Altered promoter nucleosome positioning is an early event in gene silencing

Abstract

Gene silencing in cancer frequently involves hypermethylation and dense nucleosome occupancy across promoter regions. How a promoter transitions to this silent state is unclear. Using colorectal adenomas, we investigated nucleosome positioning, DNA methylation, and gene expression in the early stages of gene silencing. Genome-wide gene expression correlated with highly positioned nucleosomes upstream and downstream of a nucleosome-depleted transcription start site (TSS). Hypermethylated promoters displayed increased nucleosome occupancy, specifically at the TSS. We investigated 2 genes, CDH1 and CDKN2B, which were silenced in adenomas but lacked promoter hypermethylation. Instead, silencing correlated with loss of nucleosomes from the -2 position upstream of the TSS relative to normal mucosa. In contrast, permanent CDH1 silencing in carcinoma cells was characterized by promoter hypermethylation and dense nucleosome occupancy. Our findings suggest that silenced genes transition through an intermediary stage involving altered promoter nucleosome positioning, before permanent silencing by hypermethylation and dense nucleosome occupancy.

Abbreviations

TSS

transcription start site; NDR, nucleosome depleted region; CGI, CpG island; CRC, colorectal carcinoma; NOMe-Seq, nucleosome occupancy and methylome sequencing; MBD-Seq, Methyl-CpG binding domain (MBD) protein DNA enrichment and sequencing; MNase-Seq, micrococcal nuclease and sequencing

Introduction

Gene expression can be regulated by changes to DNA methylation, as well as by the positioning and occupancy of nucleosomes at promoter regions.¹ The term positioning describes the precise location of a given nucleosome, whereas occupancy describes the proportion of molecules bearing a nucleosome at a specific location, at any given instant.² The positioning of nucleosomes at promoters regulates gene expression by demarcating the promoter region and transcription start site (TSS).³ At gene promoter regions, nucleosomes can be held at specific positions by DNA-binding proteins such as transcription factor complexes.⁴ While activation of gene expression correlates with nucleosome depletion at promoters, nucleosomes have been shown to rapidly reform when transcription ceases.^5,6

In cancer, many genes critical to tumor development are known to undergo epigenetic silencing. Typically, this silencing occurs in association with hypermethylation and dense nucleosome occupancy across the CpG island (CGI) promoter region.⁷ However, the majority of genes that are hypermethylated in cancer are also silenced in normal precursor cells despite no evidence of promoter methylation.^8,9 These studies support the view that hypermethylation serves to consolidate a transcriptionally silent state rather than initiate it.¹⁰ However, the mechanism by which a promoter transitions from normal to silenced and ultimately hypermethylated states remains poorly defined.

Recently, we showed that resilencing of the MLH1 gene following drug-induced demethylation was initiated by the reassembly of nucleosomes at the TSS, and was then followed by remethylation.¹¹ Deposition of nucleosomes at the NANOG gene promoter and enhancer also precedes hypermethylation and stable silencing.¹² These studies led us to question whether changes in nucleosome positioning at a promoter occur during aberrant gene silencing in neoplasia prior to hypermethylation.

In this study, we investigated nucleosome positioning at the promoters of genes that were silenced in colorectal adenomas in the absence of hypermethylation. Because promoter hypermethylation is frequent in colorectal carcinomas (CRCs), but rare in adenomas,¹³ we reasoned that promoters undergoing silencing in the absence of hypermethylation would be more common in lesions that reflect the earlier stages of colorectal neoplastic progression.

Results

Gene expression is associated with highly ordered nucleosome positioning

Using RNA-Seq and MNase-Seq data from an adenoma from one individual (patient 81), we investigated the relationship between gene expression, nucleosome positioning, and nucleosome occupancy at promoters. We use the term positioning to describe the precise location of nucleosomes and occupancy to describe the relative levels of nucleosomes at a specific location. We separated transcripts with CGI promoters into quartiles of expression and determined nucleosome occupancy across promoter regions. This showed the promoters of highly expressed (4th quartile) transcripts contained a nucleosome-depleted region (NDR) of ∼200 bp across the TSS, which was flanked by precisely positioned nucleosomes immediately upstream (labeled as the −2, −3, −4 and −5 nucleosomes in Fig. 1A) and downstream (labeled as the +1, +2, +3 and +4 nucleosomes in Fig. 1A). Transcripts with lower expression levels showed loss of this precise positioning (Fig. 1B-D), as well as increased occupancy of nucleosomes at the TSS (Fig. 1E). These data demonstrate that gene expression is associated with the precise positioning of promoter nucleosomes and the presence of a NDR across the TSS.

Figure 1.

The precise positioning of nucleosomes at gene promoters is altered at hypermethylated promoters. Panels (A-D)show nucleosome positions at CGI promoters for transcripts expressed at high (Q4, 4th quartile), moderately high (Q3, 3rd quartile), moderately low (Q2, 2nd quartile) or low levels (Q1, 1st quartile) in adenoma tissue from patient 81. Thin lines represent absolute values, whereas thick lines represent moving average trend lines of absolute values. Y-axis values indicate the 5′ site (1st bp only) of sense strand MNase-Seq reads and are plotted as reads per 5 bp. X-axis position is corrected by +75 bp to represent values at the center of nucleosomes. Nucleosomes are labeled −5 to +4 relative to the TSS. Panels (E and F) show nucleosome occupancy at promoters determined using combined sense and antisense extended (150 bp) MNase-Seq reads. In panel (E), the data is stratified by expression quartile. Panel F, compares nucleosome occupancy at unmethylated (UnCH3) with hypermethylated (HyperCH3) promoters. Pink bar indicates 150 bp. Gray solid lines indicate −2 and +1 nucleosomes. Dashed lines indicate TSS. n = number of transcripts analyzed.

Promoter hypermethylation is associated with loss of the precise positioning of nucleosomes

Next we focused on promoters that were hypermethylated in the adenoma relative to paired normal mucosa. To do this, we profiled gene expression and DNA methylation in paired normal mucosa using RNA-Seq and MBD-Seq, respectively. We compared nucleosome positioning and occupancy at hypermethylated (n = 1,725) and unmethylated (n = 11,561) promoters. Unmethylated promoters contained a NDR at the TSS bordered by well-positioned −2 and +1 nucleosomes (Fig. 1F). This positioning mirrored that seen at the promoters of transcripts in the 4^th quartile of expression (Fig. 1E). Hypermethylation was associated with the loss of precise positioning of nucleosomes at the −2 and +1 positions, as well as increased occupancy at the TSS (Fig. 1F). This positioning was similar to the positioning seen at the promoters of transcripts in the 1st quartile of expression (Fig. 1E and F).

Silenced and unmethylated promoters show altered nucleosome positioning

Next we investigated nucleosome positioning at the promoters of genes that were downregulated >50% in the adenoma relative to normal mucosa but remained unmethylated (as determined by RNA-Seq and MBD-Seq data, respectively, see Table S1). To enrich for genes in the early stages of epigenetic silencing we cross-referenced downregulated genes with the PubMeth database (www.pubmeth.org). This approach identified the CDH1 and CDKN2B tumor suppressor genes, which are known targets of promoter hypermethylation in a range of different cancer types. Using qRT-PCR we determined CDH1 and CDKN2B expression levels in 12 additional adenomas and paired normal mucosa. All 12 adenomas showed loss of CDH1 expression to between 3.4% and 25.6% (mean, 13.7%, SD ± 7.1) of the levels seen in normal mucosa (Fig. 2A). We selected at random 6 of the adenoma/normal mucosa pairs for analysis using 2 separate NOMe-Seq assays encompassing the CDH1 TSS (Fig. 2B). In 5 of the 6 adenomas we observed reduced nucleosome occupancy in the −2 position compared with normal tissue (Fig. 2C-I), whereas in one adenoma the levels at this position were similar to normal mucosa, though slightly higher in the adenoma (Fig. S1). For example, in one adenoma nucleosomes were present in 89% (8/9) of molecules 150 bp upstream of the CDH1 TSS in normal mucosa but in only 27% (6/22) of molecules at the same location in the paired adenoma (Fig. 2C and E). Of the 6 adenomas tested, 5 displayed low-level hypermethylation (defined as more than 2 methylated CpG sites per molecule) upstream of the TSS in a minority of molecules (Fig. 2C; Fig. S1A; Fig. S2), while one showed hypermethylation upstream of the TSS in most molecules (Fig. 2D and F). Interestingly, this hypermethylated adenoma also showed increased nucleosome occupancy specifically across the TSS (Fig. 2D and F). NOMe-Seq analysis of the control gene HSPA5, a constitutively and highly expressed gene that maintains a NDR at the TSS,¹⁴ confirmed that these differences in accessibility were a result of altered nucleosome positioning rather than incomplete M.CviPI treatment (Fig. S3).

Figure 2.

CDH1 silencing correlates with loss of the −2 nucleosome in adenomas relative to normal mucosa. Panel (A)shows CDH1 expression levels in adenoma and paired normal mucosa as determined by qRT-PCR normalized to GAPDH. Panel (B)shows a schematic of the CDH1 promoter indicating the NOMe-Seq assays (N1 and N2, gray bars) and their locations relative to the predicted average positions of the center of the −2 and +1 nucleosomes, determined using genome-wide MNase-Seq at highly expressed (4th quartile) genes (green line, reproduced from Figure 1, panel A). At 4^th quartile genes the average position of the −2 and +1 nucleosomes was −140 bp and +167 bp relative to the TSS, respectively. Panels (C-D)show NOMe-Seq data from the CDH1 promoter in normal mucosa and paired adenomas. Shown is one adenoma with low-level methylation (C) and an adenoma with dense hypermethylation (D). N = normal, Ad = adenoma. Red arrows indicate the CDH1 TSS [NM_004360]. Thin vertical black lines represent the positions of GpC and CpG dinucleotides, respectively. Black circles = GpC dinucleotides methylated/accessible to the GpC methyltransferase M.CviPI. White circles = GpC dinucleotides methylated/accessible to GpC methyltransferase. Black triangles = methylated CpG dinucleotides, white triangles = unmethylated CpG dinucleotides. Pink shading indicates regions of accessibility ≥ 150 bp or > 75 bp at the extreme ends of amplicons. In both adenomas, Region N1 shows marked loss of nucleosome occupancy from the −2 position upstream of the TSS relative to paired normal mucosa. Panels (E-I)summarize the NOMe-Seq data presented in panels (C) and (D), and in Figure S2. Depicted is the proportion of nucleosome-occupied GpC sites across the CDH1 promoter. Thin gray vertical bars indicate the average positions of the center of the −2 and +1 nucleosomes upstream and downstream of the TSS of 4th quartile genes. All 5 adenomas exhibited loss of the −2 nucleosome upstream of the TSS relative to paired normal mucosa. See alsoFigures S1-3.

CDKN2B was silenced in all 12 adenomas, however expression was only detectable in 5 of 12 normal mucosa specimens (Fig. 3A). We selected one adenoma and paired normal mucosa for NOMe-Seq analysis of the CDKN2B promoter using a NOMe-Seq assay that allowed us to detect nucleosomes in the −2 and +1 positions in the same DNA molecule (Fig. 3B). We again observed a striking loss of nucleosomes from the −2 position in adenoma relative to normal tissue (Fig. 3C-E). In normal mucosa we detected a nucleosome 150 bp upstream of the TSS in 58% (18/31) of molecules but in only 13% (4/31) of molecules in the adenoma (Fig. 3C-E). Low-level hypermethylation in the adenoma was observed in 23% (7/31) of molecules (Fig. 3E).

Figure 3.

Loss of −2 nucleosomes at the silenced CDKN2B promoter. Panel (A)shows CDKN2B expression levels in adenoma and paired normal mucosa as determined by qRT-PCR normalized to GAPDH. Panel (B)shows a schematic of the CDKN2B promoter showing that the region assayed by NOMe-Seq (gray bar) encompassed the predicted average positions of the center of the −2 and +1 nucleosomes, determined using genome-wide MNase-Seq at highly expressed (4th quartile) genes (green line, reproduced from Figure 1, panel A). At 4th quartile genes the average position of the −2 and +1 nucleosomes was −140 bp and +167 bp relative to the TSS, respectively. Panel (C)summarizes the NOMe-Seq data presented in panels D and E. Depicted is the proportion of nucleosome-occupied GpC sites across the CDKN2B promoter. Thin gray vertical bars indicate the average positions of the center of the −2 and +1 nucleosomes upstream and downstream of the TSS of 4th quartile genes. In this adenoma, marked loss of nucleosome occupancy from the −2 position upstream of the TSS was observed relative to paired normal mucosa. Panels (D)and (E)show NOMe-Seq data from the CDKN2B promoter in normal mucosa (panel D) and adenoma tissue (panel E). Note the clear loss of nucleosomes from the −2 position upstream of the TSS in adenoma tissue (panel E) relative to paired normal mucosa (panel D). Molecules showing low-level hypermethylation (3 or more methylated CpG dinucleotides per molecule) in the adenoma are indicated by a black box to the right of panel E. See also Figure S3.

CDH1 expression correlates with a NDR across the TSS, whereas stable silencing involves hypermethylation and dense nucleosome occupancy

To examine how nucleosome positioning changes in cells with stable silencing of CDH1 we used colorectal cell lines that express (HCT116) or do not express (RKO) CDH1 (Fig. 4A). In HCT116 cells, the TSS was within a NDR, and was accessible in the vast majority (97%; 36/37) of promoter molecules (Fig. 4B and C) as determined by combining the reads from Regions N1 and N2 across the 9 GpC sites shared between these assays. Also, this NDR was bordered by nucleosomes in the −2 or +1 positions in approximately half of molecules. All promoter molecules were unmethylated at CpG sites (Fig. 4C). This contrasted strikingly with data from RKO cells, in which the CDH1 promoter was inaccessible in 86% (31/36) of molecules and the entire promoter was densely methylated (Fig. 4B and D). These data show that in cells expressing CDH1 the promoter is unmethylated and nucleosomes border a NDR across the TSS. However, in cells showing stable silencing of CDH1, the promoter is densely methylated and occupied by nucleosomes.

Figure 4.

Permanent CDH1 silencing correlates with promoter hypermethylation and dense nucleosome occupancy. Panel A, qRT-PCR data showing CDH1 expression levels normalized to GAPDH (left) and an immunoblot of CDH1 protein in total protein lysates from HCT116 and RKO cells (right). The doublet represents processed and unprocessed CDH1 protein. Immunoblot of GAPDH protein was used to control for equal loading. Panel (B)shows the proportion of nucleosome-occupied GpC sites across the CDH1 promoter in HCT116 and RKO cells, as determined from NOMe-Seq data shown in panels C and D. Thin gray vertical bars indicate the average positions of the center of the −2 and +1 nucleosomes upstream and downstream of the TSS of 4th quartile genes. Note the presence of a nucleosome-depleted region across the CDH1 TSS and highly positioned −2 and +1 nucleosomes, whereas in RKO cells the promoter shows high levels of nucleosome occupancy. Panels (C)and (D)show NOMe-Seq data from regions N1 (left) and N2 (right) in HCT116 and RKO cells, respectively. See also Figure S3.

Discussion

This study shows that gene silencing can involve altered positioning of promoter nucleosomes prior to promoter hypermethylation. Specifically, loss of expression of CDH1 and CDKN2B was associated with the loss of −2 nucleosomes from the unmethylated promoters of these genes.

Our genome-wide analysis confirmed previous reports of a close relationship between gene expression levels and the positioning of nucleosomes at promoters.¹⁵ Most notably, gene expression correlated with nucleosome depletion at the TSS and also with well-positioned upstream and downstream nucleosomes. The tight relationship between expression and nucleosome positioning was confirmed by our finding that the promoters of transcripts that had lower levels of expression showed a loss of positioning of nucleosomes flanking the TSS, as well as increased nucleosome occupancy at the TSS.

In this study, the availability of DNA methylation data from normal colonic mucosa and a colonic adenoma from the same patient allowed us to examine promoters that were hypermethylated specifically in the neoplastic tissue. Genome-wide, hypermethylated promoters showed increased occupancy at the TSS and loss of the precise positioning of nucleosomes flanking the TSS. MNase-Seq data does not provide information regarding the methylation status of nucleosome-occupied or unoccupied DNA. To overcome this limitation, we investigated the relationship between methylation and nucleosome positioning using a method that simultaneously assesses both parameters on the same DNA molecule. In doing so, we focused on the promoters of CDH1 and CDKN2B, 2 genes known to be important in colorectal tumorigenesis. These genes were chosen because our genome-wide data showed they were downregulated in the absence of hypermethylation, yet both are known targets of hypermethylation in cancer.¹⁶ The analysis of paired normal/adenoma tissues allowed us to show that CDH1 and CDKN2B silencing correlated with reduced occupancy of nucleosomes at the −2 position in adenomas. Considering individual DNA molecules separately, this indicates that a proportion showed eviction or sliding of nucleosomes from this position. These findings reaffirm that an NDR, flanked by precisely positioned −2 and +1 nucleosomes, is required for gene expression, and that alterations in this positioning creates an intermediate stage of gene silencing. This intermediate stage does not exhibit the dense DNA hypermethylation that characterizes permanent consolidation of the silent state.

Due to the observed loss of −2 nucleosomes in different adenomas and at 2 separate promoters, it is likely to be relevant in the silencing of other genes. In this regard, You et al., showed that altered nucleosome occupancy at the NANOG enhancer preceded hypermethylation and silencing of the NANOG promoter during in vitro cell differentiation.¹² Also, we previously reported that nucleosome deposition at the TSS of the MLH1 gene preceded remethylation following decitabine treatment.¹¹ The significance of loss of −2 nucleosomes is, at present, unclear. The region of a promoter upstream of the TSS is rich in binding sites for transcriptional regulators.⁴ Presumably, therefore, the accessibility of potential upstream activator sequences suggests that the CDH1 and CDKN2B promoters may be poised for reactivation if transcription factors were recruited. However, increased accessibility upstream of the TSS may also predispose it to de novo methylation. Consistent with this hypothesis, genome-wide NOMe-Seq data reveals that DNA methylation may be phased with the peaks of methylation occurring at accessible linker DNA between nucleosomes.¹⁷ Indeed we observed some evidence that accessible regions within the CDH1 and CDKN2B promoters had become methylated at a small number of CpG sites. In one patient we did not observe the same loss of −2 nucleosomes from the CDH1 promoter despite downregulation of the gene. This result could be explained if this adenoma contained a nonsense mutation within the CDH1 gene, which would lead to normal transcription but low levels of detectable mRNA due to nonsense-mediated mRNA decay.

The precise positioning of the −2 and +1 nucleosomes at a given promoter is maintained by the continued recruitment of transcription factors, which serve as a barrier against which nucleosomes are positioned (Fig. 5A).¹⁸ We suggest that alterations in nucleosome positioning are subsequent to the loss of transcription factor binding (Fig. 5B); in this model, altered nucleosome positioning is a consequence of loss of expression rather than its cause. This is supported by previous studies showing that transcription of the yeast GAL genes can be repressed regardless of whether nucleosomes are allowed to reform at the promoter.¹⁹ Our data shows that alterations in promoter nucleosome positioning can occur in the absence of methylation, however the resulting hiatus in transcriptional activity, and the increased accessibility of DNA upstream of the TSS, may predispose to de novo methylation (Fig. 5C). This hypothesis was supported by our observation of low levels of methylation in the adenomas at the CDH1 and CDKN2B promoters, suggesting these genes were susceptible to further methylation while being held in a transcriptionally repressed state. Finally, consolidation of gene silencing is characterized by dense promoter methylation and nucleosome occupancy (Fig. 5D), as exemplified by the state of the CDH1 promoter in RKO cells. Further work will be required to confirm this sequence of events and whether loss of the −2 nucleosome provides an appropriate environment for subsequent de novo hypermethylation.

Figure 5.

Model of chromatin reorganization during gene silencing. Panels (A-D)depict the alterations in promoter nucleosome positioning and DNA methylation during de novo gene silencing, beginning with an active promoter (A) and ending with stable silencing (D). Red spheres represent nucleosomes. Black triangles represent methylated CpG dinucleotides. White triangles represent unmethylated CpG dinucleotides. Blue oval represents transcription factor complexes (TF). The green arrow and blunt red arrows indicate the transcription start site (TSS) and expression status (expressed or not), while −2 and +1 represent nucleosome positions.

Methods

Patients and tissue samples

A 50 mm adenoma from the transverse colon classified as a non-granular laterally spreading tumor (LST), Paris morphology 0-IIa, with tubular histological architecture and high-grade dysplasia and a biopsy of rectal mucosa were removed from a female patient aged 73 y (labeled patient 81) using endoscopic mucosal resection (EMR) and used for RNA-Seq, MBD-Seq and MNase-Seq (ethics numbers 2009/6/4.6 and 11194). The EMR procedure typically yields 85–95% pure lesional tissue. To validate the results obtained in this single adenoma, an additional 12 fresh colorectal adenomas and paired samples of adjacent normal mucosa were obtained from surgical resection specimens from 7 females and 5 males (mean age 68.7, SD 9.3 years) at St. Vincent's Hospital, Sydney (ethics numbers H00/022 and 00113). Nine adenomas were located on the right side of the colon, one in the sigmoid colon and 2 in the rectum. All 12 lesions showed tubulovillous histological architecture with high-grade dysplasia, and all were larger than 15 mm (mean 39.9 mm, SD 14.9 mm, range 15–60 mm). No lesion in this study showed evidence of invasive malignancy.

Cell culture

The colorectal cancer cell lines HCT116 and RKO were obtained from the American Type Culture Collection (ATCC) after cell authentication testing using microsatellite and mutation analysis. Cells were maintained in DMEM media supplemented with 25 mM glucose, 10% (v/v) FBS, 100 units penicillin, 100 μg/mL streptomycin and 2 mM glutamate (Life Technologies) and grown at 37°C in 5% CO₂.

Immunoblotting

Cells were lysed on ice in 50 mM Tris HCl pH 7.5, 150 mM NaCl, 1% (v/v) Triton X-100, 0.5% (w/v) deoxycholic acid, 0.1% (w/v) sodium dodecyl sulfate (SDS) and 1 x EDTA-free complete protease inhibitor (Roche), then vortexed and sonicated followed by centrifugation to pellet cell debris. Protein concentration was determined using the bicinchoninic acid protein assay (Pierce) following manufacturer's instructions. Proteins were resolved by SDS-PAGE, transferred to nitrocellulose membrane (Millipore) and probed with anti-E-cadherin (Cell Signaling Technology) or anti-GAPDH (Cell Signaling Technology) before incubation with anti-IgG HRP (Dako). Proteins were visualized by enhanced chemiluminescence using Image Quant TL software and an Image Quant LAS 400 (GE).

RNA-Seq

Total RNA was amplified using the Ovation RNA-Seq system V1 (NuGEN) prior to sequencing. Single-stranded cDNA libraries were generated using size-selected RNA and sequenced using the Illumina HiSeq2000 analyzer (BGI-Hong Kong). Low quality scoring reads and adaptor sequences were removed. Remaining reads (53.2 million paired-end) were aligned to the human genome (hg19) with the software package TopHat²⁰ using standard parameters. Refseq transcripts were quantified as reads across exons using the HOMER package²¹ and read counts between samples were normalized using the Trimmed Mean of M-value method.²²

Methyl-CpG binding domain (MBD) protein DNA enrichment and high-throughput sequencing (MBD-Seq)

MBD-Seq was used to profile DNA methylation genome-wide. Genomic DNA was fragmented in a water bath sonicator (BioruptorTM, Diagenode) to an average size of ∼400 bp and methylated DNA enriched using the MethylMiner^TM kit (Invitrogen). Methylated DNA was enriched by overnight incubation with MBD2 beads at 4°C and eluted using 2 M NaCl. Enriched DNA was subject to size selection (350–450 bp) followed by adaptor ligation, flow cell hybridization and cluster amplification for 11–13 cycles using TruSeq SR Cluster Kit v3-cBot-HS (Illumina) prior to sequencing using the Illumina GAIIx platform (single-end, 50 bp). Low quality scoring reads and adaptor sequences were removed. Remaining reads (∼31.52 million from normal mucosa and ∼28.2 million from adenoma tissue) were aligned to the human genome (hg19) using the BWA read alignment software²³ with extension to 200 bp. CGIs overlapping with −/+1 kb of a Refseq annotated TSS were considered promoter CGIs. MBD-Seq data from normal mucosa and adenoma tissues were normalized according to total library size. Methylated promoter CGIs were identified as those containing ≥10 reads in the adenoma. Promoter CGIs in the adenoma with no overlapping reads were identified as unmethylated. Hypermethylated promoter CGIs were identified as those containing ≥10 reads in the adenoma with ≥ 3 fold the number of reads in normal mucosa.

Real-time quantitative reverse transcriptase PCR (qRT-PCR)

RNA was extracted using PureLink Micro Kit (Life Technologies). cDNA was prepared using the SuperScript III cDNA Synthesis Kit (Life Technologies) as per the manufacturer's instructions. Real-time quantitative reverse transcriptase PCR (qRT-PCR) was performed in triplicates with 10 ng cDNA using iQ SYBR Green supermix (Bio-Rad) and a MyiQ iCycler (Bio-Rad). Gene expression was normalized to the glyceraldehyde-3-phosphate dehydrogenase (GAPDH [NM_002046]) gene. All primer sequences are available on request.

Isolation of mononucleosomal DNA using micrococcal nuclease (MNase) and sequencing (MNase-Seq)

Fresh adenoma tissue (29 mg) was homogenized with a pestle in 400 μL of lysis buffer (50 mM Tris pH 7.9, 100 mM KCl, 5 mM MgCl₂, 50% glycerol, 1.5% β-mercaptoethanol and complete protease inhibitor with EDTA). Nuclei were extracted by adding an equal volume of lysis buffer supplemented with 0.2% saponin and incubating on ice for 10 min. Nuclei were washed once in 1 mL of wash buffer I (50 mM Tris pH 7.5, 0.32 mM sucrose, 4 mM MgCl_2, 1 mM CaCl_2, and complete protease inhibitor with EDTA) and once in wash buffer II (wash buffer I without EDTA). Nuclei were re-suspended in 500 μL of wash buffer II. Chromatin was digested to mononucleosomes by incubating with 20 U MNase (Fermentas) for 5 min at 37°C. Digestion was stopped by the addition of 20 mM EDTA pH 8 and placed immediately on ice. To isolate DNA the sample was treated with 200 μg/mL proteinase K for 2 hr at 55°C and 0.2 U/μL RNase for 1 hr at 37°C followed by phenol chloroform extraction and ethanol precipitation. Mononucleosomal DNA corresponding to 150 bp was isolated by gel extraction using a QIAquick gel extraction kit (Qiagen). DNA was sequenced using the Illumina GAIIx platform (single-end, 50 bp). Low quality scoring reads and adaptor sequences were removed. Remaining reads (∼61 million) were aligned to the human genome using BWA software. For the construction of nucleosome occupancy profiles, either the 5′ end of reads (1st bp) was used or reads were first extended to 150 bp to reflect the size of mononucleosomes.

Nucleosome occupancy and methylome sequencing (NOMe-Seq)

For analysis of single promoters, candidates were shortlisted based on 2 criteria: 1) Genes showing >50% reduced expression in adenoma tissue from patient 81 versus paired normal mucosa (n = 6,106); 2) Genes that were not hypermethylated in adenoma tissue from patient 81 (n = 5,762). These criteria identified genes that were silenced in the absence of hypermethylation. For the CDH1 promoter we designed 2 overlapping NOMe-Seq assays (designated Regions N1 and N2) that encompassed the annotated TSS of CDH1 [NM_004360]. Region N1, located −214 bp to +42 bp relative to the TSS, was designed to capture nucleosomes in the −2 position. Region N2, located −65 bp to +180 bp relative to the TSS, was designed to capture nucleosomes in the +1 position. For the CDKN2B promoter we designed a single NOMe-Seq assay encompassing −209 bp to +226 bp relative to the annotated TSS [NM_004936], allowing the detection of −2 and +1 nucleosomes on the same molecule. NOMe-Seq was performed as described previously.²⁴ Briefly, intact nuclei were treated with 200–300 U GpC methyltransferase M.CviPl and 160–320 μM S-adenosylmethionine for 15 min at 37°C followed by termination of the reaction with an equal volume of 20 mM Tris HCl pH 7.9, 600 mM NaCl, 1% (w/v) SDS and 10 mM EDTA. DNA was extracted using phenol chloroform followed by ethanol precipitation. DNA was bisulfite modified using the EZ DNA methylation Gold Kit (Zymo Research). The CDH1, CDKN2B and HSPA5 promoter regions were amplified from 40 ng of bisulfite treated DNA (all primer sequences available upon request). PCR amplicons were cloned using the TOPO TA Cloning kit (Invitrogen) and individual molecules isolated by colony PCR for sequencing as described previously.²⁵ M.CviPI enzyme methylates accessible DNA at GpC sites, whereas nucleosome bound DNA is inaccessible and remains refractory to GpC methylation. Reactions without M.CviPI were routinely performed to confirmed endogenous CpG methylation levels. GpCpG sites were excluded from NOMe-Seq analysis. Nucleosome occupancy was defined as a region ≥150 bp that was inaccessible to M.CviPI, as described previously.^12,14,26 At the extreme ends of each molecule nucleosome occupied DNA was identified as M.CviPI inaccessibility >75 bp (half the size of a nucleosome occupied region of DNA). Duplicate molecules (those containing identical patterns of GpC and CpG methylation) were removed from further analysis to prevent data misinterpretation due to cloning or PCR bias. Molecules containing non-CpG methylation (except in the context of GpC sites) were also discarded to eliminate amplicons derived from incompletely converted DNA. Throughout the manuscript, the term positioning refers to the precise location of a nucleosome, whereas occupancy refers to the proportion of molecules bearing a nucleosome at a specific location, as described previously.²

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We would like to thank Neil Youngson, Robert Rapkins, Peter Zarzour and Duncan Sproul for helpful discussions and careful reading of the manuscript.

Funding

LBH is supported by a Cancer Institute New South Wales Career Development Fellowship (Grant number: 09CDF226; www.cancerinstitute.org.au). RLW is supported by funding from the Cancer Council NSW (www.cancercouncil.com.au) and Cancer Australia (http://canceraustralia.gov.au).

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.