Tuning HP1α Chromodomain Selectivity for Di- and Trimethyllysine

Abstract

Histone lysine methylation is a critical marker for controlling gene expression. The position and extent of methylation (mono-, di-, or tri-) controls the binding of effector proteins that determine whether the associated DNA is expressed or not. Dysregulation of histone protein methylation has been associated with a number of types of cancer, and development of inhibitors for the effector proteins is becoming an active area of research. For this reason, understanding the mechanism by which effector proteins obtain selectivity for the different methylation states of lysine is of great interest. To this end, we have performed mutation studies on the Drosophila HP1α chromodomain, which binds H3K9Me₂ and H3K9Me₃ with approximately equal affinities. The selectivity of HP1α chromodomain for H3K9Me₃ over H3K9Me₂ was investigated by mutating E52 to remove or weaken the hydrogen bond to K9Me₂ while maintaining affinity for K9Me_3, including E52F, E52I, E52V, E52D, an E52Q. The E52Q mutant exhibited the greatest degree of selectivity for KMe3, with 3.5-fold weaker binding to the dimethylated peptide (K_D=52 μM) compared to the trimethylated peptide (K_D=15 μM). These studies provide insight into the role of electrostatic interactions and hydrogen bonding in the differentiation of methylation states and have implications regarding the evolutionary pressure for selectivity in this protein–protein interaction. Moreover, the information from this study may help guide inhibitor development for this class of proteins.

Introduction

Lysine methylation of histone proteins is intimately involved in controlling gene expression and aberrant methylation is associated with cancer.1–3 Lysine can have up to three different methylation states including mono- (KMe), di- (KMe₂), or trimethylation (KMe₃). Depending on the methylation state and position in the histone tail, methylated lysine recruits different effector proteins that bind directly to the methylated lysine and control expression of the associated DNA. Thus, accurate read-out of the different methyl states by effector proteins is critical for proper gene expression.

Given the importance of selective recognition of the different methylation states of lysine, as well as recent efforts to develop inhibitors of the effector proteins for methylated lysine,4–7 there is great interest in understanding the mechanism by which effector proteins achieve selectivity. The recognition of lysine methyl marks by effector proteins is mediated by contacts made between the methylammonium of the methylated lysine and an aromatic cage.8, 9 The aromatic cage is prevalent in all methyllysine reader domains. Importantly, this motif is not present in domains that target unmodified lysine such as the plant homeodomain (PHD) finger found in BHC80.10 The aromatic cage is composed of two to four aromatic amino acid residues that are involved in cation–π interactions with methylated lysine.8 Several factors contribute to selectivity, including the shape and composition of the aromatic cage, the number of cation–π interactions between the aromatic rings and the polarized (δ+) methyl groups,8, 9 the depth of the binding pocket, and the presence of hydrogen-bond acceptors and negatively charged side chains.11–15 Understanding the factors that contribute to this selectivity is critical for understanding the readout of the histone code.16

The Drosophila heterochromatin protein 1α (dHP1α) chromodomain is a surface-groove binding domain with an aromatic cage composed of two tyrosine residues, a tryptophan, and a glutamate (Figure 1).13 dHP1α exhibits minimal preference for H3K9Me₃ over H3K9Me₂.13 E52 is involved in a water-mediated hydrogen bond to dimethyllysine (Figure 1). In contrast, when in complex with trimethyllysine, the water molecule is displaced from the K9 nitrogen by over 1 Å. Trimethyllysine cannot participate in hydrogen bonding with E52, but a favorable electrostatic interaction is possible. However, mutation of glutamate to alanine has previously been shown to result in only a minimal change in binding affinity to H3K9Me₃, indicating that E52 does not contribute to binding of trimethyllysine.13 Because these experiments were not repeated with H3K9Me₂, the contribution of the water-mediated hydrogen bond and electrostatic interaction between E52 and dimethyllysine has not been established.

Figure 1. Structure of H3K9Me₂ (dark gray) and H3K9Me₃ (light gray) in the aromatic cage of dHP1α chromodomain (dark and light gray, respectively). The water molecule is shown as a dark gray sphere for KMe₂ and a light gray sphere for KMe₃. The water-mediated hydrogen bond/electrostatic interaction between Glu52 and K9Me_2/3 is depicted with dashed lines. PDB IDs for chromodomain bound to H3K9Me₂ and H3K9Me₃ are 1KNA and 1KNE respectively.13

Herein, we investigate the role of E52 in recognition of K9Me₂ and K9Me₃ by the dHP1α chromodomain. E52 was mutated to residues that are unable to form strong hydrogen bonds to dimethyllysine, but might still be able to maintain favorable interactions with trimethyllysine. These mutations included E52F, E52I, E52V, E52D, and E52Q. These studies provide new insight into how effector proteins achieve selectivity for various lysine methylation states and the role of hydrogen bonding and electrostatic interactions in providing selectivity. Moreover, these studies demonstrate that the level of selectivity for KMe₃ over KMe₂ can be enhanced through modest mutations even in a surface groove binding site, suggesting that selectivity between KMe₃ and KMe₂ is not necessary in the dHP1α chromodomain.

Results and Discussion

Design

A range of hydrophobic and polar mutations were incorporated at position E52 of the dHP1α chromodomain, including phenylalanine, tryptophan, isoleucine, valine, aspartic acid, and glutamine substitutions. E52F and E52W mutations were included as these residues may provide a fourth aromatic residue to the aromatic cage, which would favor binding to KMe₃, while preventing hydrogen bonding to KMe₂.17 The hydrophobic mutants E52I and E52V were studied as these two residues may maintain favorable van der Waals contacts with the methyl groups on K9Me3. However, isoleucine and valine lack a hydrogen bond donor or acceptor, and are unable to form the water-mediated hydrogen bond to K9Me₂ and therefore should reduce binding to H3K9Me₂ relative to H3K9Me₃.

The polar mutants E52D and E52Q were investigated to determine the influence of both the charge and the water-mediated hydrogen bond formed to E52. Although aspartate is capable of forming a hydrogen bond to water, the shortened distance between E52D and K9Me₂ may disrupt the hydrogen bond network responsible for stabilizing the dHP1α chromodomain–H3K9Me₂ complex. This in turn would disfavor binding to K9Me₂ relative to K9Me3. E52Q was included because an amide is a considerably weaker hydrogen bond acceptor than a carboxylic acid and should have weakened binding to KMe₂. Given that the E52A mutant did not reduce affinity for KMe₃,9 the loss of negative charge in the E52Q mutation was not expected to impact binding to KMe₃.

Structural characterization

Circular dichroism was used to characterize the global structure of the chromodomain mutants. Because the signal from the α-helix at 208 and 222 nm was expected to mask any signal from β-sheet structure at 215 nm, only two minima at approximately 208 and 222 nm were expected to be observed.18 With the exception of E52W, these minima are present and similar in the spectra of the E52 mutants (Figure 2). The CD spectra indicate that the E52W mutation has an overall destabilizing effect on the global structure of the chromodomain and therefore no further studies were performed using this mutant. The peak at 232 nm is consistent with exciton coupling, which arises from interactions between aromatic residues. Therefore, the variation that is observed at 232 nm with E52W and E52I might be due to a perturbation of the cage.

Figure 2. Circular dichroism spectra comparison of wild-type dHP1α chromodomain and the mutants E52W, E52F, E52I, E52V, E52D, and E52Q at 25 °C in sodium phosphate (10 mm), DTT (2 mm), pH 7.4. Spectra are the averages of three scans.

Thermal denaturation experiments were conducted to assess protein stability of the mutants (Figure 3). The melting temperatures were estimated from the inflection point of each curve (Table S3). Although there is some deviation from the melting temperature of the native chromodomain (47 °C) ranging from 39–53 °C, all chromodomain mutants are fully folded at 25 °C, which is the temperature at which all subsequent binding experiments were conducted. It is interesting to note that the E52F is slightly more stable with a melting temperature of 53 °C, which may be due to aromatic interactions with the additional phenylalanine residue.

Figure 3. Thermal denaturation monitored by circular dichroism at 222 nm of wild-type dHP1α chromodomain and the mutants E52F, E52I, E52V, E52D, and E52Q in sodium phosphate (10 mm), DTT (2 mm), pH 7.4.

Binding studies

Fluorescence anisotropy was used to determine the dissociation constants between the E52 mutants and the H3K9Me_2/3 peptides modified with carboxyfluorescein at the N terminus (see the Supporting Information). The K_D values are nearly identical for both the di and tri-methylated peptides with the wild-type chromodomain (K_D=20 and 17 μM, respectively) thus demonstrating the lack of selectivity that the native chromodomain has for the higher methylation states (Table 1, Figure S1).

Table 1. Dissociation constants [μM], ΔG°_binding [kcal mol⁻¹] (in parentheses), and selectivity ratio (K_D (KMe2)/K_D (KMe3)) of the dHP1α chromodomain mutants with H3K9Me_2/3 and the mutants.^[a]

Chromodomain	KD (ΔG°binding)		Selectivity
	H3K9Me2	H3K9Me3
[a] The data were measured in potassium phosphate buffer (50 mm, NaCl (25 mM), DTT (4 mm), pH 8.0) at 25 °C. The ΔG°binding [kcal mol^−1] was calculated from the following equation: ΔG°=RT ln KD.
wild-type	20±4	17±3	1.2
	(−6.4±0.1)	(−6.5±0.1)
E52F	52±5	40±9	1.2
	(−5.8±0.1)	(−6.0±0.1)
E52I	83±20	47±8	1.6
	(−5.6±0.1)	(−5.9±0.1)
E52V	67±11	39±9	1.5
	(−5.7±0.1)	(−6.0±0.1)
E52D	19±2	9±1	2.0
	(−6.4±0.1)	(−6.9±0.1)
E52Q	52±7	15±3	3.4
	(−5.8±0.1)	(−6.6±0.1)

Chromodomain E52 mutations involving the three hydrophobic residues, E52F, E52I, and E52V, exhibit a two- to threefold decrease in binding affinity to both dimethyl and trimethyllysine histone tails, resulting in only a small increase in selectivity for KMe₃ (Table 1, Figures S2–S4). It is possible that substituting glutamate with bulky hydrophobic residues may distort the aromatic cage or simply hinder access to the binding pocket. Indeed, distortion of the aromatic cage is consistent with the circular dichroism spectra for E52I in which the exciton coupling peak at 232 nm, which is known to be influenced by the orientation of neighboring aromatic residues, is weaker than for the wild-type protein.19 Loss of the nearby negative charge may also weaken binding to both KMe₂ and KMe₃ due to loss of a favorable electrostatic interaction. This possibility is addressed more thoroughly with the E52D and E52Q mutants discussed below.

Mutant E52D had a minimal effect on binding affinity to H3K9Me₂, with a K_D of 19 μM (Table 1, Figure S5). This suggests that the aspartate is able to maintain the water-mediated hydrogen bond to the dimethyllysine. Surprisingly, binding of E52D to K9Me₃ is improved by a factor of two, with a K_D of 9 μM. This may reflect a reduced desolvation penalty of the carboxylate in E52D, since E52 is in van der Waals contact with a methyl group of KMe₃ in the wild-type protein, whereas the carboxylate of E52D is farther away from the methyl group in the mutant. Thus, E52D will likely be better solvated by water than in the wild-type protein.

Interestingly, chromodomain E52Q has a 2.5-fold weaker dissociation constant with H3K9Me₂ (52 μM) compared to the native protein. In contrast, the dissociation constant with H3K9Me₃ (15 μM) is unaffected by the mutation (Figure S6), resulting in about threefold selectivity for KMe₃ over KMe₂. Glutamine, though a polar amino acid, is not as strong a hydrogen bond acceptor as glutamate. Thus, this result indicates that the water-mediated hydrogen bond plays measurable role in chromodomain binding to H3K9Me₂, and confirms that the possible electrostatic interaction between E52 and KMe₃ is unimportant for binding to KMe₃. The energy difference between the wild-type protein–protein interaction and the E52Q–KMe₂ binding affinities gives an estimated magnitude of the water-mediated hydrogen bond between E52 and KMe₂ of −0.6 kcal mol⁻¹.

Comparison to other methyllysine effector proteins

The role of hydrogen bonding and electrostatic interactions in the recognition of methylated Lys has been investigated in several other effector proteins. Li and co-workers manipulated the composition of the BPTF PHD finger aromatic cage to reverse its selectivity for di- and trimethyllysine.20 The wild-type PHD finger binds H3K4Me₃ and H3K4Me₂ in a surface groove with K_D values of 3 μM and 6 μM respectively. However, by substituting Y17 of the aromatic cage for glutamate, the selectivity of the PHD finger was reversed to favor H3K4Me₂ (K_D=7 μM) over H3K4Me₃ (K_D=15 μM). Comparing the crystal structures of the wild-type and Y17E PHD finger confirmed that the mutation results in a direct hydrogen bond between Y17E and H3K4Me₂.20 The destabilizing effect of the Y17E mutation on binding to KMe₃ may reflect both the loss of the cation–π interaction and a steric clash between KMe₃ and glutamic acid, as the additional methyl group on KMe₃ is larger than the proton in KMe₂.

Further mutation studies of the Y17E PHD finger probed the role of hydrogen bonding and electrostatics on binding of KMe₂ and KMe₃.20 The Y17Q mutation weakens binding to both KMe₂ and KMe₃, with K_D values of 19 and 27 μM, respectively. In each case the change in binding affinity relative to the Y17E mutant is approximately 0.5 kcal mol⁻¹, suggesting that a favorable hydrogen bond and/or electrostatic interaction contributes to binding of both KMe₂ and KMe₃. This is unlike the results from the current study, in which the E52Q mutant only destabilizes binding to KMe₂, and highlights the difference between a direct hydrogen bond and a water-mediated hydrogen bond.

In the same protein, the Y17D mutation reduces binding to KMe₂ by about 1 kcal mol⁻¹ relative to Y17E, but only by about 0.4 kcal mol⁻¹ in the case of KMe₃. Although the absolute magnitudes are different than observed in the current study, the trend is the same: shortening the side chain from glutamic acid to aspartic acid results in an approximately 0.5 kcal mol⁻¹ greater destabilization for KMe₂ than KMe₃.

Patel and co-workers also investigated the role of hydrogen bonding and electrostatic interactions in the binding of KMe₂ to pocket 2 of L3MBTL1.20 This effector protein binds to KMe and KMe₂ of several different peptides in a deep binding cleft with an aromatic cage, but does not bind KMe₃. D355 makes a direct hydrogen bond to KMe and KMe₂. Mutation of D355 to asparagine weakens binding to KMe₂ by about 1 kcal mol⁻¹ for a range of different peptide sequences. This is larger than the effect observed in the dHP1α chromodomain studied here or in the PHD mutant described above. This likely reflects the difference between an electrostatic interaction in a solvent-exposed surface groove, as in dHP1α and the PHD finger, relative to a deep cleft, as in L3MBTL1, where the charges are less well solvated.

Comparison of these three systems provides insight into the role of hydrogen bonding and electrostatic interactions in direct versus water-mediated hydrogen bonds as well as the influence of buried versus surface-exposed binding pockets. Electrostatic effects were found to contribute equally when binding to KMe₂ or KMe₃ through a direct interaction, but contribute less to binding of KMe₃ (which is only electrostatic in nature) when involved in a water-mediated interaction, than to KMe₂ (in which both hydrogen bonding and electrostatic effects contribute). In addition, electrostatic interactions were found to be more significant within deeper pockets, consistent with the relative contributions of solvation to stabilization of a buried versus solvent-exposed charge.

Conclusions

The aromatic cage within the dHP1α chromodomain is critical for recognition of di- and trimethyllysine 9 on the histone H3 tail and has been shown to have only a small preference for trimethyllysine.13 Introducing glutamine in place of glutamate was found to be a successful method for enhancing the selectivity of the dHP1α chromodomain for KMe₃ over KMe₂, despite the loss of electrostatic interactions in both cases. This study clearly indicates that differentiation between di- and trimethyllysine is possible even with single mutations. The fact that such differentiation is possible suggests that specific recognition of KMe₂ or KMe₃ is not necessary in this protein–protein interaction. Indeed, biological studies support the fact that K9Me₂ and K9Me₃ behave similarly.15

This study also provides broader insight into the role of both hydrogen bonding and electrostatic interactions in differentiation of di- and trimethyllysine. Hydrogen bonding is only possible with dimethyllysine, but an electrostatic interaction is possible in both cases. Comparison of the dHP1α chromodomain, PHD finger, and L3MBTL1 suggests that the contribution of the electrostatic interaction in binding KMe₂ and KMe₃ is dependent on the depth and solvent accessibility of the binding cleft. This provides yet another method by which nature tunes for selectivity for the different methylation states of lysine, resulting in control of gene expression through these subtle molecular marks. The findings from this study should provide insights necessary for successful inhibitor development for this type of protein–protein interaction.

Experimental Section

Synthesis and purification of peptides: Peptides were synthesized using automated solid phase peptide synthesis with a Thuramed tetras synthesizer on a 0.06 mmol scale on Wang resin. Fmoc protected amino acids and resin were purchased from Peptides International, Inc. Amino acid residues were activated with O-benzotriazole-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HBTU) and N-hydroxybenzotriazole (HOBt) with diisopropylethylamine (DIPEA) in N,N-dimethylformamide (DMF). Amino acids were deprotected twice with 2 % 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) and 2 % piperidine in DMF for 15 min each step. Each amino acid was coupled on double cycles of 30 min each to improve coupling. Fmoc-protected dimethyllysine was coupled at position 9, which was methylated before the final Fmoc deprotection with 7-methyl-1,5,7-triaza-bicyclo[4.4.0]dec-5-ene (MTBD, 10.8 μL) and iodomethane (37.4 μL) in DMF (5 mL) and bubbled with N₂ for 5 h. After which, the final Fmoc was cleaved and 5(6)-carboxyfluorescein (5(6)-FAM) (from Anaspec) was coupled using 2 equivalents of the fluorophore, HOBt, HBTU, and DIPEA overnight. Cleavage of the peptides from the resin was performed in 95 % trifluoroacetic acid (TFA), 2.5 % H₂O, 2.5 % triisopropylsilane (TIPS) for three hours. TFA was evaporated with a stream of nitrogen and diethyl ether was used to precipitate the resulting product. The product was extracted with water and lyophilized to a powder.

Peptides were purified by reversed-phase HPLC. An Atlantis Prep OBD dC-18 semi-preparative column was used for separation with a gradient of 0 % to 100 % solvent B over 60 min with solvent A (water/acetonitrile 95:5, 0.1 % TFA) and solvent B (acetonitrile/water 95:5, 0.1 % TFA). Peptides were then lyophilized and the peptide sequence was confirmed by ESI-TOF mass spectrometry (Table S1).

Protein expression and purification: The dHP1α chromodomain construct with a His₆-tag was supplied by the Khorasanizadeh lab (University of Virginia). Mutagenesis was carried by QuikChange site-directed mutagenesis (Strategene). The primers that were used are listed in Table S2 (Supporting Information) and all mutations were verified by next-generation sequencing. The mutated plasmids were transformed into E. coli BL21 (DE3) gold cells. The cells were grown to an A₆₀₀ value of 0.5–0.7 in LB medium containing ampicillin. The cells were induced with IPTG (100 μm) and grown overnight at 18 °C and harvested by centrifugation (20 min at 5900 g, 4 °C). The cells were suspended in NiA buffer (KPO₄ (50 mm), NaCl (150 mm), imidazole (5 mm), pH 8.0, 40 mL) and were lysed with in the presence of DNAse, complete, EDTA-free protease inhibitor cocktail (Roche Applied Sciences) and PMSF (final concentration 1 mm). The protein was purified using Ni-NTA resin with a step gradient of 0–250 mM imidazole using NiA and NiB buffer (KPO₄ (50 mm), NaCl (150 mM), imidazole (250 mm), pH 8.0). The purity was confirmed by SDS-PAGE and the fractions were pooled and dialyzed in 100-fold excess of buffer (KPO₄ (50 mm), NaCl (25 mm), pH 8.0). The dialyzed protein was concentrated using Millipore Centricon filters (5 KDa MWCO).

Circular dichroism: CD measurements were performed on an Aviv 62DS circular dichroism spectrometer. CD data were obtained for the chromodomain (33.3 μm) in solution containing Na₂HPO₄ (10 mM), dithiothreitol (DTT, 2 mM), at pH 7.4. Wavelength scans were performed in triplicate and averaged. Scans were performed at 25 °C. All scans were corrected by subtracting the spectrum of the buffer used in the experiment. The mean residue ellipticity (MRE) was calculated from Equation (1).

• ((1))

where Θ is MRE, signal is CD signal, l is path length, c is protein concentration, and r is the number of amino acid residues.

Thermal denaturation experiments were preformed using the same buffer and concentrations as described above and measurements were taken between 3 °C and 93 °C. The melting curves were normalized to show the fraction folded by using Equation (2).

• ((2))

where Θ is the observed MRE, Θ_D is the MRE for the fully denatured protein, and Θ_F is MRE for the fully folded protein.

Fluorescence anisotropy: Peptide concentrations were determined by UV/VIS using the absorbance of 5(6)-FAM (ε₄₉₂=78 000 M⁻¹ cm⁻¹). Fluorescence anisotropy measurements were taken using a PolarStar omega plate reader by BMG Labtech. Samples were prepared by titrating chromodomain into 1 μM histone peptides labelled with 5(6)-FAM on the N terminus in buffer (KPO₄ (50 mM), NaCl (25 mm), DTT (4 mm), pH 8.0). After allowing the samples to equilibrate for 30 min at 25 °C, they were analyzed using an excitation wavelength of 485 nm and an emission wavelength of 520 nm at 25 °C. The data were analyzed in Kaleidagraph by using Equation (3). All binding curves are shown in Figures S1–S6.

• ((3))

where r is fluorescence anisotropy, r₀ is the anisotropy of free histone tail, r_∞ is the anisotropy of fully bound histone tail, p is the total concentration of fluorescein-labeled peptide, [c] is the total chromodomain concentration added, and K_D is the dissociation constant. The values for r₀, r_∞, and K_D were treated as floating variables.

Acknowledgements

We gratefully acknowledge support from the National Science Foundation (CHE-0716126). We thank Prof. Matthew Redinbo and Dr. Sarah Kennedy for sharing of equipment and assistance with mutagenesis experiments.