CG methylated microarrays identify a novel methylated sequence bound by the CEBPB|ATF4 heterodimer that is active in vivo

CG methylation of DNA microarrays

The effect of cytosine methylation of the CG dinucleotide on DNA binding of B-ZIP proteins to multiple DNA sequences was determined using two microarray designs. One probe design has 16 sectors per slide with each sector containing 40,000 features (40K) (Lam et al. 2011). Each feature contains a 60-bp DNA with a common 25-mer at the surface of the glass that hybridizes with a primer used for the DNA double-stranding reaction. The remaining 35-mer is unique in each feature and designed such that all possible 8-mers occur 32 times (Lam et al. 2011). DNA on the array was enzymatically double-stranded, a process monitored by fluorescence of a spiked Cy3 labeled cytosine in the dNTP mixture. The methyltransferase enzyme M.SssI was added to the array to methylate CG dinucleotides. DNA methylation prevented digestion by the methylation-sensitive endonuclease HpaII, which cuts only the nonmethylated CCGG 4-mer, suggesting that the methylation reaction went to completion (Fig. 1A,B). The methylation-insensitive endonuclease MspI that cuts CCGG independent of CG methylation served as a control (Supplemental Fig. 1A,B).

Initially, we found that M.SssI incubation affected B-ZIP binding to both CG- and non-CG-containing 8-mers (Supplemental Fig. 1C,D). We reasoned that the M.SssI enzyme may stick to DNA and affect subsequent B-ZIP binding. Methylated arrays were thus incubated with proteinase K to digest any bound M.SssI. Following this treatment, methylation only affected B-ZIP binding to CG-containing sequences (Supplemental Fig. 1E,F). All data reported hereafter were obtained following the proteinase K treatment.

B-ZIP homodimers that bind methylated DNA

To monitor B-ZIP binding to the arrays, glutathione S-transferase (GST) was fused to the B-ZIP C terminus, and binding was measured using a fluorescent antibody to GST. Binding of eight B-ZIP domains to both unmethylated and methylated 40K arrays (Fig. 1C–G; Supplemental Fig. 2A–D) was reproducible (Supplemental Fig. 2E). We evaluated binding of B-ZIP proteins to 8-mers using a median signal intensity (Z-score) and a rank-order-based score (E-score) (Supplemental Fig. 3; Badis et al. 2009). To determine the length of binding sites, we calculated Z-scores for 5-mers representing a half-site (Supplemental Fig. 4) to 9-mers.

Before methylation, the best-bound 9-mer for all four CEBP family members (CEBPA, CEBPB, CEBPD, and CEBPG) is the canonical CEBP site TTGC|GCAAT (Supplemental Table 1A,B; Johnson 1993). CREB1, ATF4, JUN, and JUND bound the CRE (TGAC|GTCA) 8-mer (Fig. 1F,G; Supplemental Figs. 2A,B, 3E–H; Benbrook and Jones 1994). Only CREB1 shows specificity for a 9-mer (Supplemental Table 1C). JUN and JUND also bound the 7-mer (TGA^C/_GTCA) known as the TRE or AP-1 motif that could be extended to the 8-mer (TGA^C/_GTCAT) (Supplemental Table 1C). The ATF4 homodimer does not show strong DNA binding specificity, consistent with its in vivo function as a monomer heterodimerizing with other B-ZIP monomers (Ameri and Harris 2008). DNA binding following CG methylation was enhanced for CEBPA (P < 3.2 × 10⁻²) and CEBPB (P < 1.8 × 10⁻²) and inhibited for CREB1 (P < 2.1 × 10⁻²⁸), ATF4 (P < 1.0 × 10⁻⁷⁶), JUN (P < 8.7 × 10⁻⁸⁴), JUND (P < 7.1 × 10⁻⁸¹), CEBPD (P < 2.2 × 10⁻³), and CEBPG (P < 1.2 × 10⁻³), based on the best-bound 8-mer (Fig. 1C–G; Supplemental Fig. 2A–C; Supplemental Table 1D; Rishi et al. 2010; Warren et al. 2012).

CEBPB|ATF4 heterodimers bind methylated CGAT|GCAA

CEBPB is known to heterodimerize with ATF4 and bind a chimeric site TGAC|GCAA that contains both the CRE half-site (TGAC|G) and the CEBP half-site (C|GCAA) (Vinson et al. 1993). CEBPB is constitutively expressed in cells, while ATF4 protein is induced by cellular stress (Lu et al. 2004; Vattem and Wek 2004) and potentially heterodimerizes with CEBPB to bind new DNA sequences (Vinson et al. 1993). To examine how methylation affects DNA binding of the CEBPB|ATF4 heterodimer under conditions that mimic the biological situation, we labeled CEBPB with GST and added unlabeled ATF4. On unmethylated arrays, the CEBPB|ATF4 heterodimer preferentially binds the chimeric site TGAC|GCAA and TGAT|GCAA (Fig. 1H; Supplemental Table 2). The second sequence is the deaminated product of the chimeric sequence. Following methylation, the CEBPB|ATF4 heterodimer increased binding to CGAT|GCAA (P < 1.4 × 10⁻⁸⁸) (Supplemental Table 1D). The half-site T|GCAA is the deaminated CEBP half-site C|GCAA. Thus, T|GCAA is presumably bound by CEBPB, suggesting that ATF4 binds the other half-site (mCGAT|G) (Fig. 1H). This newly identified 5-mer CGAT|G site differs from the half-CRE site TGAC|G, the known binding half-site for ATF4, with the two pyrimidines, T and C, being switched, moving the CG dinucleotide from the center of the dyad to its flank. We also used GST-labeled ATF4 and unlabeled CEBPB and obtained similar results, suggesting that the GST tag is not contributing to these results (Supplemental Fig. 5A–D). Examination of 5-mer binding fails to identify CGAT|G as the top methylated sequence bound by CEBPB|ATF4 (Supplemental Fig. 5E,F), indicating that this 5-mer cannot function alone; instead, it needs the specificity from the CEBP half-site for heterodimer binding.

180K feature array: Quantitation of CG methylation on TF DNA binding

To evaluate the magnitude of changes in DNA binding following CG methylation, we designed a 180K array where each feature contains one of the 65,536 possible 8-mers. All non-CG 8-mers are present twice, and all CG-containing 8-mers are present three times in different background sequences (Supplemental Fig. 6A,B). All features have a T at the 5′ end and an A at the 3′ end of every 8-mer (TNNNNNNNNA). Arrays were methylated and digested with both HpaII and MspI, which confirmed that the methylation reaction went to completion (Fig. 2A,B; Supplemental Fig. 6C,D). Changes in fluorescence intensity are more uniform following enzymatic digestion compared to the 40K array (Fig. 1A,B; Supplemental Fig. 1A,B) because the CCGG 4-mer is always in the middle part of the probe, whereas in the 40K array, it can be anywhere along the length of the probe. Complementary 8-mers were evaluated separately because the flanking sequences could be part of the binding site and could break the complementarity (Supplemental Fig. 7). The 180K array design is able to quantify binding intensity to different sequences but is less optimal for identifying favored motifs because all 8-mers are flanked by the same nucleotide which may be part of the TFBS. On the 40K array, in contrast, the flanking sequences of multiple (16 or 32) occurrences of each 8-mer are random, minimizing any consistent influence on binding.

CEBPA binding on the unmethylated 180K array identified ATTGC|GCA as the best-bound 8-mer, which extends to the 9-mer ATTGC|GCAA as the “A” from the flanking sequences on the array is included (Fig. 2C). This 9-mer is the same sequence identified by the 40K array. The range in CEBPA binding between features is 45-fold, demonstrating sequence-specific DNA binding (Supplemental Table 3A). CEBPB shows a 267-fold range of binding, while ATF4 is less specific with only a 22-fold range (Supplemental Table 3A). Following methylation, the canonical CEBP 8-mer is again best-bound by CEBPA and CEBPB (Figs. 2C, 3A). Binding of ATF4 to the canonical CRE motif is diminished after methylation by sixfold (P < 5.3 × 10⁻⁶⁰) (Fig. 3B; Supplemental Tables 1D, 3B).

CEBPB|ATF4 binding on 180K arrays

The 180K arrays were also used to quantify methylation-dependent binding of the CEBPB-GST|ATF4 heterodimer (Fig. 3C–E). On unmethylated arrays, the best-bound sequence is CTGAT|GCA (Fig. 3C–E), the same sequence best-bound using the 40K array (Fig. 1H; Supplemental Fig. 5A). ATF4 addition decreased binding to the canonical CEBP motif TTGC|GCAA by sevenfold (Supplemental Table 3B), indicative of its dominant negative properties. On methylated arrays, the addition of ATF4 increased binding 10-fold to CGAT|GCAA (P < 4.0 × 10⁻²⁶–CEBPB-GST|ATF4; P < 4.7 × 10⁻¹¹–CEBPB|ATF4-GST) (Supplemental Table 1D), the same methylated sequence identified using the 40K array (Figs. 1H, 3D,E; Supplemental Fig. 5A; Supplemental Table 3B). We also used GST-tagged ATF4 with either CEBPB or GST-tagged CEBPB (Supplemental Fig. 8A,B) and obtained similar results again, suggesting that GST is not contributing to these results. The methylated 8-mer CGAT|GCAA can be extended to the 9-mer VCGAT|GCAA where V = A, C, or G (Supplemental Table 3C). CEBPB-GST|ATF4 binding on unmethylated vs. methylated arrays highlights the increase in binding to some sequences and the inhibition of binding to others, including the chimeric TGAC|GCAA (P < 1.78 × 10⁻²–CEBPB-GST|ATF4; P < 2.9 × 10⁻¹⁵–CEBPB|ATF4-GST) (Fig. 3E; Supplemental Fig. 8A; Supplemental Table 1D). Heterodimer binding to methylated CGAT|GCAA is more dramatically observed using ATF4-GST (Supplemental Figs. 5A,C, 8A,B).

EMSA shows CEBPB|ATF4 heterodimers bind methylated CGAT|GCAA

The electrophoretic mobility shift assay (EMSA) indicated that methylation of CGAT|GCAA increased CEBPB|ATF4 heterodimer binding by 10-fold, the same change observed using the 180K array (Fig. 4A; Supplemental Fig. 8C,D). The CEBPB|ATF4 heterodimer bound to DNA migrates slower than the CEBPB homodimer and faster than the ATF4 homodimer (Fig. 4A). EMSA experiments using hemi-methylated DNA probes (methyl cytosine on only one of the two strands) identified that the CEBPB|ATF4 heterodimer preferentially binds to methylated CGAT|GCAA and not the complementary TTGC|ATCG (Fig. 4A; Supplemental Fig. 8D).

The effect of methylation on binding of the CEBPB|ATF4 heterodimer to a chimeric motif TTGC|GTCA containing both the CEBPB half-site TTGC|G and the ATF4 half-site C|GTCA was examined (Supplemental Fig. 9A). As seen in both the 40K (P < 3.62 × 10⁻⁹⁷) (Supplemental Tables 1D, 2) and 180K (P < 4.69 × 10⁻¹¹) (Supplemental Tables 1D, 3B) arrays, methylation inhibits binding by two to threefold to the chimeric sequence (Supplemental Fig. 9A). Examination of hemi-methylated DNA reveals that methylation of one cytosine inhibits heterodimer binding, while methylation of the second cytosine enhances binding (Supplemental Fig. 9A).

EMSA using four palindromic CEBPB 10-mers

Both 40K and 180K arrays suggest that the CEBPB consensus binding site may be the ATTGC|GCAAT 10-mer (Supplemental Table 1) used in the crystal structure (Miller et al. 2003). CEBPB was mixed with four DNA probes containing different palindromic 10-mers NTTGC|GCAAN. ATTGC|GCAATT is best-bound by CEBPB at ∼5 nM, 10-fold better than the weakest 10-mer TTTGC|GCAAA, which bound at ∼50 nM (Fig. 4B; Supplemental Fig. 9B). Preferential binding to methylated sequences was observed when 10 mM Mg²⁺ was added in both the binding reactions and the polyacrylamide gel (Fig. 4C; Moll et al. 2002).

In vivo CEBPB binding

We examined CEBPB localization in the genome of primary female mouse dermal fibroblasts with a known methylome at 50× coverage for each strand (Supplemental Figs. 10A,B, 11A; Supplemental Table 4A) to identify whether methylated sequences are bound in vivo. CEBPB ChIP-seq data identified 7317 peaks that were most enriched for the canonical CEBP motif (Fig. 5A). To examine the unique part of genome, we focused on the masked genome (Jurka 2000). We compared in vitro CEBPB binding using methylated arrays with the enriched methylated 8-mers in the in vivo CEBPB ChIP-seq peaks (Fig. 5B). Methylated 8-mers well-bound in vivo contain the CEBP half-site C|GCAA, sequences well-bound on the arrays (Supplemental Fig. 11B). Many examples exist for TF binding functional noncanonical motifs in vivo. Our data indicate that, for CEBPB, only a small fraction of noncanonical sequences are bound in vivo. Eleven percent of the methylated TTGC|GCAA 8-mers in the genome are bound by CEBPB, fewer than expected (P < 10⁻⁴⁷, hypergeometric) (Fig. 5B; Supplemental Fig. 11C,D; Supplemental Table 4B). A larger fraction of unmethylated canonical CEBP 8-mers (54%) are bound, compared to methylated 8-mers (11%), even though in vitro CEBPB preferentially binds the methylated sequence, perhaps revealing that the unmethylated 8-mers are more accessible in vivo as is observed for unmethylated sequences which tend to be in DNase I-hypersensitive sites in vivo (Fig. 5B; Supplemental Fig. 11C,D; Supplemental Table 4B; Stadler et al. 2011). However, not all sequences well-bound on the array are bound in vivo. CEBPB ChIP-seq binding sites containing unmethylated and methylated 8-mers (TTGC|GCAA) are associated with distinct classes of GO terms; unmethylated 8-mers are associated with transcription regulation and signaling pathways, and the methylated 8-mers are associated with cell motility and lymphoid organ development (Fig. 5C). On the arrays, methylation inhibits CREB1 binding, and CREB1 ChIP-seq peaks enrich only for the unmethylated canonical CRE site TGAC|GTCA. Not a single methylated canonical CRE site is bound by CREB1 (P < 10⁻⁶⁷, hypergeometric), demonstrating the profound consequence of methylation on TF localization (Supplemental Fig. 11E,F; Supplemental Table 4C).

On the 180K arrays, CEBPB prefers to bind the methylated CEBP 9-mer TTGC|GCAAT. This is also observed in vivo (Supplemental Table 5A). CEBPB preferentially binds 39% of the methylated ATTGC|GCAAT 10-mers in the genome, compared to only 2% of methylated TTTGC|GCAAT 10-mers (P < 1 × 10⁻¹³, hypergeometric), suggesting the differences in binding observed using EMSA are in the range that are biologically significant (Fig. 5D; Supplemental Table 5B).

In vivo CEBPB and ATF4 binding after thapsigargin treatment

The CEBPB|ATF4 heterodimer binds methylated CGAT|GCAA at ∼15 nM (Fig. 4A). We evaluated if this methylated sequence is also bound in vivo. ATF4 is involved in several stress pathways and typically exists in cells as an untranslated mRNA that is translated into protein following cellular stress (Yukawa et al. 1999). To induce ATF4 protein expression, we treated primary female mouse dermal fibroblasts with thapsigargin (2 μM) for 3 h (Fig. 5E) and determined the genome-wide localization of CEBPB and ATF4 by ChIP-seq. We observed 7861 CEBPB peaks; 80% are not observed before ATF4 induction (Fig. 5F). The most enriched motif using RSAT peak-motifs (Thomas-Chollier et al. 2011) in CEBPB ChIP-seq peaks after ATF4 induction is TGAT|GCAA (Fig. 5G), the identical 8-mer best-bound by the CEBPB|ATF4 heterodimer on both 40K and 180K arrays (Figs. 1H, 3C–E, 6A,B; Supplemental Fig. 12A–D; Supplemental Table 6A; Adams 2007).

CEBPB peak size before and after ATF4 induction and the presence of four motifs (unmethylated and methylated canonical CEBP motif, best-bound 8-mer, TGAT|GCAA, and the methylated CGAT|GCA) was examined (Fig. 6A). The biggest CEBPB peaks before and after ATF4 induction do not contain the canonical CEBP motif. Intermediately bound peaks before ATF4 induction contain both the unmethylated and methylated canonical CEBP motif. Following ATF4 induction, there is a depletion of CEBPB binding to both the unmethylated and methylated canonical CEBP motif, revealing dominant negative properties for ATF4 (Fig. 6C; Supplemental Fig. 13A–F; Supplemental Table 6B). After ATF4 induction, CEBPB binding shifts to the methylated CGAT|GCA sequence identified on the arrays (Fig. 6A–C; Supplemental Fig. 13A–F). The enrichment of mCGATG occurrences in CEBPB ChIP-seq before and after ATF4 induction increased substantially from 0.62 to 1.56 (P < 2.46 × 10⁻²⁸, two-proportion z-test) (Fig. 6C; Supplemental Table 6B). The genes associated with the methylated CGAT|G bound by CEBPB after ATF4 induction are enriched for the protein amino acid phosphorylation (Supplemental Table 6C), which is required for activating the signaling pathway that inhibits protein biosynthesis after the unfolded protein response induced by ER stress (Harding et al. 2003).

ATF4 ChIP-seq (Miyamoto et al. 2011) identified 10,067 ChIP-seq peaks. The most enriched motif using RSAT peak-motifs is a chimeric sequence containing an ETS and a CRE motif (Fig. 6D; Chatterjee et al. 2012). There is less enrichment for specific 8-mers than observed for CEBPB, which is reminiscent of the microarray data where ATF4 showed less specificity (Supplemental Figs. 12C,D, 13C–F). A plot of CEBPB peak reads vs. ATF4 peak reads after treatment with thapsigargin highlights the presence of the four sequences previously mentioned in the peaks; the best-bound 8-mer is abundant (Fig. 6B) and enriched (Fig. 6C). The methylated CGAT|G is overrepresented in ATF4 ChIP peaks after thapsigargin treatment (Fig. 6C).

mRNA-seq of primary female mouse dermal fibroblasts before and after ATF4 induction using Illumina next-generation RNA sequencing identified differentially expressed genes (Fig. 6E). Genes that are commonly bound by CEBPB and ATF4 to TGAT|GCAA (the best-bound 8-mer on the array) at the promoters were highly expressed and up-regulated, or remained unchanged in the thapsigargin-treated primary female mouse dermal fibroblasts. The nearest gene to ATF4 peaks with mCGAT|G motifs shows both up- and down-regulation after thapsigargin treatment, suggesting variable effects on gene expression, including ER stress-induced genes that are up-regulated (Fig. 6E).

Cloning and expression of mouse B-ZIP proteins

The DNA binding domain (DBD) of B-ZIPs as defined in the Pfam (Finn et al. 2008) and SMART (Letunic et al. 2004) database (Supplemental Table 7) were cloned into the pETGEXCT (C-terminal GST) vector (Sharrocks 1994) either by RT-PCR from pooled mouse mRNA (1 ng), followed by ligation-independent cloning, or by gene synthesis followed by conventional cloning using NotI and SacI restriction sites of the pETGEXCT vector (Sharrocks 1994). After cloning, the inserts were sequence-verified (Supplemental Table 7). The proteins were expressed using in vitro translation (IVT) reactions. IVT reactions were performed using PURExpress In Vitro Protein Synthesis Kit (NEB) as suggested by the manufacturer's protocol. For each IVT reaction, ∼180 ng (40K) and 250 ng (180K) of plasmid DNA was used.

Design of the 180,000 (180K) feature PBMs

To evaluate binding affinities to each 8-mer, all possible 8-mers (65,536) were spotted 2–3 times in the probe sequence on a microarray glass slide. All the probes were 60 bp long, and the priming sequence was 24 bp long. Each oligonucleotide sequence on the microarray slide included 14 bp of fixed nucleotides on both sides of variable 8-mers. We appended 24 nt (GGACACACTTTAACACATGGAGAG) to each of the 3′ends of probe sequence which is complementary to the primer sequence. These microarrays were designed by Agilent Technologies in their “4X180K” format with probes attached to the glass slide at the 3′ end.

Design of the 40,000 (40K) feature PBMs

The 40K array design also consists of probe sequences which are 35 bp long, and 25 nt are appended at the 3′ end of probe sequences which are complementary to primer sequence for double-stranding. The design of this array is based on de Bruijn sequence, and each 8-mer occurs 32 times, and the palindromes occur 16 times. The sequence detail of the 40K array has been described in Lam et al. (2011). Details of the design have been described elsewhere (Berger et al. 2006, 2008; Philippakis et al. 2008).

Microarray double-stranding

The single-stranded oligonucleotide microarrays were double-stranded by primer extension as described in Badis et al. (2009). Briefly, the primer extension reactions consisted of 1.17 μM HPLC-purified common primer (Integrated DNA Technologies), 40 μM dATP, dCTP, dGTP, and dTTP (GE Healthcare), 1.6 μM Cy3 dCTP (GE Healthcare), 40 Units Thermo Sequenase DNA Polymerase (USB), and 90 μl 10× reaction buffer (260 mM Tris-HCl, pH 9.5, 65 mM MgCl2) in a total volume of 900 μl. The reaction mixture, microarray, stainless steel hybridization chamber, and single chamber gasket cover slip (Agilent Technologies, Inc.) were prewarmed to 85°C. After a 2-h incubation (85°C for 10 min, 75°C for 10 min, 65°C for 10 min, and 60°C for 90 min), the hybridization chamber was disassembled in a glass staining dish in 500 mL PBS/0.01% Triton X-100 at 37°C. The microarray was transferred to a fresh staining dish, washed for 5 min in PBS/0.01% Triton X-100 at 37°C, washed once more for 1 min in SSC 0.06× at 20°C. The double-stranded slides were scanned to quantify the amount of incorporated Cy3-conjugated dCTP (Agilent Technologies).

Methylation of double-stranded microarray

The methylation of the double-stranded microarray was performed using 10 μl of CG methyltransferase enzyme M.SssI (20 units/μl) (NEB), 1μl of S-adenosylmethionine, and 15 μl of 10× NEB buffer 2 in a total volume of 150 μl at 37°C for 3 h. The arrays were washed 3 × 5 min using PBS with 0.5% Tween-20 on a rotator and once with PBS for 2 min, and finally rinsed in PBS in a 500-mL staining dish. The methylated microarrays were stripped in a 50-mL stripping solution consisting of 10 mM EDTA, 10% SDS, and 290 Units of protease (Sigma), shaking at 200 rpm in a Coplin jar at 37°C for 16 h. Microarrays were finally washed as described previously.

Protein binding reaction

The protein binding reactions were carried out as described by Badis et al. (2009). Briefly, the double-stranded microarrays were blocked with 4% nonfat dried milk (Sigma) for 1 h. Microarrays were then washed once with PBS with 0.1% (vol/vol) Tween-20 for 5 min and once with PBS with 0.01% Triton X-100 for 2 min. Sixteen microliters of IVT reactions were added to make a total volume of 150 μl protein binding reaction for 40K array and 25 μl for 180K array, containing PBS with 2% (wt/vol) milk, 51.3 ng/μl salmon testes DNA (Sigma), and 0.2 μg/μl bovine serum albumin (NEB), and incubated for 1 h at 20°C. Preincubated protein binding mixtures were applied to individual chambers of 40K and 180K arrays and incubated for 1 h at 20°C. Microarrays were washed with squeeze bottle (40K) and in a Coplin jar (180K) once with 0.5% (vol/vol) Tween-20 in PBS for 3 min, once with 0.01% Triton X-100 in PBS for 2 min, and then finally washed with PBS. Alexa Fluor 647-conjugated GST antibody (Invitrogen) was applied to each chamber and incubated for 1 h at 20°C. Finally, microarrays were washed twice with PBS with 0.05% (vol/vol) Tween-20 for 3 min each, and once in PBS for 2 min. Every protein in this study was assayed in duplicate, once on each of our two separate microarray designs described above.

Image quantification and analysis of microarray data

Protein-bound microarrays were scanned to detect Alexa Fluor 647-conjugated anti-GST using at least two different laser power settings to best capture a broad range of signal intensities and ensure signal intensities below saturation for all spots. Microarray images were analyzed using ImaGene (BioDiscovery Inc.), bad spots were manually flagged, and the extracted data were used for further analysis. To estimate the relative binding affinities of proteins using an ∼170,000 feature array, we averaged the signal median intensities for the spots containing each 8-mer. However, on a 40,000 feature array, to estimate the relative preference for each 8-mer, two different scores were calculated: The Z-score was calculated from the average signal intensity across the 16 or 32 spots containing each 8-mer; the ‘E-score’ (for enrichment) is a rank-based, nonparametric statistical measure that is invariant to protein concentration (Berger et al. 2006).

Estimation of the significance of 8-mer binding to methylated vs. unmethylated arrays

We estimated the significance of the change in the strength of binding of a given transcription factor to each 8-mer on the methylated and unmethylated arrays. For each 8-mer, we compared its Z-scores (for the standard 40K PBMs) or median 8-mer intensities (for the 180K PBMs). Since the range of scores can differ between experiments, we first transformed the methylated 8-mer values by performing a standard linear regression onto the values of the unmethylated array for each transcription factor. We then calculated a single statistic for each 8-mer representing the change in binding strength between the two arrays:

where is the transformed methylated score, and is the score on the unmethylated array. To estimate the significance of , we compared its value to that obtained on replicates of the unmethylated arrays:

where is the score on the replicate unmethylated array. As a conservative estimate of the distribution of for the preferentially bound 8-mers of each transcription factor, we calculated the mean and standard deviation of across the 100 highest-scoring 8-mers for the given experiment. Similar results were obtained when using the top 50 or top 200 8-mers (data not shown). The final significance of the difference in binding between the methylated and unmethylated arrays to each 8-mer for each transcription factor was then calculated as:

where is the mean of across the 100 highest-scoring 8-mers for the given experiment, and is the standard deviation. The resulting Z-scores were transformed into P-values using the standard normal distribution.

Electrophoretic mobility shift assay

EMSA was performed as described previously (Rishi et al. 2010). The proteins were either in vitro translated using PURExpress In Vitro Protein Synthesis Kit (NEB) according to the manufacturer's instructions or were purified as previously described (Ahn et al. 1998). Proteins were mixed with 7 pM ³²P-end-labeled double-stranded oligonucleotides in the gel shift buffer (0.5 mg/mL BSA, 10% glycerol, 2.5 mM DTT, 12.5 mM K₂HPO₄-KH₂PO₄, pH 7.4, 0.25 mM EDTA). The final volume of the reaction was adjusted to 20 μl. The reactions were incubated at 37°C for 20 min, followed by cooling at room temperature for 5 min before loading. Ten-microliter samples were resolved on 7.5% or 12% PAGE at 150 V for 1.5 h in the 1× TBE buffer (25 mM Tris-boric acid, 0.5 mM EDTA). Sequences of oligonucleotides used for EMSA experiments are listed in Table 1 (binding sites underlined).

Mouse primary dermal fibroblasts culture

Dermal fibroblasts were cultured from newborn wild type according to the protocol given elsewhere (Rishi et al. 2010). Primary dermal fibroblasts were seeded at a density of one mouse dermis per 10-cm-dish or equivalent in DMEM/F12: GlutaMAX medium (Invitrogen) with 10% FBS. For ATF4 induction, the primary cells were treated with 2 μM of thapsigargin (Sigma) for 3 h.

Western blotting

Protein extracts were prepared in RIPA buffer containing 50 mM Tris, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EGTA, 5 mM EDTA, 10 mM NaF, 1 mM β-glycerophosphate, 1 mM sodium vandate, and protease inhibitor (Roche). Equal amounts of proteins were resolved on NuPAGE 4%–12% Bis-Tris gradient gels (Invitrogen) and transferred onto PVDF membranes (Hybond-P, Amersham Biosciences). Membranes were blocked in 5% skimmed milk for 1 h at room temperature and then incubated with primary antibody at 4°C overnight with shaking. After washing using PBST 3 × 5 min, the blots were incubated for 1 h with secondary antibodies against rabbit or mouse IgG (Amersham Biosciences, 1:5000) and washed 3 × 5 min. Blots were developed using ECL plus Western Blotting detection system (Amersham Biosciences). The following primary antibodies were used: anti-CEBPB (sc-150; Santa Cruz Biotechnology), anti-ATF4 (sc-200; Santa Cruz Biotechnology), and monoclonal anti-β-actin (sc-47778; Santa Cruz Biotechnology).

Chromatin immunoprecipitation (ChIP) sequencing

Primary cultured cells either untreated or treated with thapsigargin were chemically cross-linked for 10 min by adding 0.6% formaldehyde (Sigma) directly to the medium. The cross-linking reaction was stopped by adding 125 mM glycine, and dishes were swirled for 5 min at room temperature. Cells were washed twice with ice-cold PBS and harvested in ice-cold PBS containing protease inhibitor (Roche). A total of 10⁷ cells were pelleted by centrifugation at 4°C for 5 min at 300g. Four times 300 μl of sonicated chromatin preparation was incubated overnight with CEBPB (sc-150; Santa Cruz) or ATF4 (sc-200; Santa Cruz) or CREB antibody (sc-186; Santa Cruz). Immunocomplexes were captured using protein G agarose beads (Invitrogen) and washed twice with the buffer containing 2 mM EDTA, 100 mM Tris-Cl, pH 8.0, and 0.18% Sarkosyl, and four times with the IP buffer (100 mM Tris-Cl, pH 8.5, 500 mM LiCl, 1% NP40, 1% deoxycholic acid). After incubating with RNaseA and Proteinase K, DNA was eluted using QIAquick PCR Purification Kit. Purified DNA were used to prepare the library for Illumina high-throughput sequencing using Illumina Single End ChIP-seq Sample Preparation Kit as described in the manufacturer's protocol. Libraries were sequenced to generate 35-bp single-end reads using Illumnia GAII sequencing machines. We used the Model-Based Analysis of ChIP-seq (MACS) algorithm with default parameters for detecting the ChIP-seq peaks of CEBPB before and after ATF4 induction and ATF4 after ATF4 induction by thapsigargin (Zhang et al. 2008)

RNA-sequencing of dermal fibroblasts before and after ATF4 induction

Total RNA was isolated from the mouse primary dermal fibroblasts before and after ATF4 induction using 3 h of treatment with 2 μM of thapsigargin. Purified RNA was used for generating the mRNA-seq library using the Illumina mRNA-seq kit as described in the manufacturer's protocol. Data analysis was performed using Cufflinks software with the default parameters as suggested in the Cufflinks manual (Trapnell et al. 2010). Transcript abundances were reported in fragments per kilobase of transcript per million fragments mapped (FPKM) with arbitrary units.

Determination of whole genome DNA methylation

Genomic DNA was isolated from cultured primary dermal fibroblasts and used for bisulfite sequencing using the protocol described previously (Lister et al. 2009). Approximately 10 μg of genomic DNA was sonicated to ∼300 bp using the Covaris S2 System. Sonicated DNA was purified using Qiagen DNeasy minielute columns (Qiagen). Each sequencing library was constructed using the Illumina paired end DNA sample preparation kit (Illumina Inc.) according to the manufacturer's instructions, with the following modifications: Illumina methylated adapters were used in place of the standard genomic DNA adapters. Ligation products were purified with AMPure XP beads (Beckman). Four times 500 ng of DNA were bisulfite-treated using the EpiTect Bisulfite Kit (Qiagen) following the manufacturer's guidelines, followed by PCR amplification using the Phusion Taq using the following PCR conditions: 2 min at 95°C, 4 cycles of 15 sec at 98°C, 30 sec at 60°C, 4 min at 72°C, and 10 min at 72°C. Libraries were sequenced using the Illumina HiSeq 2000 (Illumina) up to 101 cycles. For mapping, cross validation and data analysis, NIH Helix and biowulf clusters were used with our standardized custom algorithms for the whole genome bisulfite sequencing pipeline. Briefly, we aligned bisulfite-treated sequences against a reference genome (mouse, mm9) in a single pass that aligns against both C-T and G-A in silico indexed reference sequences using Novoalign (Novocraft Technologies). The base calls per reference position on each strand were used to identify methylated cytosines in the context of CpG, CHG, and CHH.