Structural impact of complete CpG methylation within target DNA on specific complex formation of the inducible transcription factor Egr-1

Abstract

The inducible transcription factor Egr-1 binds specifically to 9-bp target sequences containing two CpG sites that can potentially be methylated at four cytosine bases. Although it appears that complete CpG methylation would make an unfavorable steric clash in the previous crystal structures of the complexes with unmethylated or partially methylated DNA, our affinity data suggest that Egr-1 is insensitive to CpG methylation. We have determined, at a 1.4-Å resolution, the crystal structure of the Egr-1 zinc-finger complex with completely methylated target DNA. Structural comparison of the three different methylation states reveals why Egr-1 can recognize the target sequences regardless of CpG methylation.

Introduction

DNA methylation at the C5 position of cytosine bases in CpG dinucleotides by DNA methyltransferases is common in higher eukaryotes, particularly vertebrates [1]. A majority of gene promoters involve regions called CpG islands, where CpGs are present at a high density but rarely methylated [2–5]. Interestingly, although the overall methylation level of CpGs in human genomic DNA is 70–80% [6], methylation levels in CpG islands are typically less than 10% for active genes. CpG methylation in promoters generally causes gene repression, and is considered to play important roles in development [2, 7]. Not surprisingly, abnormality in CpG methylation is associated with major diseases such as cancer [8], atherosclerosis [9], and schizophrenia [10].

Cross-talk between site-specific transcription factors and DNA methylation states is considered to play an important role in gene regulation via CpG methylation [5, 11–13]. Some transcription factors, including those of Ets, zinc-finger, Myc, E2F, and ATF families, recognize DNA sequences containing CpG [12]. Methylation of their target DNA sites can potentially affect these transcription factors. Conversely, the transcription factors can protect bound DNA regions from methylation [5, 12]. In fact, the transcription factors Sp1 and CTCF are known to block methylation and maintain the low methylation level of the CpG island promoters [14–16]. This protection is possible because these transcription factors are constitutively expressed. For inducible transcription factors, however, their target sequences could become methylated more easily because these proteins are only transiently expressed upon particular stimuli to the cells.

Our current study is focused on the impact of methylation within target DNA on specific complex formation of the inducible transcription factor Egr-1 (also known as Zif268). This protein recognizes 9-bp sequences GCGTGGGCG via three zinc-finger (ZF) domains [17, 18]. In the brain, Egr-1 is induced via synaptic signals and plays an important role in long-term memory formation and consolidation [19, 20]. In the cardiovascular system, Egr-1 serves as a stress-inducible transcription factor that initiates defense against vascular stress [21, 22]. The Egr-1 target sequences contain two CpGs. It is known that methylation of these CpGs in the Egr-1 target sites causes repression of some genes in vivo [23–26]. To better understand this effect, the impact of CpG methylation on DNA recognition by Egr-1 should be investigated at both molecular and atomic levels.

For this purpose, we compare the dissociation constants and the three-dimensional structures of the Egr-1 zinc-finger DNA complexes containing distinct numbers of 5-methylcytosine (5mC) bases at the CpG sites. We show that Egr-1’s binding to the target DNA is insensitive to CpG methylation although the two CpGs are located at the protein-DNA interfaces. We also present the crystal structure at a 1.4 Å resolution of the Egr-1 zinc-finger-DNA complex in which all the four cytosine bases of the two CpGs in the target are methylated. Comparison of this structure with the other high-resolution structures of the complexes containing no or fewer 5mC bases explains why Egr-1 can recognize the target sequences regardless of CpG methylation states.

Materials and Methods

Preparation of protein and DNA

The sequences of the protein and DNA are shown in Figure 1A. The residue numbering schemes adopted in this paper are according to those of Elrod-Erickson et al. [17]. The Egr-1 zinc-finger protein was prepared as previously described [27–30]. Protein concentration was measured with BCA protein assay kit (Pierce). Individual DNA strands were purchased from Integrated DNA Technologies and purified with a Mono-Q anion-exchange column (GE Healthcare). After annealing of the complementary strands, the DNA duplexes were purified using Mono-Q anion-exchange chromatography to remove any excess amount of single-stranded DNA, as described [31].

Figure 1.

Egr-1 is insensitive to CpG methylation within the target DNA, although two CpGs are present at the interfaces. (A) Sequences of the protein and DNA duplexes used in this study. Numberings of protein and DNA residues are according to Pabo and co-workers [17, 18]. Positions of 5mC bases are indicated in red. (B) Binding isotherm as measured by the fluorescence anisotropy-based titration experiment. Fluorescence anisotropy for TAMRA-labeled 12-bp duplexes (10 nM) was measured as a function of the protein concentration. (C) Competition binding assays using the protein (50 nM), TAMRA-labeled 12-bp target DNA duplex (10 nM), and unlabeled competitor. Fluorescence anisotropy for the TAMRA-labeled DNA was measured as a function of the competitor concentration. The vertical axis is for factions of TAMRA-labeled DNA in the protein-bound state, which were obtained from the anisotropy data. Magenta, blue, and black circles represent data points for completely methylated, partially methylated, and unmethylated DNA duplexes, respectively, in the panels B and C. (D) The 1.4-Å-resolution crystal structure of the Egr-1 zinc-finger complex with completely methylated DNA containing four 5mC bases. The structure of this complex (green) is superposed onto the crystal structures of the complexes with unmethylated (1AAY) or partially methylated (4R2A) DNA (grey). The methyl groups of 5mC and zinc ions of the completely methylated DNA complex are shown in red and yellow, respectively.

Binding affinity measurements

Affinities of the Egr-1 zinc-finger protein for the unmethylated, partially methylated, and completely methylated target DNA duplexes were determined using fluorescence anisotropy as a function of protein concentration (0.1 – 2000 nM). Fluorescence arising from tetramethylrhodamine (TAMRA) attached to the 3′–terminus of DNA (10 nM) was measured using an ISS PC-1 spectrofluorometer, as described [27]. The assays were performed at 20 °C using a buffer of 10 mM Tris-HCl pH 7.5 and 150 mM KCl. The dissociation constant K_d was calculated from the anisotropy data via nonlinear least-squares fitting with:

$$A_{\mathit{obs}}=A_{\mathit{free}}+(A_{\mathit{bound}}-A_{\mathit{free}})\left(P_{\mathit{tot}}+D_{\mathit{tot}}+K_d-\sqrt{{(P_{\mathit{tot}}+D_{\mathit{tot}}+K_d)}^2-4P_{\mathit{tot}}{D}_{\mathit{tot}}}\right)/(2D_{\mathit{tot}}),$$ [Eq. 1]

where A_obs is the observed anisotropy; A_bound and A_free are those of protein-bound DNA and free DNA; and P_tot and D_tot are total concentrations of the protein and the probe DNA, respectively. For each DNA, the affinity measurements were repeated three times.

We also measured K_d values using a fluorescence-based competition assay. First, we prepared a solution of 10 nM 3′-TAMRA-labeled target DNA and 50 nM Egr-1, in which the vast majority of the probe DNA is in the protein-bound state. Then, we titrated this with unlabeled competitor DNA and monitored a change in fluorescence anisotropy as a function of the competitor concentration (0.1 – 256 μM). The competitor DNA used in this assay was a 12-bp DNA duplex with the sequence of 5′-TATGTAGGCGGT-3′, which is similar to the Egr-1 target. We determined the dissociation constants from probe and competitor DNA using the following equation:

$$A_{\mathit{obs}}=\frac{(1-C_{\mathit{tot}}/K_{d(\mathit{comp})})A_{\mathit{free}}+(P_{\mathit{tot}}/K_{d(\mathit{probe})})A_{\mathit{bound}}}{1+C_{\mathit{tot}}/K_{d(\mathit{comp})}+P_{\mathit{tot}}/K_{d(\mathit{probe})}},$$ [Eq. 2]

where K_d(comp) and K_d(probe) are the dissociation constants for the competitor and probe DNA duplexes, respectively, and C_tot is the concentration of the competitor DNA. Eq. 2 is valid only under the conditions of D_tot ≪ P_tot ≪ C_tot. The K_d(comp) constant was separately measured by a protein titration assay using a 3′-TAMRA-labeled version of the competitor DNA.

Crystallization of the completely methylated DNA complex of Egr-1

A solution of the Egr-1 zinc-finger protein was mixed with a solution of the DNA duplex containing four 5mC bases at the two CpG sites in the target sequence at a 1:1.1 molar ratio of protein to DNA. Using Amicon Ultra (Millipore), the complex was concentrated to 1.0 mM in a buffer of 125 mM bis-trispropane•HCl (pH 8.0) and 500 mM NaCl. The high-concentration solution was mixed with a reservoir buffer of 25 mM bis-trispropane•HCl (pH 8.0), 350–550 mM NaCl, and 6–10 % PEG 400, as described by Pavletich and Pabo [18]. Crystals were grown at 17 °C over two weeks using the sitting-drop vapor diffusion method.

Crystallographic data collection and structure determination

The data were collected using a Mar Mosaic 225 detector at the Advance Photon Source (APS) beamline 21-ID-F (wavelength, 0.97872 Å) at Argonne National Laboratory. In order to produce complete high-resolution data with high multiplicity, datasets were collected from 4 crystals, but one was discarded due to non-isomorphism. Each dataset consisted of 360 frames of ½° width, with varying resolution. Images were processed and scaled with HKL2000. An analysis of the merging CC1/2 indicated that the resolution was limited by the detector distance and geometry, as was the completeness in the high resolution bins. Refinement, starting from the 4R2A structure, was performed using Phenix [32], with TLSMD [33] determined TLS parameters, weight optimization, and DNA restraints. Model building and validation was performed in Coot [34]. The atomic coordinates of the crystal structure have been deposited in the PDB with accession code 4X9J.

Results

Before our current study, two crystal structures of the Egr-1 zinc-finger–DNA complexes different in terms of DNA methylation were available. One was a 1.6-Å-resolution structure of the complex with unmethylated DNA (PDB 1AAY) [17]. The other was the 1.6-Å-resolution structure of the complex in which only one of the two CpGs in the target DNA is methylated (PDB 4R2A) [35]. If cytosine-3 (Cyt3) is methylated in these structures, the added CH₃ is too close to the E77 Cγ atom and creates a steric clash with a C…C distance (~2.4 Å) shorter than the sum of van der Waals radii (3.4 Å). Given this, one may expect that methylation of the second CpG would diminish Egr-1 – DNA association. However, as demonstrated below, our data show that Egr-1 is capable of recognizing the target sequence regardless of the CpG methylation state.

Impact of CpG methylation on binding affinities of the Egr-1 zinc-finger protein

Using two different fluorescence-based assays, we examined the influence of CpG methylation on Egr-1’s binding to the target DNA at physiological ionic strength (i.e., 150 mM KCl). By protein titration assays, in which TAMRA fluorescence anisotropy is monitored as a function of the protein concentration, we directly measured the dissociation constants K_d for the complexes with 12 bp DNA duplexes in different methylation states (Figure 1B). The K_d values for the complexes with the unmethylated, partially methylated, and completely methylated DNA duplexes were determined to be 6 ± 1, 7 ± 1, and 10 ± 2 nM, respectively, suggesting that Egr-1 is insensitive to CpG methylation.

Because these K_d values were close to the lower limit of the measurable range in the protein titration assay, we also analyzed affinities by competition assays that allow for K_d determination for high affinity systems. In these assays, a solution of the complex of the protein and fluorescent DNA was initially made, and an unlabeled 12-bp DNA duplex with a weaker affinity was added as the competitor. The fluorescence anisotropy changes as the unlabeled competitor increasingly outcompetes the probe DNA (Figure 1C). From these data together with the affinity of the competitor DNA (K_d = 29 ± 3 nM), we determined the K_d values for the unmethylated, partially methylated, and completely methylated complexes to be 5.1 ± 0.3, 6.1 ± 0.3, and 4.9 ± 0.3 nM, respectively. Thus, both datasets indicate that Egr-1 can recognize the target DNA regardless of its methylation state, although the CpG sites are at the protein-DNA interfaces.

Overall structure of the completely methylated DNA complex of Egr-1

To investigate structural impact of complete CpG methylation within the target DNA, we determined at a 1.4-Å resolution the crystal structure of the Egr-1 zinc-finger protein bound to completely methylated DNA containing four 5mC bases at the two CpG sites. The structure of the complex, involving 208 water molecules, was refined to R_work = 17% and R_free = 20%. Other information on crystallographic data and refinement are given in Table 1. In Figure 1D, this structure is superimposed onto the crystal structures of the complexes with unmethylated DNA (PDB 1AAY) or with partially methylated DNA containing two 5mC base (PDB 4R2A). As is evident from this superimposition, the overall structures in these three states are very similar: the values of root-mean-square deviation (RMSD) of atomic positions for the protein and DNA backbone were 0.71 Å for the completely methylated vs. unmethylated DNA complexes and 0.50 Å for the completely methylated vs. partially methylated DNA complexes. These results indicate that DNA methylation does not significantly alter the overall structure of the complex. However, owing to the high-resolution data, we were able to identify significant structural differences and common features in great details, as described below.

Table 1.

Crystallographic data collection and refinement statistics

Crystallographic data collection
X-ray Source	APS 21-ID-F
Wavelength (Å)	0.97872
Space Group	C2221
Unit cell parameters (A, °)	α = 44.01, b = 55.99, c = 128.95, α = β = γ = 90
Sample temperature (K)	100
Resolution range (Å)	30.49 – 1.41
Total reflections	46629
Non-anomalous reflections	26831
Completeness (%)	90.6 (46)a)
Multiplicity	15.4 (5.7)a)
Rmerge	0.127 (0.65)a)
Rpim	0.036 (0.29)a)
Refinement
Rwork (%)	17.2
Rfree (%)	20.0
Bond RMSD from ideal values (Å)	0.013
Angle RMSD from ideal values (°)	1.324
Ramachandran plot:
Favored (%)	100%
Allowed (%)	0
Outliers (%)	0
No. non-H atoms [Average B (Å2)]
Protein	723 [20.7]
DNA	389 [16.8]
Water	208 [25.8]

^a)Numbers in parentheses are values for the last shell (1.42 – 1.40 Å).

Egr-1’s structural plasticity to recognize target sequences regardless of methylation states

As mentioned above, simplistic addition of a CH₃ group to Cyt3 of the crystal structures of the unmethylated (1AAY) or partially methylated (4R2A) complexes cause a steric clash with the E77 side chain of zinc finger 3 (ZF3). However, our crystal structure of the completely methylated complex shows that Egr-1 adapts the conformation of the E77 side chain so that favorable interactions are made with the methylated DNA (Figure 2A). By this adaption, which accompanies changes in the χ₁ and χ₃ torsion angles by 98° and 84°, respectively, the E77 Cγ methylene group makes hydrophobic interactions with the CH₃ group of the 5mC3 base and the Cδ atom of the F72 aromatic ring. A similar conformational change upon methylation of Cyt9 is also seen for E21, the corresponding residue in ZF1 that interacts with the other CpG site (Figure 2B). Just as seen for the Cyt3 – E77 interaction, methylation of Cyt9 causes a change in the E21 χ₁ and χ₃ torsion angles by 105° and 60°, respectively, whereby the E21 side chain makes van der Waals contact with the CH₃ group of the 5mC3 base and the F16 aromatic ring. The conformational change of the E21 side chain is also found between the structures of unmethylated and partially methylated DNA complexes. These results show that the structural plasticity of Egr-1 allows for target recognition insensitive to CpG methylation.

Figure 2.

Structural changes at the Egr-1 zinc-finger–DNA interfaces upon CpG methylation. Carbon bonds and hydration water molecules are shown in green for the completely methylated DNA complex; in white for the partially methylated DNA complex; and in purple for the unmethylated DNA complex. Curved arrows show methylation-induced changes in the E77 and E21 side-chain conformations. Water molecules at the canonical hydration positions (wA and wB) for Cyt and 5mC bases [36] are shown. Dotted arrows show methylation-induced changes in hydration.

Role of hydration water in methylation-insensitive recognition of target by Egr-1

In general, methyl groups of methyl-CpGs are well hydrated [36]. Recent crystallographic studies suggest that hydration around the 5mC methyl groups plays important roles in DNA recognition by methyl-CpG-binding domain (MBD) proteins such as MeCP2 and MBD4 [37, 38]. Taking advantage of the high-resolution structures in the three different methylation states, we examined how the hydration water molecules are involved in methylation-insensitive recognition of target DNA by Egr-1. Around cytosine bases in DNA, there are two major hydration sites, which Mayer-Jung et al. referred to as wA and wB [36]. Upon CpG methylation, the position of wB is shifted toward DNA phosphate while the position of wA is virtually unaffected [36]. Figure 2 shows wA and wB water molecules around CpG cytosine bases in the three Egr-1 zinc-finger–DNA complexes. Cyt4′ (5mC4′) and Cyt10′ (5mC10′) bases, which are not in direct contact with the protein, exhibit canonical hydration at the wA and wB positions. A shift of wB upon methylation (indicated by dotted arrows in Figure 2) is also seen for these bases. For Cyt3 (5mC3) and Cyt9 (5mC9) bases, the interaction with the E77 or E21 side chain perturbs wB. Interestingly, due to this perturbation, both unmethylated and methylated states exhibit the wB water molecules in the same region. Regardless of DNA methylation, the water molecules wA at these bases remain at the same position, and importantly, form hydrogen bonds with the D76 or D20 side chain as well as with the other wA water molecule in the same CpG dinucleotide. Thus, in the target recognition, Egr-1 uses the same water-mediated interactions that are unaffected by CpG methylation. This feature appears to be important for Egr-1’s capability of recognizing the target DNA regardless of CpG methylation states.

5mC-Arg-G triads at the two CpG sites

Liu et al. recently identified a structural motif, “5mC-Arg-G triads”, common to the proteins bound to DNA at methylated CpGs [39]. In this structural motif, the methyl group of a 5mC makes van der Waals contact with the guanidino moiety of the Arg side-chain that forms two hydrogen bonds with 3′-guanine (Gua) base of methylated CpG. This structural motif is commonly found in currently available crystal structures of methyl-CpG DNA-protein complexes [35, 37, 38, 40–42]. Our crystal structure of the completely methylated DNA complex of Egr-1 also shows two 5mC-Arg-G triads: one with 5mC3, R74, and Gua4 (Figure 3A), and the other with 5mC9, R18, and Gua10 (Figure 3B). The center distances between the Cyt pyrimidine rand Arg Nη1/Nη2 groups are 3.7 and 3.8 Å for the 5mC3-R74 and 5mC9-R18 pairs, respectively, suggesting that these 5mC-Arg-G triads are stabilized by the cation–π interactions [43]. The presence of this structural motif at each CpG methylation site suggests that Egr-1 is well suited to recognize the methylated target sequences.

Figure 3.

Two 5mC-Arg-G triads in the 1.4 Å resolution crystal structure of the Egr-1 complex with completely methylated DNA. The electron density map together with ball and stick representation are shown for the triads formed by 5mC3, R74, and Gua4 (the panel A) and by 5mC9, R18, and Gua10 (the panel B). The hydrogen bonds between Arg and G are indicated with a dotted line. The 5mC-Arg- G triad motif is commonly found in three-dimensional structures of the proteins that specifically recognize methyl-CpG DNA [39].

Discussion

Our current study shows that, even when the target sequence is completely methylated at the two CpG sites, Egr-1 can adapt its conformation to recognize the target DNA and retain the same affinity. Given these in vitro data, one may expect that Egr-1’s function is insensitive to DNA methylation in vivo as well. However, some studies suggest otherwise. For example, CpG methylation of Egr-1 sites causes repression of the protein kinase Cε (PKCε) gene in the heart [23, 24, 26]. A decreased level of CpG methylation in the Egr-1 target site causes hyperactive expression of the heparanase gene in bladder cancer [25]. These in-vivo data suggest that the CpG methylation in the Egr-1 target site causes a decrease in expression of some target genes. Why can this occur, although Egr-1 retains its intrinsic binding affinity for the target sites regardless of methylation?

We speculate that this discrepancy between the in-vitro and in-vivo data could be due to competition with proteins that bind specifically to methyl-CpGs. In fact, blocking of transcription factors by MBD proteins has been proposed as an indirect mechanism for gene repression [5]. For example, the MeCP2 protein is abundant in many cell types and causes gene repression via its transcriptional repression domain [44, 45]. Judging from the crystal structure of the MeCP2 MBD–DNA complex [37], simultaneous binding of Egr-1 and MeCP2 to the same sites appears impossible, and therefore, Egr-1 should compete with MeCP2 for the methylated target sequences. When the Egr-1 target sites are methylated, MeCP2 could outcompete Egr-1 for these sites due to the abundance of MeCP2. In this indirect manner, CpG methylation of Egr-1 sites can still cause gene repression even though the CpG methylation itself does not directly affect binding affinities of Egr-1 for the target sequences. Additional studies are required to examine this hypothetical mechanism.

It should be noted that CpG methylation is known to directly impact some other zinc-finger proteins even in vitro. For example, the C2H2 zinc-finger proteins Kaiso and Zfp57 exhibit substantially stronger affinity for CpG-methylated target DNA [46, 47]. In contrast, the CXXC zinc-finger protein Cfp1 is known to selectively bind to unmethylated CpGs [48]. Recent crystallographic studies on the DNA complexes of these zinc-finger proteins provide structural basis for the direct impact of CpG methylation [40, 42, 49]. Our current study shows that, unlike these proteins, Egr-1 is insensitive to CpG methylation of its target DNA sequences, owing to its ability to adapt the side-chain conformations depending on the methylation states of the two CpGs. Other studies also show a relatively minor impact of CpG methylation on other C2H2 zinc-finger proteins, Wilms tumor protein 1 (WT1) [35] and Krüppel-like factor 4 (Klf4) [41]. Thus, there are diverse impacts of CpG methylation on transcription factors, even within the zinc-finger family alone. This diversity might be important for elaborate epigenetic programming in higher eukaryotes such as mammals.

Highlights.

• Egr-1 is intrinsically insensitive to CpG methylation of the target DNA.

• Structures are compared for the Egr-1-DNA complexes in distinct methylation states.

• Structural plasticity allows Egr-1 to bind to target DNA regardless of methylation.

• Egr-1 binds to completely methylated target DNA via two 5mC-Arg-G triads.

• Recognition by Egr-1 involves water molecules unaffected by DNA methylation.

Abbreviations

5mC

5-methylcytosine

MBD

methyl-CpG-binding domain

RMSD

root-mean-square deviation

TAMRA

tetramethylrhodamine