-invasive" cells were stained on the topside of the membrane, while "invasive cells" were stained on the underside of the membrane. Controls using the secondary antibody and no primary antibody indicated that little, if any, fluorescence was contributed by non-specific binding of this antibody (data not shown).

Protein was extracted using RIPA buffer (Sigma) and lysates were incubated with either SOX1, STAT3 or BMX (same antibodies used in Western Blotting) overnight at 4°C with rotation. The next day Protein A sepharose beads were added to the lysate and incubated for 3 hours with rotation at 4°C. The lysate was then spun at 13,000 rpms in a benchtop centrifuge and washed 3× with RIPA buffer. Before loading on a 4-20% Tris-Glycine SDS-Page gel (Invitrogen) 2× loading buffer was added and upon completion the gel was transferred to a PVDF membrane. The membrane was blocked for 45 minutes using 5% non-fat milk in TBS-T (0.1%). The membrane was then incubated overnight at 4°C using either primary antibodies SOX1 or STAT3 diluted in blocking buffer to confirm a direction interaction. The membrane was washed 3× for 10 minutes each using TBS-T (0.1%). Secondary antibody was applied for 1 hour at room temperature (infrared goat-anti-rabbit in the 800 channel) and washed. The membrane was developed using the Odyssey from Licor. Protein loading was normalized using actin from pervious Westerns.

The Licor EMSA buffer kit was used according to the manufacturer's instructions. Infrared (IR) and unlabeled STAT3 oligos were ordered from IDT and used at 0.625 fmoles/reaction.

Wildtype probes (WT): (800-IR channel and unlabeled)

F- 5'-GATCCTTCTGGGAATTCCTAGATC-3';

R- 5'-GATCTAGGAATTCCCAGAAGGATC-3'

Mutant probes (MU): (700-IR channel)

F- 5'-GATCCTTCTGGGCCGTCCTAGATC-3';

R- 5'-GATCTAGGACGGCCCAGAAGGATC-3';

Mutant oligos and unlabled wildtype oligos were used at 200-fold molar excess. A total of 20 μg of nuclear protein extract was incubated with 1× binding buffer (100 mM Tris, 500 mM KCl, 10 mM DTT; pH 7.5), Poly (dl·dC) 1 μg/μL (in 10 mM Tris, 1 mM EDTA, pH 7.5), 25 mM DTT/2.5% Tween-20, 1% NP-40, 100 mM MgCl₂, and 50% glycerol for 20 minutes at room temperature shielded from light. For supershift experiments, extracts were pre-incubated with 5 μg of STAT3 antibody at 4°C for 30 minutes. DNA/protein complexes were visualized on a native 6% Tris-Borate-EDTA polyacrylamide gel. Gels were immediately removed from cassettes and scanned using the Odyssey in both the 700 and 800 channels.

Oncomine and Gene Expression Omnibus (GEO) databases were queried to identify associations between genes. GEO database is available at http://​www.​pubmed.​org (GEO profiles) and provides raw expression data from several gene expression arrays. Oncomine 4.2 database analysis tool is available with a subscription at http://​www.​oncomine.​org Selected data was compared for gene expression levels in prostate primary tumor samples as well as their respective metastatic specimens. Data have been selected from [17] because this study was an integrated molecular profiling of gene expression in prostate cancer samples. In this work, a significant concordance between expression of Sox1 and Stat3 mRNA was found to correlate with the aggressiveness of the sample.

All statistical calculations were performed using GraphPad Prism Version 5. Comparisons between groups were carried out using either a Student's pair-wise t-test, or a One or Two-way ANOVA with a Bonferroni post-test wherever each test was applicable. Error bars represent the Standard Error of the Mean (SEM) and each experiment has been completed at least twice with samples in triplicate.

Individual promoter tiling arrays were performed to analyze global CpG promoter methylation for both non-invasive and invasive cell isolates from both LNCaP and DU145 (Figure 1). The cells were allowed to invade the Matrigel toward a highly defined media called stem cell media (SCM) [18]. It was then determined which genes were methylated in the non-invasive cells and not in the invasive fraction of cells. This analysis determined that 869 probes were differentially methylated in the non-invasive LNCaP fraction compared with the invasive and 1015 for DU145 (Additional File 1, Tables S1A and S1B). A very small subset of 44 overlapping genes was methylated in the non-invasive cells and not in the invasive population from both of the prostate cancer lines analyzed. These included genes involved in development such as Irx3, Six1 and Sox1, as well as a type-III 5" deiodinase (Dio3), and an embryonic version of myosin (Myh3) (Table 1). Using the Oncomine database we investigated changes in expression patterns for these methylated targets, and we found a significant association between progression of prostate cancer and metastasis with expression of a number of genes including G protein, beta-1 subunit (Gnb1), retinoblastoma binding protein 8 (rbbp8), secretogranin III (Scg3) and Sox1 (Figure 2). Albeit a number of these proteins have been shown to play a role in cancer, we chose to investigate the role of Sox1 in our model since it is very homologous to the induced pluripotent stem cell (iPS) regulator Sox2, and has been shown to play a role in progression of lung and nasopharyngeal cancer [19, 20]. We also chose to investigate bone marrow tyrosine kinase gene in chromosome X protein (Bmx) since it has been shown to regulate hematopoiesis [21] and play a role in the regulation of prostate cancer [22]. However, from our Oncomine analysis Bmx was not shown to significantly affect prostate cancer metastasis (Figure 2).

To verify the results from our methylation specific promoter tiling arrays, we performed methylation specific PCR (MS-PCR) where primers were designed around the probe sequences identified from the arrays. Both Bmx and Sox1 were found to be methylated in the parental (total) LNCaP and DU145 cell lines (Figure 3A), representing the non-invasive phenotype. To determine if this pattern of methylation correlated with the level of gene expression, real time quantitative PCR (qRT-PCR) was performed. Significant differences in the expression of Bmx and Sox1 were seen when comparing the expression in non-invasive and invasive cell populations in both LNCaP and DU145 cell lines (Figure 3B) (Two-way ANOVA; *compares non-invasive to parental and ** compares invasive to parental, p < 0.05). To further validate the results, immunocytochemistry (ICC) was performed to analyze differences in protein expression between non-invasive and invasive cells. There is significantly higher expression of activated BMX and SOX1 in the invasive versus non-invasive cells (Figure 3C). Therefore, we validated the methylation and resultant decreased expression of BMX and SOX1 in the non-invasive cells.

To further determine the role of Bmx and Sox1 during the process of invasion we performed the invasion assay with DU145 cells stably infected with shRNAs directed against Sox1 or Bmx (Figure 4). A significant decrease in expression of SOX1 and BMX following induction with 1 μg/mL of doxycycline (Dox) for 24 hours was first verified using western blotting. Upon induction with Dox, the shRNA is turned on and a downstream red fluorescent protein (RFP) demonstrates efficiency of this induction (Figure 4A). Densitometry analysis was performed to compare expression of individual clones with the NS cells, and no significant differences in protein expression were seen using the non-silencing (NS) controls (Figure 4B). In addition, SOX1 shRNA cells demonstrated a significant decrease in proliferation compared to either the parental cell line (total cells) or the NS infected line (Figure 4C), as well as a significant decrease in invasion toward SCM (Figure 4D) (p-value < 0.05). However, there was not a significant difference using the shBMX lines, except for a slight reduction in invasion using clone #3. Interestingly, a small increase in proliferation was seen with the shBMX clones (Figure 4C). Further promoter tiling array analysis using two short term cultures primary prostate tumor cell lines, PCSC1 and PCSC2, determined that Sox1, and not Bmx, was methylated in the invasive population of cells (Additional File 2, Table S2A and B). Overall, we demonstrate that Sox1 is differentially methylated within the invasive CSC population and the shRNA studies indicate it could be selectively targeted to block invasion.

In addition to the method presented here, prostate TICs (tumor initiating cells) can also be isolated by culturing total cells in SCM where structures called prostatospheres are generated [18, 23‐25]. The prostatospheres are multicellular globes that develop from cells that survive anchorage-independent conditions in vitro, and are frequently used when analyzing the ability of TICs to self-renew or differentiate upon the addition of serum. Using this assay as a model, a greater number of prostatospheres were isolated from DU145 NS cells compared to shSOX1 cells (Additional File 3, Figure S1A and B). When invasive DU145 cells were isolated and cultured in SCM, prostatospheres were maintained for up to 3 passages (Additional File 3, Figure S1C) and if these cells were further cultured in the presence of 1% human serum (Additional File 3, Figure S1C), the vector control (NS) cells rapidly differentiated and proliferated, while the shSOX1 cells did not (Additional File 3, Figure S1D). These observations suggest that not only does Sox1 play a role in regulating invasion, but it can also regulate the maintenance of 'stem-ness' in culture.

Each data set of differentially methylated genes was then extracted and uploaded to the Ingenuity server to identify common gene pathways that are regulated during the process of invasion. The most conserved functional pathways between the cell lines are cellular development, cell growth and proliferation, as well as organismal development, nervous system development and function, and tissue development (Additional File 4, Table S3). The full list from the Ingenuity pathway analysis is also included (Additional File 5, Figure S2A and S2B). Additionally, the IL-6 signaling pathway involving STAT3 had a significant number of contributing methylated genes, a pathway recently found to play a significant role in cancer stem cell regulation [26‐34] (Additional File 6, Figure S3A).

Based on the information generated from Ingenuity, we chose to determine how the IL-6 pathway might be regulating this process of invasion. A number of inhibitors of downstream targets of IL-6 regulation were tested for their ability to block invasion toward SCM. We included a neutralizing antibody to interleukin-6 (IL-6) to test what effect this may have upstream. Downstream of the receptor, the following inhibitors were used; the PI3K inhibitor LY294002, small molecular inhibitor of MEK called U0126 (thus downstream inhibition of extracellular-related kinase (ERK1 and ERK2) mediated responses), a small molecule inhibitor of JAK (Janus Kinase) called AG490 and an inhibitor of its partner signal transducers and activators of transcription-3 (STAT3) called Stattic (Figure 5A). Additionally, we tested the ability of the Tec kinase family inhibitor LFM-A13 based on the potential involvement of BMX during invasion (Figure 5A). The inhibitors which demonstrated the greatest effect at blocking invasion included Stattic, LY294002, and LFM-A13 (Figure 5A). However, a proliferation assay determined that Stattic could be preventing invasion because it was either cytotoxic to the cells or causing them to undergo apoptosis (Additional File 6, Figure S3B). To eliminate this possibility, viable cells were isolated after treating the DU145 cell line with Stattic for 24 hours (data not shown). These cells, although viable as determined by trypan blue staining, were still unable to invade.

Since inhibition of STAT3 demonstrated such a profound effect on invasion toward SCM, we questioned its involvement with the epigenetically regulated targets. Although we did not observe methylation of Stat3 itself, in both cell lines, the mRNA expression of Stat3 was increased (p-value < 0.03) when comparing invasive cells to their non-invasive counterpart (Figure 5B). Protein expression of pSTAT3 was also found to be increased in the invasive cells (Figure 5C). Since both SOX1 and STAT3 are known to act as transcriptional activators after forming protein complexes with other proteins [35‐40], and BMX is known to activate STAT3 itself [40], we determined whether STAT3 directly interacts with either SOX1 or BMX. An interaction between SOX1 and STAT3 was observed (Figure 5D), however not between STAT3 and BMX (Figure 5D). In addition, a significant decrease in the expression of activated pSTAT3 was seen in both sub-cellular fractions of the BMX and SOX1 shRNA infected cells (Figure 5E). However, there was no change in total expression of STAT3. Additionally, a significant decrease in STAT3 DNA binding activity was observed in both BMX and SOX1 shRNA infected cells (Figure 5F). Overall, we see an interaction between SOX1 and STAT3, and upon loss of either BMX1 or SOX1 expression we observe a loss of STAT3 activation.

To further elucidate the connection between the SOX1 and STAT3, a decrease in the STAT3 target gene Mcl-1 and Stat3 itself were observed by qRT-PCR in shSOX1 clone #7 cells (Figure 6A). However, no change was observed for the STAT3 targets genes Survivin or Myc (Figure 6A). Finally, since prostatospheres are also a model for generating aggressive populations of cells in culture, we generated them from LNCaP cells and asked if STAT3 genes were affected. qRT-PCR analysis was performed and compared to adherent LNCaP cells, expression of Stat3 and Stat3 target genes Mcl-1, Myc, and Survivin were increased as well as Bmx and Sox1 (Figure 6B).

In order to determine what might be regulating the increased expression of Stat3 and Sox1, transcription factor binding sites were analyzed using Genomatix software. In both the Stat3 and Sox1 promoters there are a number of overlapping binding sites for transcription factors with a significant matrix value such as GATA-binding factors, RNA polymerase II transcription factor IIB (TFIIB), NeuroD/Beta2, TALE homeodomain class recognizing TG motifs, TCF11 transcription factor otherwise known as Nrf2, Nkx homeodomain factors, and finally the Zinc finger transcription factor RU49 also called Zipro1 (Additional File 7, Table S4). With this information, we can begin to understand why the methylation of Sox1 could serve as a master regulator of CSC invasion, thereby controlling its potential to undergo EMT and further metastasize.

Additional analysis using the GEO database determined that both Sox1 and Stat3 are expressed at higher levels in metastatic prostate cancer tissues and not Bmx (Figure 6C and 6D). Overall, we demonstrate that SOX1 is an epigenetically regulated target involved in the progression of prostate cancer, and is involved in signaling via the STAT3 pathway.