DNA methylation—the addition of a methyl group to a cytosine or adenine base within DNA—has a key role in regulating the expression of genes as proteins. It contributes to processes such as X-inactivation, in which one copy of the X chromosome is silenced in females, and genomic imprinting, in which the expression of a gene depends upon which parent it was inherited from. DNA methylation has also been implicated in the development of cancer. However, the molecular mechanisms by which it produces these effects are not fully understood.

In mammals, the methylation of CpG sites—which consist of a cytosine base next to a guanine base—is typically thought to reduce gene expression by preventing proteins called transcription factors from binding to regions of DNA called promoters. This can occur directly if methylation disrupts interactions between the DNA and the transcription factors, or indirectly if other proteins that bind to the methylated DNA compete with the transcription factors for binding sites. However, only a small number of proteins that bind to methylated DNA have so far been identified.

Now, Hu et al. have screened the entire family of roughly 1300 human transcription factors and 210 co-factors (proteins that interact with transcription factors) for their ability to bind to some 150 different stretches of methylated DNA. They found that 47 of the proteins could bind to methylated CpG sites, with the majority showing a preference for specific DNA sequences. Moreover, some transcription factors and co-factors bind to methylated and non-methylated DNA targets with distinct sequences. These two types of binding are largely independent, as illustrated by the fact that mutations that prevent a transcription factor called KLF4 from binding to methylated DNA do not prevent it binding to unmethylated DNA, and vice versa.

The work of Hu et al. suggests that methylated cytosine can effectively act as a ‘fifth base’—in addition to adenine, cytosine, guanine and thymine—and emphasizes the importance of DNA methylation for regulating gene expression.

DNA methylation is an ancient and major epigenetic modification that plays an important role in key biological processes, including genomic imprinting, X-chromosome inactivation, suppression of transposable elements, and carcinogenesis (Jaenisch and Bird, 2003; Egger et al., 2004; Robertson, 2005; Feinberg, 2007; Reik, 2007). In higher eukaryotes, methylation of CpG sites, especially at promoter regions, is generally considered as the hallmark of gene silencing (Baylin, 2005). The molecular consequence of CpG methylation is generally believed to disrupt TF–DNA interactions either directly (Nan et al., 1998), or indirectly by recruiting sequence-independent methylated DNA-binding proteins that occupy the methylated promoters and compete for the TF binding sites (Boyes and Bird, 1991). So far, only MeCP2, MBD1, MBD2, and a few zinc finger proteins have been identified as bona fide methylated DNA-binding proteins (Lewis et al., 1992; Meehan et al., 1989; Filion et al., 2006; Bartke et al., 2010; Bartels et al., 2011; Quenneville et al., 2011; Spruijt et al., 2013). It is unclear whether the methylated DNA binding activity is widespread among different TF subfamilies. Furthermore, the transcriptional regulatory activity of these methylation-dependent TF–DNA interactions has not been explored. Finally, the structural basis of these methylation-dependent TF–DNA interactions remains elusive.

Here we employed a protein microarray-based approach to characterize the entire human TF repertoire for their direct binding capacity to a large set of methylated DNA motifs. In this work, we simultaneously tested the methylated DNA-binding activity for the majority of human TFs. Since we examined each methylated DNA motif individually on the protein microarray, we were able to determine the specific sequences surrounding the methylated CpG that is recognized by TFs. Furthermore, we demonstrated that the methylated- and unmethylated-DNA binding activities of KLF4 could be decoupled by protein mutagenesis.

We selected a total of 154 DNA motifs (Supplementary file 1A) with the following characteristics: (1) predicted to be potential TF-binding sites in promoter regions of the human genome (Xie et al., 2005; Yu et al., 2006; Elemento et al., 2007); (2) representative of a subset of the 460 DNA motifs that have been examined for protein-binding activity in our previous study (Hu et al., 2009; Xie et al., 2010); and (3) carrying at least one CpG site. Because our goal was to identify DNA methylation-dependent binding activity, we developed a competition assay on a protein microarray to identify TFs that prefer DNA motifs carrying mCpGs (Figure 1A). Each synthesized double-stranded DNA motif was end-labeled with Cy5 (Hu et al., 2009) and converted to the double-stranded mCpG form using a bacterial CpG DNA methylase SssI. Each methylated motif was incubated with the protein microarray in the presence of its unlabeled (cold), unmethylated counterpart in 10-fold excess. The competition assay was first tested on a pilot array, comprised of MBD1, MBD2, MeCP2, and several negative control proteins. As illustrated in Figure 1B, strong binding signals were observed in a protein concentration-dependent manner, while control proteins, such as GST and BSA, did not show any detectable signals, confirming the reliability of the competition assay.

Figure 1 with 8 supplements see all

Download asset Open asset

We then probed for direct interactions between 154 methylated motifs and the human TF repertoire using the TF protein microarray, which includes 1321 TF and 210 co-factor proteins (Supplementary file 1B). These full-length human proteins were overexpressed and individually purified in yeast as N-terminal GST fusion proteins as previously reported (Hu et al., 2009). The quantity and purity of the purified proteins were examined with Coomassie and silver stains against a random set of purified proteins (data not shown). Each of the labeled 154 methylated motifs was separately incubated on the human TF protein microarray in the presence of the unlabeled counterparts in 10-fold excess. Each reaction was performed in duplicate to ensure reproducibility. After a washing step, binding signals were acquired and processed using the GenePix software (Figure 1—figure supplement 1). After the binding signals were further normalized by correcting local background signals with a 9 × 9 sub-grid, Z-scores of each protein on the array were determined as previously described (Hu et al., 2009). A Z-score of 3 was used as the cutoff to identify the positives. Correlation analysis of the binding signals showed high reproducibility between each duplicated assay (Figure 1—figure supplement 2). Of the 154 methylated motifs examined, 150 (97%) showed significant binding signals to at least one protein on the microarrays (median = 8), suggesting binding to methylated DNA is prevalent, at least in this in vitro competition assay (see ‘Materials and methods’, Figure 1—figure supplement 3). In addition, 41 TFs and 6 TF co-factors (3% of all factors examined) showed mCpG-dependent binding activity (Figure 1C). Among these, 15% showed a broad binding activity to over 50% of all methylated motifs tested, suggesting that these proteins are not sensitive to the sequence context surrounding the mCpG site. For example, two zinc finger TFs, ZNF114 and ZNF416, could recognize almost all of the motifs tested in the competition assay. However, the majority of the 47 mCpG-binding proteins showed mCpG- and sequence-dependent binding activity, with 22 TFs exhibiting binding signaling to fewer than three motifs (Figure 1—figure supplement 4). Notably, the mCpG binding activity is widespread among various TF subfamilies, such as the zf-C2H2, Homeobox, bHLH, Forkhead, bZIP, and HMG box subfamilies (Figure 1C); no TF subfamily was found significantly enriched in the hit list (Figure 1—figure supplement 5). On the other hand, consistent with the important role of DNA methylation in the development of cancer (Egger et al., 2004; Reik, 2007), there was a significant enrichment for known oncogenes and tumor suppressors (10/47; p<0.01; hypergeometric model) and factors involved in tissue development (25/47; p<0.015; hypergeometric model; Figure 1C). Together, these results suggest that methylation-dependent direct TF–DNA interaction is a widespread phenomenon among various TF subfamilies in humans.

To validate the protein microarray results, we selected 11 TFs with different mCpG-dependent DNA-binding behaviors (i.e., sequence-dependent or -independent). Among them, six (e.g., ARNT2) bound to fewer than three methylated motifs, while the others (e.g., HOXA5) bound to at least 66 motifs (Supplementary file 1C). Using an electrophoretic mobility shift assay (EMSA), we confirmed that 8 of the 11 tested TFs could readily form a protein–DNA complex with their corresponding methylated motifs, but not with the unmethylated form (Figure 1D; Figure 1—figure supplement 6). This result indicates either a 27% of false positive rate of the protein microarray assay or a higher sensitivity of protein microarray than EMSA. While the binding activities to the mCpG-carrying motifs could be readily competed off with the unlabeled mCpG-containing analogs in 10-fold excess, the unmethylated counterparts showed no obvious impact, indicating that the complex formation between these TFs and specific DNA motifs is not non-specific but requires CpG methylation (Figure 1E).

To determine whether mCpG-dependent TF–DNA interactions could convey transcriptional activity in vivo, we selected HOXA5 and DIDO1 to perform cell-based luciferase assays. Motifs M305 and M24 recognized by HOXA5 and DIDO1, respectively, were separately cloned into the promoter region of a CpG-free luciferase construct (pCpGL) (Klug and Rehli, 2006), which was then methylated with SssI and transfected with the HOXA5 and DIDO1 expression constructs at different concentrations into a mammalian cell line. The complete methylation of the motifs before transfection was confirmed using Sanger-bisulfite sequencing (Figure 1—figure supplement 7). Unmethylated luciferase construct was used as a negative control. Comparing with the luciferase activity of the negative controls, we found that both HOXA5 and DIDO1 demonstrated dose- and methylation-dependent enhancement of the luciferase activity (Figure 1F), indicating that the two TFs could specifically bind to the methylated promoters and enhance the downstream gene transcription.

We then asked whether these mCpG-binding TFs also bound to the same consensus motifs in an unmethylated form. Since our previous genome-wide analysis of human TF–DNA interactions included these 150 unmodified CpG-containing motifs (Hu et al., 2009), we integrated both protein microarray datasets. We discovered that 17 TFs recognize both methylated and unmethylated motifs (Figure 2A). Of the 435 protein–DNA interactions (PDIs) examined, only a small fraction (4%) exhibited motif-binding activity irrespective of the CpG methylation status (Figure 2A; yellow bars). Even in these cases, TF binding to methylated forms is much stronger than to the unmethylated forms based on the inability of non-methylated analogs to compete effectively even when present in 10-fold excess. Interestingly, 321 (74%) and 97 (22%) PDIs exhibit specificity to methylated (Figure 2A; red bars) and unmethylated motifs (Figure 2A; blue bars), respectively. To evaluate sequence specificity for the newly discovered methylation-dependent PDIs, we determined the consensus methylated motif sequences for the 17 TFs. The comparison with their known consensus sequences indicated that, in most cases, the sequences of methylated and unmethylated motifs recognized by the same TFs are distinct (Figure 2A; right panel; ‘Materials and methods’). For example, nuclear factor I/C (NFIC) binds to a methylated motif M209, GTmCpGCC, while its known consensus sequence is TTGGC, suggesting that many TFs are of dual-specificity, recognizing methylated and unmethylated DNA motifs of distinct sequences.

Figure 2 with 4 supplements see all

Download asset Open asset

To experimentally confirm this apparent dual-specificity, we performed EMSA assays with two motifs, recognized by the same TF, in both methylated and unmethylated forms (Figure 2B). For instance, KLF4 could form a protein–DNA complex with the methylated motif M197 (TCCmCpGCCC), but not with its unmethylated form (Figure 2B). In contrast, methylation on motif M412 (GCTTTTACG) disrupted its interaction with KLF4 (Figure 2B). The same phenomenon was confirmed for three additional TFs, namely TFAP2A, ARID3B, and ZMYM3 (Figure 2B).

These results raised an interesting question: does binding by a dual-specificity TF to one motif affect binding to the other of different methylation status? By performing a competition EMSA assay we observed two scenarios for different TFs (Figure 2C–F). In the first scenario, binding to one motif interferes the binding activity to the other motif. For example, the formation of protein–DNA complex between ARID3B and a methylated motif M319 (AAAmCpGCTTCC) could be readily competed off with the unmethylated motif M47 (GTGGGCGAAA) in 10-fold excess, and vice versa (Figure 2E), suggesting that ARID3B either uses the same DNA-binding domain to interact with both motifs, or binding to one motif inhibits binding to the other, presumably via conformational changes or steric hindrance (Figure 2C). The specificity of these assays was further confirmed with a competition EMSA assay with ARID3B and ZMYM3 (Figure 2—figure supplement 1). In the second scenario, binding activities of a dual-specificity TF are independent of each other (Figure 2D). For instance, KLF4–DNA complex formation with the methylated motif M197 was not affected by adding unmethylated motif M412 in 10-fold excess in the competition assay, suggesting that it may use two different domains to distinguish methylated from unmethylated motifs (Figure 2F; left panel). As expected, addition of 10-fold excess of methylated M197 could readily compete off the complex formation of KLF4 with labeled M197 (Figure 2F; left panel). The same results were observed for TFAP2A (Figure 2F; right panel). The binding specificities of KLF4 and TFAP2A were further confirmed by adding both methylated and non-methylated competitor DNA, resulting in disruption of the complex formation (Figure 2—figure supplement 2). To facilitate direct comparison, we reproduced all combinations of the competition EMSA assays for KLF4 in parallel on the same gel. The results confirmed its dual-specificity in a non-competitive fashion (Figure 2—figure supplement 3). The fact that KLF4 could form protein–DNA complexes with both motifs equally well, regardless of the presence of the competitors, also suggested that it is highly unlikely that these two motif sequences could bind to the same domain at different affinities.

We next employed an oblique incidence reflectivity difference (OIRD) system (Landry et al., 2004; Zhu et al., 2007; Fei et al., 2011) to determine binding affinity (i.e., K_D values) of ZMYM3, TFAP2A, and KLF4 with their corresponding methylated DNA motifs M203, 213, and M197, respectively. As a comparison, the well known methylated DNA-binding protein, MBD2b, was also included. Because the OIRD system can monitor binding events in a real-time, label-free fashion, we therefore obtained the K_on and K_off values, and determined the K_D values of ZMYM3, TFAP2A, and KLF4 as 460 nM, 399 nM, and 479 nM, respectively (Figure 2G; Figure 2—figure supplement 4A–D). As expected, none of them showed any significant binding activity to the unmethylated DNA motifs in the OIRD measurements, confirming our previous observations. On the other hand, the K_D values of MBD2b measured against the same motifs are in close range to the three TFs tested (Figure 2—figure supplement 4A), suggesting that these TFs could bind to methylated DNA motifs almost as well as MBD2b (Figure 2H).

To further dissect the molecular mechanism of this phenomenon, we focused on KLF4 to identify the key residue(s) responsible for interacting with the methylated cytosine, in order to decouple its dual-specificity. Intriguingly, KLF4 encodes a classic zf-C2H2 domain, a truncated and a full zf-H2C2 domain at its very C-terminus (Figure 3A and Figure 3—figure supplement 1). Because the X-ray structure of KLF4 bound to methylated CpG sequences is not available, we used a modeling approach. These efforts were facilitated by crystal structures of other, unrelated 5mC-binding proteins (e.g., MeCP2 and ZFP57) (Ho et al., 2008; Liu et al., 2012), as well as the adduct composed of mouse KLF4 and unmethylated DNA, which was employed as a modeling template. Using the existing KLF4 structure (Schuetz et al., 2011), we identified Arg458 and Asp460 in the last H2C2 zinc finger as a candidate 5mCpG recognition motif. In the KLF4 model, DNA binding is stabilized by a 5mC-Arg-G via improved van der Waals interaction and an Arg458-Asp460 salt-bridge (Figure 3A). Binding is further stabilized by a CH•••O (H₂O-5mC) H-bond contact of the methylated cytosine on the complementary strand (Figure 3B). Interestingly, the local atomic environment of the KLF4 model is rather similar to those of MeCP2 and ZFP57 in their mCpG-bound forms (Figure 3—figure supplement 2). Therefore, Arg458 and Asp460 were predicted to be crucial for KLF4 and mCpG-dependent interaction (Figure 3—figure supplement 1).

Figure 3 with 4 supplements see all

Download asset Open asset

To test our prediction, we generated a series of KLF4 mutants, including three single (i.e., R458A, R458K, D460A), one double (R458A::D460A) mutants, and one truncation (Δ432). EMSA assay with purified KLF4 mutant proteins showed that all point mutations at either or both residues completely abolished KLF4’s binding activity to methylated motif M197 (CCmCpGCC), while none of them had any detectable impact on the binding activity to the unmethylated motif M412 (TACpGCC; Figure 3C and Figure 3—figure supplement 3). The R458A and R458K mutations showed similar impact on mCpG-dependent binding activity, suggesting that charge on the residue is not critical. Similarly, KLF4 truncation showed no binding activity to methylated motif M197 (CCmCpGCC), but still maintained its binding activity to the unmethylated motif M412 (TACpGCC; Figure 3—figure supplement 3), indicating that the C2H2 zinc finger in KLF4 mediates KLF4’s interaction with unmethylated motif M412.

To examine whether KLF4’s binding activity to M197 (CCmCpGCC) is important for transcription regulation, we employed cell-based luciferase assays using both wild-type (WT) and mutated KLF4 expression constructs. Indeed, when co-transfected with the pCpGL construct carrying methylated motif M197 in the promoter, WT KLF4 showed dose-dependent increase of the luciferase activity (Figure 3D). This enhancement of transcription is also methylation-dependent, because WT KLF4 could not increase luciferase activity when motif M197 was not methylated in the construct (Figure 3D). More importantly, both KLF4^R458A and KLF4^D460A constructs completely lost their ability to modulate transcription from the pCpGL construct carrying methylated motif M197; whereas they both showed dose-dependent enhancement of gene transcription in the luciferase assay, like WT KLF4, when co-transfected with the pCpGL construct carrying unmethylated motif M412 (Figure 3E). Consistent with the EMSA results, neither WT nor mutant KLF4 constructs showed any significant increase of transcription when motif M412 was methylated in the construct (Figure 3E). Western blot analysis demonstrated an equal transfection efficiency of WT KLF4, KLF4^R458A and KLF4^D460A constructs (Figure 3—figure supplement 4). Altogether, these results suggested that KLF4 could bind to methylated and unmethylated CpG sites in different sequence contexts, resulting in activation of downstream gene expression in heterologous cells in vivo, and that these dual specificities are independent and achieved via two functionally separable domains encoded by KLF4.

Finally, we asked whether KLF4 binding preferences to both methylated and unmethylated motifs that we identified in vitro reflect sequences preferred in vivo. Since KLF4 is well known to play an important role in embryonic stem cell (ESC) maintenance, we therefore examined whether mCpG-dependent KLF4–DNA interactions occur in human ESCs (H1). First, we performed bioinformatics analysis of in vivo KLF4-binding sites in H1 cells based on their methylation status (Figure 4—figure supplement 1). We superimposed the published in vivo KLF4 binding sites, as determined by the ChIP-seq approach (Lister et al., 2009), with the published methylome dataset of single-base resolution obtained with whole-genome bisulfite sequencing in H1 hESCs (Lister et al., 2009). We identified numerous cases in which KLF4 tends to bind to those genomic regions containing a methylated CCCpGCC sequence, consistent with our protein microarray results (Figure 4—figure supplement 1; lower panel). Among the KLF4-binding sites containing at least one CpG, we found that the majority of these sites could be categorized into two groups across the whole-genome: 48% have high methylation levels of over 80% and 38% have methylation levels lower than 20% (Figure 4A). We identified statistically significant consensus motifs in methylated group (‘Materials and methods’; Supplementary file 1D). The top consensus motif, CmCpGC, discovered in the high methylation group is embedded in motif M197 (CCmCpGCC) that was recognized by KLF4 in the protein microarray assays (Figure 4A). Taken together, these integration analyses demonstrate that KLF4’s preference to methylated consensus sequence identified in vivo is highly similar to that determined in vitro.

Figure 4 with 3 supplements see all

Download asset Open asset

To provide direct experimental evidence that KLF4 binds to endogenous CCmCpGCC in H1 hESCs in vivo, we performed KLF4 ChIP coupled with bisulfite sequencing (Figure 4B), using primer pairs designed on the basis of the above bioinformatics analysis. First, chromatin IP (ChIP) demonstrates specific and direct binding of KLF4 to five selected regulatory loci based on KLF4 ChIP-seq signals (Figure 4C; Figure 4—figure supplement 2). Next, quantitative real-time PCR (q-PCR) also showed the high enrichment (≥9-fold) on the specific loci after KLF4 ChIP (Figure 4D). KLF4 ChIP-bisulfite sequencing analyses showed that KLF4 could readily ChIP those loci (i.e., L1 and L2 in Figure 4E) with TACpGCC at two completely unmethylated fragments, and with CCCpGCC at ∼100% methylation level in other two loci tested (L3 in Figure 4E, and Figure 4—figure supplement 3), indicating that KLF4 binds to both unmethylated and highly methylated CpGs in different sequence contexts in vivo. To rule out the possibility that other sites within the ChIP’ed fragments might be responsible for KLF4 binding, we searched the published dataset (Lister et al., 2009) and did not find any other KLF4 binding peaks within the surrounding 500 bp, supporting that KLF4 indeed binds to these unmethylated or methylated sites. We also examined KLF4 methylated target sites that exhibit medium methylation levels (e.g., ∼50%) based on the published methylome dataset. Importantly, KLF4 ChIP-bisulfite sequencing analyses showed that the methylation level of CCCpGCC consensus was significantly increased from 30% and 55% of the genomics inputs to 65% (p=0.001) and 73% (p=0.04) in the ChIP’ed samples for two examined regions, respectively (L4 and L5 in Figure 4F; upper panel). As predicted, those nearby CpG sites not embedded in the KLF4 consensus did not show any significant increase in methylation (Figure 4F; lower panel). Together, the above analyses provided direct evidence that KLF4 preferentially binds to CCmCpGCC consensus sequence in vivo.

In this study we have identified numerous human TFs across various subfamilies that showed mCpG- and sequence-dependent binding activity, a much more prevalent phenomenon than previously appreciated (Karlsson et al., 2008; Rishi et al., 2010; Quenneville et al., 2011; Liu et al., 2012). It is likely that more proteins with the similar activity are yet to be discovered (Figure 1—figure supplement 8), because (1) we only surveyed a tiny fraction of methylated CpG space; (2) we did not screen for TFs that can recognize methylated CHG or CHH sequences in this study, which are present in several cell types including ESCs, induced pluirpotent stem cells, germline and neural cells (Ramsahoye et al., 2000; Ziller et al., 2011); (3) we did not include in the screen any hemimethylated DNA motifs, which occurs in vivo on newly synthesized DNAs during replication; and (4) some of the unconventional DNA-binding proteins we discovered in an earlier study might also encode this property (Hu et al., 2009).

We estimated false positive and false negative rate for our protein microarray assays. Since 8 of the 11 methylated DNA-binding TFs were confirmed by EMSAs, we estimated that this protein microarray strategy might produce ∼27% of false positives. Our approach also showed some false negatives because some known methylated DNA binding proteins were not identified in this study (Mann et al., 2013; Spruijt et al., 2013). There are several possible reasons to explain this discrepancy. First, since we employed a competition assay on the human TF protein microarrays, only those that preferentially bind to methylated DNA motifs will be identified. In other words, if a protein binds to a DNA sequence irrespective of its methylation status, it will not show strong signals in our screen. Second, the number of DNA sequences we tested is rather small. Based on our simulation analysis (Figure 1—figure supplement 8), we expect that there exist more such human TFs yet to be discovered with more DNA motifs tested. Third, because we do not have any TF heterodimers on our arrays, it may also generate false negatives. For example, a recent study (Spruijt et al., 2013) employed a generic methylated DNA probe to pull down potential methylated DNA-binding proteins in mouse cell lysates, resulted in the identification of 19 proteins. Among them, 11 have human orthologs and are presented on our human protein microarrays. Five of them, namely MeCP2, NRF1, MBD1/4, and KLF4, showed methylated-specific binding activities in our study. The other six proteins showed weak binding signals in our competition assay, presumably due to the reasons discussed above (Figure 1—figure supplement 1E).

Regardless of all the above limitations, we still discovered many novel methylation-dependent DNA-binding activities. For example, Znf114 and Znf416 were identified as new generic methylated-DNA binding proteins, with binding activities to almost all methylated DNA motifs tested in this study. The reason that they were not discovered in previous studies might be due to the fact that most of the previous studies were not performed in a systematic way. In our systematic survey, we identified a total of 47 novel methylated DNA-binding proteins, which significantly expanded the methylation-dependent protein–DNA interaction landscape.

Our findings provide a new framework to better understand mechanisms of DNA methylation-dependent regulation of gene expression, especially in cancer, stem cell and development biology. For example, recent DNA methylome studies in cancer have identified dynamic epigenetic changes, resulting in global reprogramming in gene expression (Jones and Baylin, 2007; Figueroa et al., 2010; Noushmehr et al., 2010). However, CpG methylation does not always correlate with transcription repression in cancers (Everhard et al., 2009). Our unbiased screen of the entire human TF family identified many sequence-specific mCpG ‘readers’, which can presumably interpret the DNA methylation changes and in turn, regulate gene expression in a dynamic environment. Therefore, by presenting distinct binding sites for TFs, methylated cytosines may serve as the fifth alphabet that changes the landscape of TF–DNA interactions and increases the complexity and diversity of gene regulation.

1. 1. Bartels SJ
   2. Spruijt CG
   3. Brinkman AB
   4. Jansen PW
   5. Vermeulen M
   6. Stunnenberg HG

   (2011) A SILAC-based screen for Methyl-CpG binding proteins identifies RBP-J as a DNA methylation and sequence-specific binding protein

   PLOS ONE 6:e25884.

   https://doi.org/10.1371/journal.pone.0025884

   • Google Scholar
2. 1. Bartke T
   2. Vermeulen M
   3. Xhemalce B
   4. Robson SC
   5. Mann M
   6. Kouzarides T

   (2010) Nucleosome-interacting proteins regulated by DNA and histone methylation

   Cell 143:470–484.

   https://doi.org/10.1016/j.cell.2010.10.012

   • Google Scholar
3. 1. Bateman A
   2. Coin L
   3. Durbin R
   4. Finn RD
   5. Hollich V
   6. Griffiths-Jones S

   et al. (2004) The Pfam protein families database

   Nucleic Acids Res 32:D138–D141.

   https://doi.org/10.1093/nar/gkh121

   • Google Scholar
4. 1. Baylin SB

   (2005) DNA methylation and gene silencing in cancer

   Nat Clin Pract Oncol 2:S4–S11.

   https://doi.org/10.1038/ncponc0354

   • Google Scholar
5. 1. Boyes J
   2. Bird A

   (1991) DNA methylation inhibits transcription indirectly via a methyl-CpG binding protein

   Cell 64:1123–1134.

   https://doi.org/10.1016/0092-8674(91)90267-3

   • Google Scholar
6. 1. Chen X
   2. Xu H
   3. Yuan P
   4. Fang F
   5. Huss M
   6. Vega VB

   et al. (2008) Integration of external signaling pathways with the core transcriptional network in embryonic stem cells

   Cell 133:1106–1117.

   https://doi.org/10.1016/j.cell.2008.04.043

   • Google Scholar
7. 1. Chiang CH
   2. Su Y
   3. Wen Z
   4. Yoritomo N
   5. Ross CA
   6. Margolis RL

   et al. (2011) Integration-free induced pluripotent stem cells derived from schizophrenia patients with a DISC1 mutation

   Mol Psychiatry 16:358–360.

   https://doi.org/10.1038/mp.2011.13

   • Google Scholar
8. 1. Egger G
   2. Liang G
   3. Aparicio A
   4. Jones PA

   (2004) Epigenetics in human disease and prospects for epigenetic therapy

   Nature 429:457–463.

   https://doi.org/10.1038/nature02625

   • Google Scholar
9. 1. Elemento O
   2. Slonim N
   3. Tavazoie S

   (2007) A universal framework for regulatory element discovery across all genomes and data types

   Mol Cell 28:337–350.

   https://doi.org/10.1016/j.molcel.2007.09.027

   • Google Scholar
10. 1. Everhard S
    2. Tost J
    3. El Abdalaoui H
    4. Criniere E
    5. Busato F
    6. Marie Y

    et al. (2009) Identification of regions correlating MGMT promoter methylation and gene expression in glioblastomas

    Neuro Oncol 11:348–356.

    https://doi.org/10.1215/15228517-2009-001

    • Google Scholar
11. 1. Fei YY
    2. Schmidt A
    3. Bylund G
    4. Johansson DX
    5. Henriksson S
    6. Lebrilla C

    et al. (2011) Use of real-time, label-free analysis in revealing low-affinity binding to blood group antigens by Helicobacter pylori

    Anal Chem 83:6336–6341.

    https://doi.org/10.1021/ac201260c

    • Google Scholar
12. 1. Feinberg AP

    (2007) Phenotypic plasticity and the epigenetics of human disease

    Nature 447:433–440.

    https://doi.org/10.1038/nature05919

    • Google Scholar
13. 1. Figueroa ME
    2. Lugthart S
    3. Li Y
    4. Erpelinck-Verschueren C
    5. Deng X
    6. Christos PJ

    et al. (2010) DNA methylation signatures identify biologically distinct subtypes in acute myeloid leukemia

    Cancer Cell 17:13–27.

    https://doi.org/10.1016/j.ccr.2009.11.020

    • Google Scholar
14. 1. Filion GJ
    2. Zhenilo S
    3. Salozhin S
    4. Yamada D
    5. Prokhortchouk E
    6. Defossez PA

    (2006) A family of human zinc finger proteins that bind methylated DNA and repress transcription

    Mol Cell Biol 26:169–181.

    https://doi.org/10.1128/MCB.26.1.169-181.2006

    • Google Scholar
15. 1. Guo JU
    2. Su Y
    3. Zhong C
    4. Ming GL
    5. Song H

    (2011) Hydroxylation of 5-methylcytosine by TET1 promotes active DNA demethylation in the adult brain

    Cell 145:423–434.

    https://doi.org/10.1016/j.cell.2011.03.022

    • Google Scholar
16. 1. Ho KL
    2. McNae IW
    3. Schmiedeberg L
    4. Klose RJ
    5. Bird AP
    6. Walkinshaw MD

    (2008) MeCP2 binding to DNA depends upon hydration at methyl-CpG

    Mol Cell 29:525–531.

    https://doi.org/10.1016/j.molcel.2007.12.028

    • Google Scholar
17. 1. Hu S
    2. Xie Z
    3. Blackshaw S
    4. Qian J
    5. Zhu H

    (2011) Characterization of protein-DNA interactions using protein microarrays

    Cold Spring Harb Protoc 5:499–507.

    https://doi.org/10.1101/pdb.prot5614

    • Google Scholar
18. 1. Hu S
    2. Xie Z
    3. Onishi A
    4. Yu X
    5. Jiang L
    6. Lin J

    et al. (2009) Profiling the human protein-DNA interactome reveals ERK2 as a transcriptional repressor of interferon signaling

    Cell 139:610–622.

    https://doi.org/10.1016/j.cell.2009.08.037

    • Google Scholar
19. 1. Jaenisch R
    2. Bird A

    (2003) Epigenetic regulation of gene expression: how the genome integrates intrinsic and environmental signals

    Nat Genet 33:245–254.

    https://doi.org/10.1038/ng1089

    • Google Scholar
20. 1. Jones PA
    2. Baylin SB

    (2007) The epigenomics of cancer

    Cell 128:683–692.

    https://doi.org/10.1016/j.cell.2007.01.029

    • Google Scholar
21. 1. Karlsson QH
    2. Schelcher C
    3. Verrall E
    4. Petosa C
    5. Sinclair AJ

    (2008) Methylated DNA recognition during the reversal of epigenetic silencing is regulated by cysteine and serine residues in the Epstein-Barr virus lytic switch protein

    PLOS Pathog 4:e1000005.

    https://doi.org/10.1371/journal.ppat.1000005

    • Google Scholar
22. 1. Klug M
    2. Rehli M

    (2006) Functional analysis of promoter CpG methylation using a CpG-free luciferase reporter vector

    Epigenetics 1:127–130.

    https://doi.org/10.4161/epi.1.3.3327

    • Google Scholar
23. 1. Landry JP
    2. Zhu XD
    3. Gregg JP

    (2004) Label-free detection of microarrays of biomolecules by oblique-incidence reflectivity difference microscopy

    Opt Lett 29:581–583.

    https://doi.org/10.1364/OL.29.000581

    • Google Scholar
24. 1. Lewis JD
    2. Meehan RR
    3. Henzel WJ
    4. Maurer-Fogy I
    5. Jeppesen P
    6. Klein F

    et al. (1992) Purification, sequence, and cellular localization of a novel chromosomal protein that binds to methylated DNA

    Cell 69:905–914.

    https://doi.org/10.1016/0092-8674(92)90610-O

    • Google Scholar
25. 1. Lister R
    2. Pelizzola M
    3. Dowen RH
    4. Hawkins RD
    5. Hon G
    6. Tonti-Filippini J

    et al. (2009) Human DNA methylomes at base resolution show widespread epigenomic differences

    Nature 462:315–322.

    https://doi.org/10.1038/nature08514

    • Google Scholar
26. 1. Liu XS
    2. Brutlag DL
    3. Liu JS

    (2002) An algorithm for finding protein-DNA binding sites with applications to chromatin-immunoprecipitation microarray experiments

    Nat Biotechnol 20:835–839.

    https://doi.org/10.1038/nbt717

    • Google Scholar
27. 1. Liu Y
    2. Toh H
    3. Sasaki H
    4. Zhang X
    5. Cheng X

    (2012) An atomic model of Zfp57 recognition of CpG methylation within a specific DNA sequence

    Genes Dev 26:2374–2379.

    https://doi.org/10.1101/gad.202200.112

    • Google Scholar
28. 1. Ma DK
    2. Jang MH
    3. Guo JU
    4. Kitabatake Y
    5. Chang ML
    6. Pow-Anpongkul N

    et al. (2009) Neuronal activity-induced Gadd45b promotes epigenetic DNA demethylation and adult neurogenesis

    Science 323:1074–1077.

    https://doi.org/10.1126/science.1166859

    • Google Scholar
29. 1. Majka J
    2. Speck C

    (2007) Analysis of protein-DNA interactions using surface plasmon resonance

    Adv Biochem Eng Biotechnol 104:13–36.

    https://doi.org/10.1007/10_026

    • Google Scholar
30. 1. Mann IK
    2. Chatterjee R
    3. Zhao J
    4. He X
    5. Weirauch MT
    6. Hughes TR

    et al. (2013) CG methylated microarrays identify a novel methylated sequence bound by the CEBPB|ATF4 heterodimer that is active in vivo

    Genome Res 23:988–997.

    https://doi.org/10.1101/gr.146654.112

    • Google Scholar
31. 1. Meehan RR
    2. Lewis JD
    3. McKay S
    4. Kleiner EL
    5. Bird AP

    (1989) Identification of a mammalian protein that binds specifically to DNA containing methylated CpGs

    Cell 58:499–507.

    https://doi.org/10.1016/0092-8674(89)90430-3

    • Google Scholar
32. 1. Nan X
    2. Cross S
    3. Bird A

    (1998)

    Gene silencing by methyl-CpG-binding proteins

    6–16, Novartis Found Symp, 214, discussion 16-21, 46-50, 10.1002/9780470515501.ch2.

    • Google Scholar
33. 1. Nelson JD
    2. Denisenko O
    3. Bomsztyk K

    (2006) Protocol for the fast chromatin immunoprecipitation (ChIP) method

    Nat Protoc 1:179–185.

    https://doi.org/10.1038/nprot.2006.27

    • Google Scholar
34. 1. Noushmehr H
    2. Weisenberger DJ
    3. Diefes K
    4. Phillips HS
    5. Pujara K
    6. Berman BP

    et al. (2010) Identification of a CpG island methylator phenotype that defines a distinct subgroup of glioma

    Cancer Cell 17:510–522.

    https://doi.org/10.1016/j.ccr.2010.03.017

    • Google Scholar
35. 1. Quenneville S
    2. Verde G
    3. Corsinotti A
    4. Kapopoulou A
    5. Jakobsson J
    6. Offner S

    et al. (2011) In embryonic stem cells, ZFP57/KAP1 recognize a methylated hexanucleotide to affect chromatin and DNA methylation of imprinting control regions

    Mol Cell 44:361–372.

    https://doi.org/10.1016/j.molcel.2011.08.032

    • Google Scholar
36. 1. Ramsahoye BH
    2. Biniszkiewicz D
    3. Lyko F
    4. Clark V
    5. Bird AP
    6. Jaenisch R

    (2000) Non-CpG methylation is prevalent in embryonic stem cells and may be mediated by DNA methyltransferase 3a

    Proc Natl Acad Sci USA 97:5237–5242.

    https://doi.org/10.1073/pnas.97.10.5237

    • Google Scholar
37. 1. Reik W

    (2007) Stability and flexibility of epigenetic gene regulation in mammalian development

    Nature 447:425–432.

    https://doi.org/10.1038/nature05918

    • Google Scholar
38. 1. Rishi V
    2. Bhattacharya P
    3. Chatterjee R
    4. Rozenberg J
    5. Zhao J
    6. Glass K

    et al. (2010) CpG methylation of half-CRE sequences creates C/EBPα binding sites that activate some tissue-specific genes

    Proc Natl Acad Sci USA 107:20311–20316.

    https://doi.org/10.1073/pnas.1008688107

    • Google Scholar
39. 1. Robertson KD

    (2005) DNA methylation and human disease

    Nat Rev Genet 6:597–610.

    https://doi.org/10.1038/nrg1655

    • Google Scholar
40. 1. Schuetz A
    2. Nana D
    3. Rose C
    4. Zocher G
    5. Milanovic M
    6. Koenigsmann J

    et al. (2011) The structure of the Klf4 DNA-binding domain links to self-renewal and macrophage differentiation

    Cell Mol Life Sci 68:3121–3131.

    https://doi.org/10.1007/s00018-010-0618-x

    • Google Scholar
41. 1. Shields JM
    2. Yang VW

    (1998) Identification of the DNA sequence that interacts with the gut-enriched Krüppel-like factor

    Nucleic Acids Res 26:796–802.

    https://doi.org/10.1093/nar/26.3.796

    • Google Scholar
42. 1. Spruijt CG
    2. Gnerlich F
    3. Smits AH
    4. Pfaffeneder T
    5. Jansen PW
    6. Bauer C

    et al. (2013) Dynamic readers for 5-(hydroxy)methylcytosine and its oxidized derivatives

    Cell 152:1146–1159.

    https://doi.org/10.1016/j.cell.2013.02.004

    • Google Scholar
43. 1. Xie X
    2. Lu J
    3. Kulbokas EJ
    4. Golub TR
    5. Mootha V
    6. Lindblad-Toh K

    et al. (2005) Systematic discovery of regulatory motifs in human promoters and 3’ UTRs by comparison of several mammals

    Nature 434:338–345.

    https://doi.org/10.1038/nature03441

    • Google Scholar
44. 1. Xie Z
    2. Hu S
    3. Blackshaw S
    4. Zhu H
    5. Qian J

    (2010) hPDI: a database of experimental human protein-DNA interactions

    Bioinformatics 26:287–289.

    https://doi.org/10.1093/bioinformatics/btp631

    • Google Scholar
45. 1. Yu X
    2. Lin J
    3. Zack DJ
    4. Qian J

    (2006) Computational analysis of tissue-specific combinatorial gene regulation: predicting interaction between transcription factors in human tissues

    Nucleic Acids Res 34:4925–4936.

    https://doi.org/10.1093/nar/gkl595

    • Google Scholar
46. 1. Zhu X
    2. Landry JP
    3. Sun YS
    4. Gregg JP
    5. Lam KS
    6. Guo X

    (2007) Oblique-incidence reflectivity difference microscope for label-free high-throughput detection of biochemical reactions in a microarray format

    Appl Opt 46:1890–1895.

    https://doi.org/10.1364/AO.46.001890

    • Google Scholar
47. 1. Ziller MJ
    2. Müller F
    3. Liao J
    4. Zhang Y
    5. Gu H
    6. Bock C

    et al. (2011) Genomic distribution and inter-sample variation of non-CpG methylation across human cell types

    PLOS Genet 7:e1002389.

    https://doi.org/10.1371/journal.pgen.1002389

    • Google Scholar

Author details

1. Shaohui Hu

   1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Center for High-Throughput Biology, Johns Hopkins University School of Medicine, Baltimore, United States

   Present address

   CDI Laboratories, Baltimore, United States

   Contribution

   SH, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Jun Wan and Yijing Su

   Competing interests

   The authors declare that no competing interests exist.

2. Jun Wan

   The Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   JW, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Shaohui Hu and Yijing Su

   Competing interests

   The authors declare that no competing interests exist.

3. Yijing Su

   1. Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   YS, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Contributed equally with

   Shaohui Hu and Jun Wan

   Competing interests

   The authors declare that no competing interests exist.

4. Qifeng Song

   1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Center for High-Throughput Biology, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   QS, Acquisition of data, Analysis and interpretation of data

   Competing interests

   The authors declare that no competing interests exist.

5. Yaxue Zeng

   1. Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   YZ, Acquisition of data, Analysis and interpretation of data

   Competing interests

   The authors declare that no competing interests exist.

6. Ha Nam Nguyen

   1. Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Cellular and Molecular Medicine Graduate Program, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   HNN, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

7. Jaehoon Shin

   1. Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Cellular and Molecular Medicine Graduate Program, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   JS, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

8. Eric Cox

   1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Center for High-Throughput Biology, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   EC, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

9. Hee Sool Rho

   1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, United States
   2. Center for High-Throughput Biology, Johns Hopkins University School of Medicine, Baltimore, United States

   Contribution

   HSR, Acquisition of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

10. Crystal Woodard

    1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, United States
    2. Center for High-Throughput Biology, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    CW, Acquisition of data, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

11. Shuli Xia

    1. Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, United States
    2. Hugo W Moser Research Institute at Kennedy Krieger, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    SX, Conception and design, Acquisition of data

    Competing interests

    The authors declare that no competing interests exist.

12. Shuang Liu

    Institute of Physics, Chinese Academy of Sciences, Beijing, China

    Contribution

    SL, Conception and design, Acquisition of data, Analysis and interpretation of data

    Competing interests

    The authors declare that no competing interests exist.

13. Huibin Lyu

    Institute of Physics, Chinese Academy of Sciences, Beijing, China

    Contribution

    HL, Conception and design, Acquisition of data, Analysis and interpretation of data

    Competing interests

    The authors declare that no competing interests exist.

14. Guo-Li Ming

    1. Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, United States
    2. Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, United States
    3. Cellular and Molecular Medicine Graduate Program, Johns Hopkins University School of Medicine, Baltimore, United States
    4. The Solomon Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    G-LM, Analysis and interpretation of data, Drafting or revising the article

    Competing interests

    The authors declare that no competing interests exist.

15. Herschel Wade

    Department of Biophysics and Biophysical Chemistry, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    HW, Conception and design, Analysis and interpretation of data, Drafting or revising the article

    For correspondence

    herschel.wade@jhmi.edu

    Competing interests

    The authors declare that no competing interests exist.

16. Hongjun Song

    1. Institute for Cell Engineering, Johns Hopkins University School of Medicine, Baltimore, United States
    2. Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, United States
    3. Cellular and Molecular Medicine Graduate Program, Johns Hopkins University School of Medicine, Baltimore, United States
    4. The Solomon Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    HS, Conception and design, Analysis and interpretation of data, Drafting or revising the article

    For correspondence

    shongju1@jhmi.edu

    Competing interests

    The authors declare that no competing interests exist.

17. Jiang Qian

    The Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    JQ, Conception and design, Analysis and interpretation of data, Drafting or revising the article

    For correspondence

    jiang.qian@jhmi.edu

    Competing interests

    The authors declare that no competing interests exist.

18. Heng Zhu

    1. Department of Pharmacology and Molecular Sciences, Johns Hopkins University School of Medicine, Baltimore, United States
    2. Center for High-Throughput Biology, Johns Hopkins University School of Medicine, Baltimore, United States

    Contribution

    HZ, Conception and design, Analysis and interpretation of data, Drafting or revising the article

    For correspondence

    hzhu4@jhmi.edu

    Competing interests

    The authors declare that no competing interests exist.

• 5,799

  views

• 1,052

  downloads

• 281

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.