An engineered variant of SETD3 methyltransferase alters target specificity from histidine to lysine methylation

Most characterized SET domain (SETD) proteins are protein lysine methyltransferases, but SETD3 was recently demonstrated to be a protein (i.e. actin) histidine-N3 methyltransferase. Human SETD3 shares a high structural homology with two known protein lysine methyltransferases—human SETD6 and the plant LSMT—but differs in the residues constituting the active site. In the SETD3 active site, Asn255 engages in a unique hydrogen-bonding interaction with the target histidine of actin that likely contributes to its >1300-fold greater catalytic efficiency (kcat/Km) on histidine than on lysine. Here, we engineered active-site variants to switch the SETD3 target specificity from histidine to lysine. Substitution of Asn255 with phenylalanine (N255F), together with substitution of Trp273 with alanine (W273A), generated an active site mimicking that of known lysine methyltransferases. The doubly substituted SETD3 variant exhibited a 13-fold preference for lysine over histidine. We show, by means of X-ray crystallography, that the two target nitrogen atoms—the N3 atom of histidine and the terminal ϵ-amino nitrogen of lysine—occupy the same position and point toward and are within a short distance of the incoming methyl group of SAM for a direct methyl transfer during catalysis. In contrast, SETD3 and its Asn255 substituted derivatives did not methylate glutamine (another potentially methylated amino acid). However, the glutamine-containing peptide competed with the substrate peptide, and glutamine bound in the active site, but too far away from SAM to be methylated. Our results provide insight into the structural parameters defining the target amino acid specificity of SET enzymes.

Enzymes that create posttranslational methylations on proteins, particularly histones, play a pivotal role in regulating gene expression and chromatin organization (1)(2)(3). For example, the histone lysine methyltransferases that catalyze site-specific lysine methylation have been extensively studied over the past two decades, ever since the discovery of the Suv39H1 as a histone H3 lysine 9 -specific methyltransferase (MTase) 3 (4). The vast majority of characterized histone lysine MTases contain an ϳ130-residue SET domain that possesses the methylation activity (5,6), with the exception of DOT1L (on H3K79) (7)(8)(9) and KMT9 (on H4K12) (10). In humans, approximately half of the 55 SET domain family members methylate lysine residues on histone and/or nonhistone proteins (11,12). Recently, SETD3 was identified as the first metazoan histidine MTase that works on an actin histidine residue, which promotes signal-induced smooth muscle contraction and in a catalytic-independent manner is important for virulence of enteroviruses (13)(14)(15). Structurally, SETD3 shares a high degree of global similarity with two characterized SET domain proteins (14,16,17), namely human SETD6 and rubisco LSMT, which act respectively on lysine residues of the RelA subunit of nuclear factor NF-B and the large subunit of rubisco (18 -22). Here we investigate the structural and molecular determinant(s) of target specificity of histidine versus lysine versus glutamine in the active site of SETD3.
Most, if not all, enzymatic reactions of SAM-dependent MTases, including those of histidine methylation catalyzed by SETD3 and the lysine methylation catalyzed by SETD6 and LSMT, are thought to proceed with direct transfer of the methyl group to substrate from the methyl donor S-adenosyl-L-methionine (SAM) (23). This reaction also requires a deprotonation step, in which a proton is removed before, concurrent with, or after methyl transfer. Even within the structurally conserved family of SET domain MTases, a variety of mechanisms have evolved to activate the catalytic nucleophile, dependent on the polarizability of the target atom. For example, lysine methylation by DIM-5, LSMT, and SETD6 showed maximal in vitro activity at approximately pH 10, principally because of the pK a value of ϳ10 for a lysine substrate (19,24,25). Histidine methylation by SETD3 has an optimum pH of 7 and above, in agreement with the imidazole ring having a typical pK a value near 6 (17). Another aspect of these reactions is that the MTases have to position the target atom such that the lone-pair electrons on the (nitrogen) nucleophile point toward, and within a short distance of, the incoming methyl group. The methyl transfer is followed by an attack on the positively charged sulfonium of SAM with an inversion of symmetry in a S N 2-like mechanism (26). Here we compare the active-site configurations of the histidine methyltransferase SETD3 with the lysine methyltransferases SETD6 and LSMT by mutagenesis and assay its activities on histidine, lysine, and glutamine (as another physiologically methylated amino acid) in the context of actin peptide.

Comparison of active sites of SETD3, SETD6, and LSMT
Our previous structural characterizations of SETD3 included the prereactive substrate complex and the postreactive product complex containing an actin peptide encompassing target His 73 of actin (17). The histidine imidazole ring contains two nitrogen atoms (N 1 and N 3 ), both of which can be protonated. To ensure that the target atom N 3 is deprotonated prior to the methyl transfer, Asn 255 forms a hydrogen bond to the protonated N 1 nitrogen (Fig. 1A). With substitution of Asn 255 to alanine (N255A), valine (N255V), or phenylalanine (N255F), SETD3 loses the ability to form this hydrogen bond; this results in reduced activity for His 73 methylation with slower k cat values (N255A and N255V), and surprisingly for N255F, no measurable activity was observed for histidine methylation ( Fig. 1, B-D). We reasoned that the bulky side chain of phenylalanine at residue 255 could collide with neighboring Trp 273 (Fig. 1E). Structure-based sequence alignment indicated that three of four residues (Asn 255 , Trp 273 , Ile 310 , and Tyr 312 ) that form the active-site pocket in SETD3 are not conserved in LSMT and SETD6 (Fig. 1F). The corresponding SETD3 residues Asn 255 and Trp 273 are phenylalanine and alanine in both LSMT and SETD6 (Fig. 1G). Thus, we generated a variant of SETD3 with two alterations in the active site, N255F and W273A.

SETD3 and the N255F/W273A variant have opposite activity on His and Lys methylation
We first measured the activities of SETD3 (WT) and the double mutant on actin peptide residues 66 -80 containing His 73 . Unlike the N255F single mutation, the double mutant showed a measurable activity on His 73 with k cat of 2.8 h Ϫ1 and K m of 54 M ( Fig. 2A). Nevertheless, the double mutant exhibited a reaction rate of ϳ15-fold slower (k cat ) and an affinity for the substrate ϳ2-fold weaker (K m ) than that of WT SETD3 under the same conditions, resulting in ϳ36-fold loss in catalytic efficiency (comparing k cat /K m values) for His 73 methylation ( Fig. 2A).
Because the double mutant was designed to mimic the active sites of LSMT and SETD6, both of which catalyze protein lysine methylation, we asked whether the double mutant is capable of methylating lysine in the place of His 73 (H73K) in the context of the same actin peptide residues 66 -80. Interestingly, the activ-

SETD3 target specificity
ities of WT and double mutant SETD3 on Lys 73 methylation flipped the order of preference (Fig. 2B). The double mutant demonstrated a ϳ500-fold gain in catalytic efficiency for Lys 73 methylation, compared to that of WT (k cat /K m value of 0.66 h Ϫ1 M Ϫ1 for the mutant and 1.3 ϫ 10 Ϫ3 h Ϫ1 M Ϫ1 for the WT). The gain in catalytic efficiency is driven by an improved reaction rate (ϳ105ϫ in k cat ) and stronger binding for the substrate (ϳ5ϫ in K m ). In essence, the double mutant variant of SETD3 has switched the identity of SETD3 from a histidine MTase into a lysine MTase. We note that the WT SETD3 demonstrated ϳ1385-fold preference of His 73 over Lys 73 methylation in catalytic efficiency, whereas the mutant showed a ϳ13-fold preference of Lys 73 over His 73 methylation. Thus, comparing the mutant to WT, the ratio of catalytic efficiency on Lys 73 to His 73 methylation is increased by ϳ18,000-fold ( ϭ 1385 ϫ 13).

Structure of the N255F/W273A variant in complex with Lys 73containing peptide
To facilitate co-crystallization, we increased the peptide length to actin residues 66 -88. As observed in the previous structural work, the additional C-terminal actin residues engage in additional intermolecular interactions with SETD3 (16,17). The enhanced enzyme-peptide interactions increased binding affinity by decreasing the K m value but do not change the reaction rate by maintaining the same k cat value (Fig. 2C).
The double mutant was readily crystallized with the Lys 73 (66 -88) peptide in the presence of SAH (Fig. 3A). The mutant-Lys 73 structure is highly similar to those of WT enzyme in com-plex with His 73 (PDB ID 6MBL), with pairwise comparison of ϳ0.4 Å across 458 pairs of C␣ atoms. The target lysine residue is inserted into the active-site channel, where at the end of the channel the terminal ⑀-amino nitrogen atom meets the cofactor from the opposite end (Fig. 3B). The channel is bordered by the aromatic residues of Tyr 312 and N255F, which pack against the aliphatic portion of the target lysine. The side chains of the two mutated residues, N255F and W273A, are sufficiently separated (Fig. 3C). Superimposition of WT-His 73 and the mutant-Lys 73 complex structures indicated that the two target nitrogen atoms, i.e. the terminal ⑀-amino nitrogen of lysine and the N 3 atom of unmethylated histidine ring, reside at nearly the same position (Fig. 3D). The side chain aliphatic carbons of lysine trace along the edge of superimposed histidine ring in five-bond distances from the main chain C␣ atom to the target nitrogen atom (Fig. 3D). The ⑀-amino group of the target lysine is 3.4 Å away from the sulfur atom of SAH, where a transferable methyl group would be attached (Fig. 3E). The distance between the sulfur of SAH and N⑀ of lysine is approximately the sum of the bond distance of donor-methyl (S ϩ -CH 3 ϭ 1.82 Å) and the bond distance of acceptor-methyl (CH 3 -N ϩ ϭ 1.47 Å). A water molecule (w1), coordinated by the main-chain carbonyl oxygen atoms of Cys 276 and W273A, could facilitate the deprotonation of the target ⑀-amino group of lysine during catalysis (Fig. 3, E  and F).
Comparisons of the double mutant structure of SETD3 to that of LSMT and SETD6 show that the features of the active-

SETD3 target specificity
site configurations for lysine methylation are highly preserved (Fig. 4). The aromatic side chains of Tyr-Phe pair (Tyr 312 -N255F in SETD3, Tyr 287 -Phe 224 in LSMT, and Tyr 285 -Phe 225 in SETD6) guide the target lysine into a narrow channel. In addition, the hydroxyl oxygen atom of the Tyr interacts and stabilizes the positive charge on the SAM methylsulfonium group (CH 3 -S ϩ ). As a result, the deprotonated amino group (NH 2 ) of the target lysine is positioned at the right distance to be able to nucleophilically attack the positively charged SAM methylsulfonium without any general base.

SETD3 is not active on glutamine methylation
Human HemK2 has recently been documented to be a histone H4 lysine 12 MTase (renamed as KMT9) (10) in addition to its known activity of glutamine methylation of eukaryotic release factor eRF1 (27, 28). We confirmed that HemK2 is active on glutamine and lysine (29) and note the unique property of this MTase by working on two different substrates with different target residues; one is located in the cytoplasm (eRF1, glutamine) and the other in the nucleus (histone H4, lysine). The common feature of the two potential substrates is the amino group (NH 2 ) of glutamine and lysine (Fig. 5A). Unlike SET domain MTases, HemK2/KMT9 is a seven-␤-stranded family MTase (23). As shown in Fig. 3D, the side chain of the lysine residue, with a length of 5-bond distance between the C␣ atom and the terminal N⑀ nitrogen, adopts a bent conformation such that its aliphatic carbons could seemingly trace along one edge of the histidine imidazole ring. We asked whether the side chain of a glutamine residue, with a length of 4-bond distance, could trace along the shorter edge of the imidazole ring to reach the target position in the active site of SETD3 (Fig. 5A).
In the context of the same actin peptide residues 66 -88, we replaced the His 73 by glutamine (Gln 73 ). Under the saturating conditions of an overnight reaction, in which SETD3 totally completes reaction on all given His 73 substrate, we observed no activity of SETD3 or its mutants on the Gln 73 peptide (Fig. 5B). However, we did observe direct binding between SETD3 and the Gln 73 peptide with a dissociation constant (K D ) of ϳ0.7 M by isothermal titration calorimetry (Fig. 5C). The Gln 73 peptide competes with the substrate peptide by inhibiting SETD3 activity on His 73 with a half-maximal inhibitory concentration (IC 50 ) of ϳ1 M (Fig. 5D).
To further understand the structural basis of the binding of SETD3 with Gln 73 -containing peptide, we crystallized and determined the inactive complex structure at the resolution of 2.0 Å (Table S1). The Gln 73 structure is essentially the same to that of His 73 -containing structure (root mean square deviation ϭ 0.169 Å over 886 pairs of C␣ atoms with two complexes per crystallographic asymmetric unit). The side chain of Gln 73 occupies the active site, but its amide group is Ͼ5 Å away from the sulfur of SAH (Fig. 5E). At the resolution of 2.0 Å, we were not able to determine the exact nature of oxygen versus nitrogen atom of Gln 73 side chain. However, the chemical nature of interacting functional groups, particularly the side chain of Asn 255 and its associated water molecule, allowed us to position the Gln 73 side chain as shown in Fig. 5E. The water molecule (w2) is saturated with tetrahedral coordination of four hydrogen bonds: two H-bond donors to the main-chain carbonyl oxygen atom of Arg 253 and one of the side chain carboxylate oxygen atoms of Asp 274 and two H-bond acceptors from the main chain amide nitrogen of Ile 270 and side chain amide nitrogen of Asn 255 (Fig. 5E). Asn 255 forms a weak hydrogen bond (via its side chain oxygen atom) with the amide group of Gln 73 , which is positioned more than 5 Å away from the sulfur atom of SAH where a transferable methyl group would be attached. In com-

SETD3 target specificity
parison, the corresponding distance is ϳ3.4 Å between the ⑀amino group of the target lysine (or the N 3 of histidine) and the sulfur atom of SAH (Fig. 3D).
Although we initially used SAH in the mixture for the crystallization, we were surprised to find that the difference electron density clearly shows one distinguishable location with

SETD3 target specificity
F o Ϫ F c at 4 above the mean, ϳ1.8 Å from the sulfur atom of SAH ( Fig. 5F; we note again the bond distance of S ϩ -CH 3 is 1.82 Å). We interpreted that this density most likely be the methyl group of donor SAM (Fig. 5G), which might co-purify with SETD3 endogenously (as with SETD6 (19)) (Fig. 5H). We reasoned that the endogenous SAM cofactors were not fully exchanged during the sample preparation because Gln 73 -containing peptide is not a substrate.
We superimposed the Gln 73 structure to that of SETD3 bound with methylated His 73 (PDB ID 6OX2), which revealed the difference density is immediately adjacent to the methyl group attached to the N 3 atom of the methylhistidine (Fig. 5I). This comparison illustrated that the target nitrogen atom has to be primed in a position toward the incoming methyl group within a short distance during the catalysis. The inactivity on Gln 73 presumably reflects the fact that the amide nitrogen of Gln 73 is not positioned in close enough proximity to the donor methyl group of SAM for nucleophilic attack (Fig. 5I). We rotated the side chain torsion angles of Gln 73 to trace the edge of the histidine ring and forced the amide nitrogen onto the N 3 atom of His 73 (Fig. 5J). In this assumed conformation, the side chain oxygen atom of Gln 73 points directly toward and smashes with the aromatic ring of Tyr 312 . As noted above, Tyr 312 is the only active-site residue conserved among SETD3, SETD6, and LSMT (Fig. 1G), as well as almost all SET domain MTases (6), and is essential for catalysis by binding to SAM (24) with the hydroxyl oxygen fitting between the two moieties of the cofactor (Fig. 5E). Thus, the rigid Tyr 312 would push away the side chain of Gln 73 and restrain it in nonproductive position as observed in the crystal.

Discussion
A wide variety of macromolecules, including DNA, RNA, proteins, polysaccharides, lipids, and a range of small molecules are subject to methylation by highly specific SAM-dependent MTases acting on a particular target atom. Examples of methylation targets include nucleic acids (cytosine-C5, cytosine-N4, and adenine-N6), protein residues (arginine-N, lysine-N, glutamine-N, and histidine-N), and small molecules (catechol-O, histamine-N, glycine-N, and thiopurine-S). The SET domaincatalyzed protein methylation has demonstrated that these enzymes frequently have high substrate specificity because of recognition of the various sequences surrounding the target residue (30 -34). SETD3 displays a greater substrate specificity than most by recognition of 17 ordered residues of a 23-residue long peptide used in this study (Fig. 3A), suggesting that SETD3 would inefficiently accommodate substrates divergent from actin.
Here, we investigated the determinant(s) of target specificity of histidine versus lysine versus glutamine in the active site of SETD3 by replacing the His 73 of actin with lysine (Lys 73 ) or glutamine (Gln 73 ). We engineered variants of SETD3 that have altered specificity. Once in the active site, the N 1 atom of the imidazole ring hydrogen bonds to Asn 255 of SETD3 that makes the histidine residue a preferred target (17). Substitution of Asn 255 with phenylalanine (N255F) together with substitution of Trp 273 with alanine (W273A) generates an active site mimicking known lysine MTases. The aromatic side chains of N255F and Tyr 312 form the wall of a narrow channel that the aliphatic part of the target lysine side chain is well-accommodated. We have also shown that positioning the target nitrogen atom in the right position and distance is fundamental for its acceptance of the methyl group from the donor SAM. The N 3 atom of histidine ring and the terminal N⑀ of lysine, both of which are substrates, can occupy the same position, whereas the amide nitrogen atom of glutamine is displaced because of the physical hindrance of an essential Tyr 312 conformation. In summary, our studies delineate the governing principles of how SETD3 and its derivatives discriminate different target residues by organizing the architectures of the active sites.
Single crystals were flash frozen in liquid nitrogen by equilibrating in a cryoprotectant buffer containing the crystallization solution and 25% (v/v) ethylene glycol. X-ray diffraction data were collected at the SER-CAT beamline 22ID of the Advanced Photon Source at Argonne National Laboratory. Crystallographic datasets were first processed with HKL2000 (35). Molecular replacement was performed with PHENIX PHASER module (36) by using the known structure of human SETD3 (PDB ID 6OX3) as the search model. Structure refinement was performed with PHENIX Refine (37) with 5% randomly chosen reflections for the validation by the R free value. COOT (38) was used for the manual building of the structure model and cor-SETD3 target specificity rections between refinement rounds. Structure quality was analyzed during PHENIX refinements and finally validated by the PDB validation server. Molecular graphics were generated by using PyMol (Schrödinger, LLC).

Steady-state kinetics
Reaction mixtures contained 20 mM Tris/HCl, pH 8.0, for His 73 peptide or 20 mM glycine/NaOH, pH 10.5, for Lys 73 peptide, 50 mM NaCl, 0.1 mg/ml BSA, 1 mM DTT, 0.18 -15 M enzymes (SETD3 WT or mutants; details in legends of Figs. 1C and 2), 40 M SAM, and varying concentration of peptides. The reactions were carried out at 37°C for different times (20 min to 3 h; details in legends of Figs. 1C and 2) with a total reaction volume of 20 l, and terminated by the addition of TFA to 0.1% (v/v) for Tris/HCl, pH 8.0, or 0.4% (v/v) for glycine/NaOH (pH 10.5). Samples terminated with the addition of 0.4% (v/v) TFA were then further diluted 4ϫ with buffer 20 mM Tris/HCl, pH 8.0, 50 mM NaCl, 0.1 mg/ml BSA, 1 mM DTT to reduce the TFA concentration. The methylation activity was measured using the Promega bioluminescence assay (MTase-Glo TM ) in which the reaction byproduct SAH is converted into ATP in a twostep reaction and ATP can be detected through a luciferase reaction (39). In general, 5 l of reaction mixture was transferred to a low volume 384-well plate and the luminescence assay was performed according to the manufacturer's protocol. A Synergy 4 Multi-Mode Microplate Reader (BioTek) was used to measure luminescence signal. The dependence of the velocity of product formation per enzyme on substrate concentration was analyzed according to the Michaelis-Menten equation.

Assay of SETD3 on Gln 73 peptide
To test whether SETD3 could methylate Gln 73 in the context of actin peptide residues 66 -88, a reaction mixture containing 20 mM Tris/HCl, pH 8.0, 50 mM NaCl, 0.1 mg/ml BSA, 1 mM DTT, 3 M SETD3, 40 M SAM, and 10 M peptide was performed overnight at room temperature (Fig. 5B). Then reactions were terminated by the addition of TFA to 0.1% (v/v).

Isothermal titration calorimetry
ITC experiments (Fig. 5C) were performed using a MicroCal PEAQ-ITC automated system (Malvern Instrument Ltd) at 25°C. Purified SETD3 (20 M in 20 mM Tris, pH 8.0, 50 mM NaCl, 0.1 mg/ml BSA, 1 mM DTT) was maintained in the sample cell. The actin peptide Gln 73 (residues 66 -88, 400 M in the same buffer) were injected into the cell by a syringe under continuous stirring (750 rpm) with the reference power set as 8 cal/s. The volume of each injection was 2 l with a fixed duration time of 4 s and the spacing time between the injections was 250 s to achieve equilibrium. Binding constants were calculated by fitting the data using a single binding-site model by the ITC data analysis program supplied by the manufacturer.
After 20 min at room temperature, the reactions were terminated by the addition of TFA to 0.1% (v/v).

Data availability
The X-ray structures (coordinates and structure factor files) have been submitted to the PDB under accession numbers 6V62 (N255F/W273A-Lys 73 ) and 6V63 (SETD3-Gln 73 ).
Author contributions-S. D. data curation; S. D. and J. R. H. investigation; J. R. H. formal analysis; A. W. W., O. G., and X. Z. writingreview and editing; X. Z. and X. C. conceptualization; X. Z. and X. C. supervision; X. C. funding acquisition.