The Eaf3 chromodomain acts as a pH sensor for gene expression by altering its binding affinity for histone methylated-lysine residues

Post-translational modifications of histones play an essential role in the regulation of transcription, DNA repair, replication and heterochromatin formation by changing the chromatin structure. For example, histones are specifically acetylated and methylated to induce gene expression [1,2]. Saccharomyces cerevisiae shows a global pattern of histone H3 and H4 acetylation, whereby acetylation is higher at promoters and lower in coding regions [3,4]. Acetylation loosens the chromatin structure, allowing RNA polymerase II (Pol II) to easily access the promoter; in transcribing regions, by contrast, de-acetylation tightens the chromatin structure, meaning that Pol II can read the gene information but cannot access the gene to start transcription from a cryptic site. Similar to these patterns of acetylation, lysine 4 of histone H3 (H3K4) is methylated by Set1 methyltransferase at promoter sites, whereas Set2 methyltransferase mediates H3K36 methylation in coding regions [5].

In addition to the acetylation pattern for gene expression, histone acetylation is also affected by intracellular pH (pH_i) in both normal and cancer cells [6]. When pH_i decreases, histones are globally de-acetylated by histone deacetylases (HDACs), and the released acetate anions are co-exported with protons from cells by monocarboxylate transporters to maintain pH_i. Further decreases in pH_i have been shown to trigger apoptosis [7]. Alongside the rising pH_i that occurs when resting cells are induced to proliferate, differentiate and progress in the cell cycle [8], global acetylation of histone increases.

Although an apparent relation between pH_i and histone acetylation responsible for gene expression has been reported, the pH_i-dependent molecular mechanism of gene expression due to histone acetylation and methylation remains elusive. In terms of the relation between acetylation and methylation, Eaf3 (essential Sas2-related acetyltransferase1-associated factor 3) is a key protein in S. cerevisiae. Eaf3 is a component of both the NuA4 histone acetyltransferase (HAT) complex and the Rpd3S HDAC complex. It is thought that the N-terminal chromodomain (CD) of Eaf3 recognizes H3K36 methylation, which then positions Rpd3S on the nucleosome to suppress intragenic transcription initiation [9–11]. However, binding between Eaf3 CD and methylated H3K36 (H3K36me) is relatively weak; therefore, it is thought that Eaf3 CD determines the specificity for H3K36me but the strong binding affinity is cooperatively mediated by the Rco1 subunit of Rpd3S via its plant homeodomain (PHD) [12,13]. Because of the weak interaction, the structure of Eaf3 CD bound to the H3K36me peptide was determined by NMR using a fusion protein of the CD and a model H3K36me peptide [14].

Notably, Eaf3 is also a subunit of the NuA4 HAT complex, which recognizes H3K4me at promoter site nucleosomes. In fact, Eaf3 CD can also bind weakly to H3K4me, facilitating recruitment of NuA4 to the promoter region. Similar to Rpd3S, the PHD of the Yng2 subunit of NuA4 plays an essential cooperative role in binding of Eaf3 CD to H3K4me [15,16]. Thus, Eaf3 CD seems to have no strong sequence specificity and can bind to both H3K4me and H3K36me, while the specific recruitment of NuA4 and Rpd3S to H3K4me and H3K36me, respectively, is cooperatively determined by the PHD of the respective Yng2 and Rco1 subunits. Nevertheless, the cooperative effect of Eaf3 is important because cells lacking Eaf3 have an even distribution of histone acetylation levels across the genome, whereas wild-type (WT) cells have higher histone acetylation levels at promoter regions than at coding sequences [17].

Here, we have explored the role of Eaf3 CD binding to H3K4me and H3K36me. We examined the complexes of Eaf3 CD bound to lysine-methylated histones, which showed that Eaf3 CD is a unique module containing a histidine, His18, in its binding pocket for methylated lysine residues of histones, even though a positively charged residue would be expected to repulse the positively charged histone methyl-lysine. Protonation of His18 was found to critically control both the binding activity and the pH sensitivity of Eaf3 CD: at high pH_i, the deprotonation of Eaf3 CD causes strong binding to H3K4me at promoters via NuA4, and to H3K36me at transcribing regions via Rpd3S. Thus, the role of Eaf3 CD is to act as a sensor of pH_i to control gene expression, rather than to act as a determinant of binding between H3K4me and H3K36me, which is achieved by other components in NuA4 and Rpd3S.

WT and mutants of Eaf3 CD (residues 1−120) were expressed as hexa-histidine-tagged products in pET15b vectors (Novagen) in Escherichia coli BL21(DE3)pLysS (Novagen). Lysed supernatant was loaded onto a Ni-nitrilotriacetic acid (NTA)-agarose (Qiagen) column. Eluted His-tagged Eaf3 CD was then treated with thrombin to remove the 6xHis-tag. The sample was again loaded onto the Ni-NTA agarose column. Fractions passing through the column were concentrated and applied onto Superdex75 (GE Healthcare).

Histone H3 and H4 peptides were synthesized and purchased from Qiagen and SIGMA Genosys. The histone peptide was added to 0.1 mM ¹⁵N-labeled Eaf3 CD at a molar ratio of 1:4, or 1:8, or 1:16 (Eaf3:histone) in 10 mM potassium phosphate (pH 6.8), 20 mM NaCl, and 5 mM deuterated DTT dissolved in 90% H₂O/10% D₂O. ¹H,¹⁵N-HSQC spectra were acquired before and after peptide addition at 25°C on a Bruker AVANCE-600 spectrometer equipped with a cryogenic probe. Chemical shift change was calculated as ∆δ = {(∆δ¹H)² + (∆δ¹⁵N/5)²}^1/2.

For the structure determination, 1.0–1.5 mM Eaf3 CD in 10 mM potassium phosphate (pH 6.8), 20 mM NaCl and 5 mM deuterated DTT dissolved in either 90% H₂O/10% D₂O or 99.9% D₂O was used. NMR experiments were performed at 25°C on a Bruker AVANCE-500, AVANCE-600 or AVANCE-800 spectrometer equipped with a cryogenic probe. Backbone and side chain resonances were assigned by using CBCA(CO)NH, CBCANH, HN(CO)CA, HNCA, HN(CA)CO, HNCO, HBHA(CO)NH, HCCH-COSY, HCCH-TOCSY, CCCONH and HCCCONH [18]. Aromatic side chain resonances were assigned by using DQF-COSY, TOCSY, (HB)CB(CGCD)HD, (HB)CB(CGCDCE)HE [19], CG(CB)H, CG(CD)H and CG(CDCE)H [20]. Stereospecific assignments were obtained from a combination of HNHB, HN(CO)HB, HNCG, HN(CO)CG [21] and ¹³C-edited NOESY-HSQC (τ_m = 50 ms) [18]. Distance restraints were obtained from NOESY, ¹⁵N-edited NOESY-HSQC (τ_m = 150 ms) and ¹³C-edited NOESY-HSQC (τ_m = 50 and 100 ms). Side-chain torsion angles, χ1 and χ2, were obtained from a combination of HNHB, HN(CO)HB, HNCG, HN(CO)CG [21] and ¹³C-edited NOESY-HSQC (τ_m = 50 ms) [18]. Hydrogen bond restraints were obtained by backbone amide H/D-exchange experiments. Spectra were processed by NMRPipe [22] and analyzed by NMRView [23].

Interproton distance restraints derived from NOE intensities were grouped into four distance ranges, 1.8–2.7 Å (1.8–2.9 Å for NOEs involving HN protons), 1.8–3.3 Å (1.8–3.5 Å for NOEs involving HN protons), 1.8–5.0 Å and 1.8–6.0 Å, corresponding to strong, medium, weak and very weak NOEs, respectively. The upper limit was corrected for constraints involving methyl groups, aromatic ring protons and non-stereospecifically assigned methylene protons. Dihedral angle restraints for φ and ψ were obtained by analyzing the backbone chemical shifts with TALOS [24]. χ1 and χ2 angles were restrained ± 30 ° for three side-chain rotamers. Structure calculations were performed by distance geometry and simulated annealing by using Xplor-NIH [25,26], and all structures were subjected to water refinement [27]. A total of 200 structures were calculated. Structural statistics for the 20 best structures are summarized in Supplementary Table S2. Structures were analyzed and displayed using PROCHECK-NMR [28], MOLMOL [29] and PyMOL (Schroedinger).

The order parameter, S², was obtained from the model-free analysis assuming axially symmetric rotation. The backbone ¹⁵N relaxation parameters of Eaf3 CD were measured by using a ¹⁵N-labeled sample of 0.30 mM at 25°C on a Bruker AVANCE-600 (¹⁵N frequency, 60.8 MHz) spectrometer equipped with a cryogenic probe. Relaxation durations of 24, 256, 512, 768, 1024, 1280, 1536, 2048 and 2560 ms for the longitudinal relaxation rates (R1), and 16.98, 33.95, 50.93, 67.90, 101.86, 135.81, 203.71 and 339.52 ms for the transverse relaxation rates (R2) were used. Heteronuclear ¹⁵N-{¹H} NOE experiments were recorded in the presence and absence of proton saturation, which was achieved with a 5.0 s duration consisting of 120° ¹H pulses applied every 5.0 ms. R1 and R2 were obtained by fitting peak intensities at a series of relaxation durations to an exponential decay curve using CurveFit [30]. The uncertainties of the peak intensities were estimated by using duplicated data from the shortest relaxation delay. The uncertainties of the relaxation rates were determined by CurveFit using a Monte–Carlo simulation. The steady-state ¹⁵N-{¹H} NOE values were determined from peak intensity ratios obtained from spectra acquired with and without proton saturation. The uncertainties were determined from the standard deviation in background noise levels by using NMRView [23]. Initial values of the overall correlation time and the axially symmetric rotational diffusion tensor were estimated by using r2r1_diffusion [31]. Model-free analysis was performed by using Modelfree [30] and FAST ModelFree [32].

A 4-fold molar excess of the spin label reagent MTSL (1-oxy-2,2,5,5-tetramethyl-D3-pyrroline-3-methyl) methanethiosulfonate (Toronto Research Chemicals) dissolved in acetonitrile was added to H3K36me3 peptide (residues 32-41) with a Gly33Cys substitution (SIGMA Genosys), dissolved in buffer [10 mM potassium phosphate (pH 6.8), 20 mM NaCl], and incubated at 25°C for 16 h in the dark. Unreacted MTSL was removed by using a PD MidiTrap G-10 column (GE Healthcare) equilibrated with distilled water. The spin labeled peptide eluted with water was lyophilized and dissolved in buffer, and the pH was adjusted to 6.8. A paramagnetic sample was obtained by adding MTSL-H3K36me3 peptide to ¹⁵N-labeled Eaf3 CD dissolved in buffer at a molar ratio of 1:1 (final concentration 0.3 mM), and the ¹H,¹⁵N-HSQC spectrum was recorded. A diamagnetic sample was generated by reducing the paramagnetic sample with a 3.2-fold molar excess of ascorbate (Wako) at 25°C for 4 h in the dark. After the reduction, the ¹H,¹⁵N-HSQC spectrum was recorded.

It has been reported that Eaf3 CD binds weakly to the H3K4me peptide as well as the H3K36me peptide [9,10,14,33]. To confirm this, we monitored NMR signal changes of ¹⁵N-labeled Eaf3 CD upon the addition of histone peptides by using ¹H,¹⁵N HSQC spectroscopy. Initially, we examined the methylation-dependent specificity of binding to the N-terminal fragment (residues 1−42) of histone H3 (H3) (Figure 1A and Supplementary Table S1). The addition of a 4-fold excess of an H3 peptide di-methylated (me2) at K4, K36, or both lysine residues gave rise to a small but significant signal change of specific residues such as Tyr81, Trp84 and Trp88, which formed a binding pocket (Supplementary Figure S1a). By contrast, the addition of a 4-fold excess of unmodified H3 peptide did not result in any altered signals.

To characterize the weak binding of Eaf3 CD in more detail, we observed the interaction of Eaf3 CD with various histone peptides (Figure 1B and Supplementary Table S1). Eaf3 CD interacted with many dimethyl-lysine peptides, and even with arginine-asymmetrically dimethylated (ame2) peptides, but not with unmethylated, phosphorylated (pho) or acetylated (ac) peptides. Thus, the binding of Eaf3 CD to histone is highly methylation-specific but promiscuous.

Because the NuA4 and Rpd3S complexes are closely related to H3K4me and H3K36me, respectively, we focused on these methylation marks and evaluated their quantitative binding to Eaf3 CD by NMR titration (Table 1 and Figure 2). We found that Eaf3 CD bound more strongly (i) to the trimethylated form than to the dimethylated form of both H3K4 and H3K36 and (ii) to di- and trimethylated H3K36 than to the corresponding methylated forms of H3K4. In addition, binding to H3K4me3 was strengthened by R2ame2 modification, but weakened by T3pho modification. Although the different binding features observed were interesting, in all cases the dissociation constant (K_d) values were in the millimolar range. Thus, Eaf3 CD can be categorized as a very weak histone methyl-lysine-binding module.

All methylated-lysine histone-binding modules have a characteristic binding cage consisting of two to four aromatic residues (Phe, Tyr and Trp), which contact a methylammonium moiety through cation–π and van der Waals interactions [34].

The structure of Eaf3 CD has been previously determined and the binding of Eaf3 to H3K36me found to be weak [14,33]. We re-examined both the dynamic and static structures of Eaf3 CD by using NMR spectroscopy (Supplementary Table S2 and Figure S1a). The present NMR structure was consistent with the previous solved structures, although it resembled the crystal structure more than the solution one (Supplementary Figure S2). Unlike HP1 and Polycomb CDs, where a methylated H3 tail is inserted in a cleft, the corresponding cleft in Eaf3 CD is occupied by β strand S1, and helix H2 is positioned above the S1 strand (Supplementary Figure S1c). In addition, relaxation analysis indicated that there is restricted backbone motion on the picosecond to nanosecond timescale for all regions except for the N- and C-termini (residues 1−8 and 115−120) and a long loop between β strands S3 and S4 (residues 43−57) (Supplementary Figure S1d). The two turns between β strands S1 and S2 and between S5 and S6, which form the binding pocket (Supplementary Figure S1a, marked by dashed line), show a relatively low degree of flexibility as normal (Supplementary Figure S1d, marked by dashed lines). It is therefore clear that the origin of the weak binding of Eaf3 CD is not related to the dynamic character of this domain.

Eaf3 CD possesses an aromatic cage formed by Tyr81 at the bottom with Tyr23, Trp84 and Trp88 around the walls (Figure 3A). A previous study of Eaf3 CD bound to a H3K36me2 analog, in which Eaf3 CD was engineered to link to a H3K36 fragment to a chemically incorporated methyl-lysine analogue produced by cysteine alkylation, revealed that these four aromatic residues (Tyr23, Tyr81, Trp84 and Trp88) surround the dimethyl-lysine analog [14]. In fact, each of the four aromatic residues was found to be essential for the interaction because the alanine mutants Y23A, Y81A, W84A and W88A, each of which maintain the proper conformation (Supplementary Figure S3), failed to bind to a trimethylated (me3) H3K36, H3K4me3 or H3R2ame2K4me3 peptide (Table 2 and Supplementary Figure S4). By contrast, W84Y and two alanine mutants of Gly19 and Gln82, both of which are near to the aromatic cage but not involved in direct binding, maintained the binding ability of Eaf3 CD.

Taken together, the above results indicate that Eaf3 CD per se has few direct contacts with histone residues and relies heavily on the interaction with a methyl-lysine. We therefore focused on the residues inside the binding cage of Eaf3 CD. To identify unique features, we first compared the structures of histone methyl-lysine peptide complexes determined so far. Supplementary Table S3 summarizes the results of a systematic survey of residues in which the heavy atoms of the side-chains of the methyl-lysine-binding domains are within 6 Å of the Nζ atom of the methyl-lysine of histone. As expected, all histone-binding domains contain aromatic residues [34]. Curiously, however, Eaf3 CD possesses a histidine, His18, and not an acidic residue as the charged amino acid (Figure 3A and Supplementary Table S4). In general, a histidine residue will be protonated at physiological pH, which would produce a repulsive Coulombic barrier to interacting with the positively charged methyl-lysine of histone. Contrary to Eaf3 CD, acidic residues are often observed in the binding aromatic cages of other methyl-lysine-binding modules (Supplementary Table S3 and Figure 3B). Notably, the electrostatically unfavorable histidine residue in the binding cage is not limited to yeast Eaf3, and is seen in other species such as animals and plants (Supplementary Figure S5).

The structural analysis suggested that the unusual His18 residue in the aromatic cage might be a key determinant for the weak interactions of Eaf3 CD. However, a histidine residue has a wide range of acid dissociation constant (pK_a) values depending on its environment. Next, therefore, we determined the pK_a value of His18 (Figure 4A–D) and its tautomeric state (Figure 4E) by using NMR [36]. We confirmed that no disruption of structure was caused by the pH changes throughout the experiments (not shown). The observed pK_a value of His18 was approximately 6.8 (Figure 4D), indicating that the binding activity of the aromatic cage of Eaf3 CD is sensitive to changes in pH under physiological conditions.

To verify the pH-dependent binding of Eaf3 CD suggested by the above observations, we examined the interaction of Eaf3 CD with methylated histone peptide under different pH (Figure 5A). An increase in pH from 6.8 to 7.5 enhanced the binding of Eaf3 CD to H3K36me3 2.2-fold. Upon a further increase to pH 8.5, the binding was enhanced a further 3.6-fold. In contrast, a decrease from pH 6.8 to 6.0 weakened the binding 2.3-fold. Decreasing the pH to 5.2, where almost all Eaf3 CD molecules would contain protonated His18, abolished the binding. Thus, the protonation state of His18 is negatively correlated with the binding activity of Eaf3 CD. At a physiological pH of 6.8, almost half of all Eaf3 CD molecules will be protonated at His18 and thus in an inactive binding form, while the rest will be deprotonated at His18 and therefore in an active binding form.

We further confirmed the necessity of protonation of His18 in the observed pH sensitivity of binding by examining the binding activity of H18A and H18D at pH 6.8, 8.5 and 5.2 (Figure 5B,C). In contrast with the alanine mutants of the four cage aromatic residues, none of which showed binding (Table 2), H18A showed 2.9-fold stronger binding to the H3K36me3 peptide as compared with the WT peptide at pH 6.8 (Figure 5A,B). Relative to pH 6.8, the binding of H18A to H3K36me3 peptide was slightly enhanced by a factor of 1.2-fold at pH 8.5, and decreased by 1.9-fold at pH 5.2 (Figure 5B), although pH 5.2 did not result in a complete loss of activity of Eaf3 CD, unlike WT (Figure 5A). H18D enhanced the binding to H3K36me3 peptide even more (Figure 5C,B), with an 11.7-fold increase relative to WT at pH 6.8 (Figure 5A,C). H18D maintained high-binding activity at pH 8.5 and to a lesser extent at pH 5.2 (Figure 5C). In addition, the electrostatic state in the aromatic binding cage of the mutant proteins altered the preference for H3K36me3 (Supplementary Figure S6). As described above (Table 1), WT Eaf3 CD showed a preference for H3K36me3 over H3K36me2 (Supplementary Figure S6a). H18A exhibited the same preference (Supplementary Figure S6b), whereas H18D showed a preference for the dimethyl form over the trimethyl one (Supplementary Figure S6c). For binding to H3K36me2 peptide, H18D exhibited a 71.1-fold stronger binding as compared with WT (Supplementary Figure S6c and a). Thus, the introduction of negative charge into the binding cage of Eaf3 CD greatly augmented the binding activity, disrupted the pH sensitivity and altered the preference to H3K36me3.

We also tested the pH dependence of W84A as a representative of alanine mutants of the four binding aromatic residues in the binding site. Similar to the observations at pH 6.8 (Table 2), W84A showed no binding at pH 8.5 or at pH 5.2, demonstrating the pH independence of mutant W84A regardless of His18 de-protonation (Figure 5D).

The above findings showed that His18 is the crucial structure determinant of the weak binding of Eaf3 CD. At the same time, they demonstrated a novel function of Eaf3 CD as a pH_i sensor with histone methyl-lysine-binding ability. We therefore investigated this dual function from the viewpoint of the structure by using paramagnetic relaxation enhancement (PRE) experiments. In PRE, effects are observed depending only on the distance between the paramagnetic center and the proton; thus, it is possible to obtain structural information about low-population states in an exchanging system such as that of Eaf3 CD and H3K36me3 peptide [37].

The addition of an H3K36me3 peptide carrying an MTSL spin label to 0.3 mM ¹⁵N-labeled Eaf3 CD at a molar ratio of 1:1 gave rise to only subtle chemical shift changes of the backbone amide signals for a few residues of Eaf3 CD (Figure 6). At pH 5.2, no PRE effects were observed, indicating no binding (Figure 6A,D). At pH 6.8, by contrast, signals from several specific residues around the binding site were broadened upon addition of the paramagnetically labeled H3K36me3 peptide (Figure 6B,E). For the H18D mutation at pH 6.8, many broadened signals were also observed in the binding site and surrounding area (Figure 6C,F). Taking all the findings together, we conclude that Eaf3 CD is a previously unknown pH_i sensor that detects a change in pH_i and simultaneously converts it into methylated histone-binding ability via protonation of His18 in the binding site.

As described above and in previous studies [9,10,14,33], Eaf3 CD itself has high methylation specificity but low sequence specificity; therefore, combinatorial actions with the PHD of Yng2 in NuA4 and with the PHD of Rco1 in Rpd3S are necessary for the overall affinity and specificity of the H3K4me- and H3K36me-nucleosome interaction [12,16], and activation [15,38,39] of the complexes (Figure 7). A biochemical study previously demonstrated that Rpd3S preferentially binds di-nucleosomes by contacting two nucleosomes simultaneously via Eaf3 CD and Rco1 PHD [40]. Deuterium exchange mass spectrometry study also found that Rpd3S undergoes conformational changes upon contact with nucleosomes, and the Sin3-interacting domain of Rco1 allosterically stimulates preferential binding of Eaf3 CD to H3K36me [13]. These findings strongly suggest that the precise positioning of the complex on nucleosomes by multiple domains is crucial for regulation of the enzymatic activity. Our findings now open up the possibility that Eaf3 CD provides pH sensitivity to this regulation. We suspect that Eaf3 CD is needed to swiftly respond to cellular signals of pH_i perturbation and coordinate co-transcriptional chromatin remodeling. For instance, during cell proliferation when the pH_i increases, subsets of genes should be activated, requiring both dynamic opening of promoter regions by HAT for easy access of Pol II, and tight closing of coding regions by HDAC after Pol II passes through to avoid aberrant gene transcripts. Via His18, Eaf3 CD senses increasing pH_i and simultaneously strengthens its affinity for methylated nucleosomes, thereby placing NuA4 and Rpd3S in the appropriate position on the nucleosomes in combination with other subunits. Activated NuA4 on promoter regions and Rpd3S on coding regions establish an environment for efficient transcription. The generation of acetate during the histone deacetylation process might promote dissociation of Rpd3S from nucleosomes in the gene bodies. Further studies are needed to verify this notion.

In addition to Eaf3, S. cerevisiae has two well-known CD-containing proteins: Chd1, an ATP-dependent chromatin remodeling factor; and Esa1, the catalytic subunit of HAT NuA4. In previous studies, we found that both CDs are structurally incompetent to bind to lysine-methylated histone [41,42]. Given the fact that Eaf3 is a limited CD protein possessing methylated histone-binding ability in yeast, its pH-responsive interactions may have a fundamental role. This idea is supported by the fact that the significance of fluctuations in pH_i is not limited to yeast. In higher eukaryotes, changes in pH_i have been correlated with cell proliferation [43], differentiation [44], cell cycle progression [45], apoptosis [7] and cancer [46]. In this regard, it is worth noting that His18 of Eaf3 CD is highly conserved (Supplementary Figure S5) and the corresponding histidine of Eaf3 orthologs – namely, His21 in human MRG15, His62 in Arabidopsis MRG2, and His37 in rice MRG701 – actually lies at a structurally equivalent position [47–49]. In addition, the interactions between MRG15 CD and methyl-lysine peptides have been found to be strongly sensitive to the pH used for NMR titrations, with little or no interaction between MRG15 CD and H3K36me3 peptide at acidic pH [50]. On the basis of these observations, we predict that all MRG protein CDs essentially functions as a pH_i sensor identical with the role of Eaf3 CD described here.

Finally, to explore pH_i sensor activity and the correlation between pH_i and histone acetylation, we focused here on the histone methyl-lysine-binding domain and one of its histidine residues. However, this does not mean that Eaf3 CD is the only domain that acts as a pH_i sensor. It is noteworthy that the PHDs of human Death Inducer Obliterator 3 (Dido3) and its Drosophila ortholog PPS use a histidine to bind trimethylated H3K4 and exhibit pH dependence [51,52]. The Arabidopsis SHH1 tandem Tudor domain also contains a histidine in its aromatic binding cage for methylated H3K9 [53]. We speculate that other chromatin-associated proteins might use a similar pH-responsive mechanism to regulate chromatin structure.

The authors declare that there are no competing interests associated with the manuscript.

This work was supported in part by Grants-in-Aid for an NMR platform [grant number 07022019 (to Y.N.)] from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan; Basis for Supporting Innovative Drug Discovery and Life Science Research (BINDS) from the Japan Agency for Medical Research and Development (AMED), Japan under [grant number JP17am0101073 (to Y.N.)].

M.O. conceptualized the study, performed the experiments, analyzed the data, and wrote the manuscript. Y.N. formulated the project, supervised the research, and reviewed the manuscript.

Coordinates for the solution structures of Eaf3 CD have been deposited in the RCSB Protein Data Bank with accession number 6K5W. NMR factors have been deposited in the Biological Magnetic Resonance Bank with accession number 36261.

ac

acetylated

ame2

arginine-asymmetrically dimethylated

CD

chromodomain

Eaf3

essential Sas2-related acetyltransferase1-associated factor 3

H3

histone H3

H4

histone H4

HAT

histone acetyl transferase

HDAC

histone deacetylase

HSQC

heteronuclear single quantum coherence

K_d

dissociation constant

me2

dimethylated

me3

trimethylated

NMR

nuclear magnetic resonance

NOE

nuclear overhauser effect

pH_i

intracellular pH

pho

phosphorylated

Pol II

RNA polymerase II

pK_a

acid dissociation constant

PRE

paramagnetic relaxation enhancement

WT

wild-type