Structural insights into SETD3-mediated histidine methylation on β-actin

SETD3 is a member of the SET (Su(var)3–9, Enhancer of zeste, and Trithorax) domain protein superfamily and plays important roles in hypoxic pulmonary hypertension, muscle differentiation, and carcinogenesis. Previously, we identified SETD3 as the actin-specific methyltransferase that methylates the N3 of His73 on β-actin (Kwiatkowski et al., 2018). Here, we present two structures of S-adenosyl-L-homocysteine-bound SETD3 in complex with either an unmodified β-actin peptide or its His-methylated variant. Structural analyses, supported by biochemical experiments and enzyme activity assays, indicate that the recognition and methylation of β-actin by SETD3 are highly sequence specific, and that both SETD3 and β-actin adopt pronounced conformational changes upon binding to each other. In conclusion, this study is the first to show a catalytic mechanism of SETD3-mediated histidine methylation on β-actin, which not only throws light on the protein histidine methylation phenomenon but also facilitates the design of small molecule inhibitors of SETD3.


INTRODUCTION
Microfilaments, as the building blocks of cytoskeleton, are composed of actin proteins (dos Remedios et al., 2003;Theriot and Mitchison, 1991). Six different isoforms of actin are identified in mammals, which could be divided into three groups: alpha, beta and gamma (Gunning et al., 1983;Herman, 1993). Among the three types, β-actin is ubiquitously expressed and plays important roles in forming cytoskeleton, mediating cell motility and maintaining cell stability (Leterrier et al., 2017;Nudel et al., 1983).
Many different types of post-translational modifications (PTM) are found in actin, such as acetylation, methylation, SUMOylation, ubiquitination, etc. (Terman and Kashina, 2013). N 3 -methylation of His73 in β-actin is of special interest since its identification in 1967 (Johnson et al., 1967), not only because it is highly conserved in almost all eukaryotic actins except the actin from Naegleria gruberii (Sussman et al., 1984), but also because it affects hydrolysis rate of the actin-bound ATP through mediating the inter-domain mobility around the terminal phosphate of the ATP molecule (Kabsch et al., 1990), evidenced by the higher ATP exchange rate caused by the actin-H73A mutant (Nyman et al., 2002). Aside from actin proteins, N 3 -methylhistidine is also reported to exist in other proteins, such as myosin (Asatoor and Armstrong, 1967) and yeast Rpl3 (Webb et al., 2010). The effects of post translational modifications on proteins could be classified as follows: some modifications lead to a conformational change of the target molecule, while others may exclude existing binding partners and/or recruit new ligands (Bonasio et al., 2010). Methylation of the His73 residue of β-actin was reported to influence polymerization of the actin molecules and rate of ATP hydrolysis (Nyman et al., 2002;Terman and Kashina, 2013). Although some actin methyltransferases had been identified decades ago (Raghavan et al., 1992;Vijayasarathy and Rao, 1987), whether these enzymes specifically catalyze the methylation of His73 of β-actin, and the molecular mechanism of N 3 -methylation on His73, are largely unknown.
In all types of methylation, the molecular mechanism, as well as the biological functions, of histone lysine methylation has been well established because of its crucial roles in gene regulation and cell development (Martin and Zhang, 2005;Sims et al., 2003).
The SET domain is a ~130 aa motif and is initially identified in Trithorax, with its acronym derived from Su(var)3-9, Enhancer of zeste, and Trithorax (Dillon et al., 2005). It utilizes AdoMet as the co-factor by transferring the methyl moiety of AdoMet to substrates.
There are ~50 SET domain proteins in the human genome (Dillon et al., 2005) and whether all of them act as lysine methyltransferases, are largely unknown. SETD3 is a protein abundantly expressed in muscles and its knock down impairs differentiation of muscle cells (Eom et al., 2011). Previously, SETD3 has been identified as a histone methyltransferase that methylates H3K4 and H3K36 based on bioinformatics and peptide analyses (Eom et al., 2011;Wagner and Carpenter, 2012) and recently it is also found to be associated with lymphoma (Chen et al., 2013). After extensive screening on those putative methyltransferases, we found that SETD3 specifically recognized a His73containing fragment of β-actin, and catalyzed the N3-methylation of His73.
To understand how SETD3 methylates β-actin, we crystallized and solved two structures of AdoHcy-bound SETD3, one with the peptide containing unmodified His73 and the other with the same peptide containing methylated His73 (His73me). On the basis of our two β-actin bound SETD3 complex structures, we confirmed that SETD3 is a histidine methyltransferase and it specifically methylates His73 of β-actin.
By structural, mutagenesis and enzymatic analysis, we pinpointed some key active residues of SETD3 important for β-actin His73 methylation and substrate selectivity, and uncovered its methylation mechanism. This study would not only provide insights into the role of SETD3 in mediating the function of actins, but also guide the future design of small molecules towards SETD3.

SETD3 binds to and methylates β-actin
Since SETD3 has been identified as a histone lysine methyltransferase catalyzing methylation of H3K4 and H3K36, we purified the core region of SETD3 (aa 2-502) ( Figure 1A), and synthesized two histone H3 peptides, containing histone H3K4 and H3K36 residues, respectively. Surprisingly, our ITC binding results revealed that neither of two peptides displayed detectable binding to SETD3 (Supplementary table   S1). Furthermore, we also checked the methylation activity of SETD3 on H3K4 (1-10) and H3K36 (31-44) peptides by the mass spectrometry technique, and found that in contrast to β-actin, neither of these two peptides could be methylated by SETD3, at least under our experimental conditions (Supplementary Figure S1A-S1B). SETD3 has been shown to play an important role in muscle cell differentiation (Eom et al., 2011). Besides, the interactions between actin and myosin are also essential for smooth muscle cell differentiation (Skalli et al., 1986). Therefore, β-actin His73 methylation by SETD3, which are critical in mediating actin polymerization, directly link the methylation activity of SETD3 with actin functions. We set out to explore if βactin acts as the substrate of SETD3. Markedly, we found by ITC that SETD3 indeed displayed high binding affinity towards a His73-containing peptide of β-actin (aa 66-88) (Kd = 0.17 μM) ( Figure 1B), and our mass spectrometry based enzymatic assay shows that this peptide could be efficiently methylated (Supplementary Figure S1C).
The same β-actin peptide with His73 mutated to alanine (H73A) binds to SETD3 only slightly weaker (Kd = 0.45 μM)( Figure 1B), however, we failed to detect any methylated products from this mutant peptide (Supplementary Figure S1D), implicating that it is His73 of β-actin, which was specifically methylated by SETD3.

Overall structure of SETD3
To uncover the underlying recognition and methylation mechanisms of β-actin by SETD3, we tempted to crystalize SAM-bound SETD3 (2-502) with full length β-actin, but failed to get any diffractable crystals. Since SETD3 bound to β-actin peptide (66-88) and methylated His73 on it, we crystallized and determined a 1.95 Å structure of SAM-bound SETD3 (2-502) with β-actin (aa 66-88) ( Table 1). There are 4 molecules in a crystallographic asymmetric unit and residues 22-501 of SETD3 and 66-84 of βactin are visible in the complex structures (Supplementary Figure S2A). Markedly, based on the density map, we found that His73 of β-actin had been methylated in the complex structure (Supplementary Figure S3A), although we used an unmethylated peptide. The reason could be that SETD3 bound to the methyl donor AdoMet from the E. coli cells, which we used to express our protein, and the AdoMet-bound SETD3 methylated His73 during crystallization. Therefore, this complex represents the snapshot of post-methyl transfer state, which prompts us to crystallize the substrate bound enzyme next.
To avoid the methylation reaction, SETD3 was purified with adding 5-fold excess of AdoHcy in the buffer to compete off AdoMet. Then the unmethylated peptide was mixed with the purified SETD3 to form the complex. The complex was crystallized and solved at a resolution of 2.15 Å (Table 1). In this complex structure, the density map of the peptide indicates that it was unmodified, suggesting the methylation did not occur (Supplementary Figure S3B). The complex structure hereby reflects the image of the pre-methyl transfer state.

AdoHcy binding pocket within SETD3
In both complex structures, SETD3 binds AdoHcy in a cleft formed between SET and C-SET and buttressed by iSET at the bottom ( Figure 1C). Specifically, the adenine ring of AdoHcy is sandwiched between the side chain of Glu104 and the aromatic ring of Phe327, forming stacking and π-π interactions, respectively ( Figure Figure   S4B). Arg75 is a unique AdoHcy binding residue only found in SETD3 (Supplementary Figure S4B).

β-actin binding mode of SETD3
In both complexes, the β-actin peptide lies in a narrow groove formed by SET, iSET and C-SET, with the unmodified or methylated His73 residue accommodated in a pocket which is built by α11 of iSET, β6 and β12 of SET, and α12 of C-SET, and further buttressed by ƞ5 of SET ( Figures 1C and 3A). When we compared the ternary complex structure of SETD3 with AdoHcy and unmodified β-actin peptide with the previously determined AdoMet-SETD3 binary structure (PDB id: 3SMT), we found that, despite the overall similarity between these two structures (with a RMSD of 0.66 Å in Cα positions), the region containing two β-sheets of the SET domain (β4-β10-β9 and β3-β11) takes a conformational change upon peptide binding, due to its direct interactions with the N-terminal end of the peptide, with the three loops preceding β4, β9 and β11 shift 4.2 Å, 4.4 Å and 6.8 Å towards the peptide, respectively (Supplementary Figure   S6). The detailed interactions between SETD3 and the N-terminal end of the peptide will be described in more detail next.

A unique histidine recognition pocket in SETD3 confers SETD3 its histidine methyltransferase activity
In the substrate complex, residues Leu67-Glu83 of the peptide are visible, with His73 occupying in a hydrophobic pocket ( Figure 3A). The main chain of His73 forms several hydrogen bonds with SETD3, with its main chain amide, Cα and main chain carbonyl H-bonded to the main chain carbonyl of Tyr313, the side chain carboxyl of Asn256, and the guanidino of Arg316, respectively ( Figure 3A). The imidazole ring of His73 is parallel to the aromatic ring of Tyr313, with its orientation determined by two hydrogen bonds, one between the N 1 atom of the ring and the guanidino of Arg316 mediated by a water molecule, the other between the N 3 atom and the main chain carbonyl of Asp275 ( Figure 3A). In addition to the stacking interactions with Tyr313, His73 also makes hydrophobic contacts with Trp274 and Ile311 of SETD3. The distance between the N 3 atom of His73 and the sulphur of the AdoHcy is 3.7 Å, suggesting that it mimics the pre-methyl transfer state ( Figure 3A).
In the product complex, residues Leu67-Lys84 and the main chain of Ile85 of the peptide are visible, with His73 methylated to N 3 -methylated His73 (His73me) (Supplementary Figure S7). Although the peptide adopts a 3 10 helix at its C-terminal end, the overall structures of the two complexes are highly similar, with a RMSD of 0.19 Å over protein Cα atoms and a RMSD of 0.32 Å over peptide Cα atoms.
His73me is inserted into the same pocket of SETD3 as shown in the product complex, However, Superposition of the two complexes clearly indicated that while the main chain carbonyl of Tyr313 and the guanidino of Arg316 are still hydrogen bonded to the main chain of His73me, the imidazole ring of His73me rotates by ~90° comparing to that of the unmodified His73, with its C δ H pointing toward the ring of Tyr313 ( Figure   3B). Accordingly, Asn256 of SETD3 flips its side chain by 180° and makes one hydrogen bond with the N 1 atom of His73me imidazole ring, which determines the orientation of the imidazole ring of His73 and allows it to make stacking and hydrophobic interactions with Tyr313, and Trp274 and Ile311, respectively ( Figure 3B).
However, the flipping of the Asn256 side chain not only disrupts the hydrogen bond between Asn256 and the main chain Cα of His73me, but also makes the side chain amide groups of Asn256 close to the His73me Cα atom with a H-H distance of 1.7 Å (Supplementary Figures S8a-S8b), which would destabilize the enzyme-product complex and lead to subsequent release of the product.
In comparison with the other SET domain containing lysine methyltransferases, we found that there are two main reasons that SETD3 is not likely to be a lysine methyltransferase. First, the histidine binding pocket of SETD3 is too shallow to accommodate the aliphatic side chain of lysine. Replacement of His73 with a lysine, not only disrupts the water mediated hydrogen bond formed between N 1 of His73 and Arg316, but also leads to the steric clash between the N ε of the lysine and the main chain of the bottom carbonyl cage residue, Asp275. Second, phenylalanine of LSMT and SETD6 that makes cation-π interaction with the lysine, is replaced by asparagine (Asn256) in SETD3, and Asn256 is critical for substrate binding and histidine methylation ( Figure 3 and Supplementary Figure S4B). Therefore, the substitution establishes the imidazole ring-specific interaction with histidine by impairing the lysine specific interaction.

SETD3 exhibits extensive interactions with β-actin other than H73
Besides His73, the other residues of the β-actin peptide interacts extensively with the N-terminal lobe of SETD3, with Leu67-Glu72 and Gly74-Glu83 mainly contacting the SET and iSET motifs, respectively ( Figure S9A). The extensive interactions between the C-terminal residues of β-actin and SETD3 explains why deletion of the C-terminal residues of β-actin reduces its binding to SETD3 significantly (Kd = 2.9 μM, Figure 1B), suggesting that the C-terminal residues far from the target His73 residue also contribute to binding. Of note, the peptide sequence is also conserved in α-and γ-actins, suggesting that the other actins may be potential SETD3 substrates as well.

Key residues of SETD3 involved in β-actin binding and methylation
To evaluate the roles of the SETD3 residues around the active site in β-actin methylation, we made several single mutants and tested their binding affinities to the β-actin peptide. All SETD3 mutants displayed weaker binding affinities towards β-actin peptide (Supplementary table S1). Among these residues, the R215A and R316A mutants exhibited the most significantly reduced binding affinities by 21-and 42-folds, respectively. Our complex structures showed that both Arg215 and Arg316 make several hydrogen bonds with β-actin, implicating that they play a critical role in β-actin recognition and methylation (Supplementary table S1). N256A, which disrupts the hydrogen bond to the main chain of the His73 Cα atom, also reduces the β-actin binding affinity by 12-folds (Kd = 2.1µM) (Supplementary table S1). In the substrate complex, the imidazole ring of His73 stacks with the aromatic ring of Tyr313, with the N 3 atom of His73 in proximity to the hydroxyl group of Tyr313, which partially neutralizes the charge of N 3 and stabilizes the complex (Supplementary table S1). The favorable charge-charge interaction was also observed in the complex structures of protein with m 7 G base (Xu et al., 2016), verified by the fact that mutating Tyr313 to phenylalanine diminishes binding affinity by 17-fold (Kd = 3.0 µM) (Supplementary table S1).
Next we performed mutagenesis and ITC studies to corroborate the roles of β-actin residues in binding to SETD3. Since the β-actin His73 is mainly recognized by SETD3 via several main chain hydrogen bonds, it is not unexpected that H73A only slightly reduces the SETD3 binding by ~2-folds (Kd = 0.36 μM) (Supplementary Table S1).
Tyr69, Ile71 and Met82 of β-actin all contact SETD3 via hydrophobic interactions, with replacement by Ala dramatically reducing the binding affinities for SETD3 by 5-90 folds (Supplementary table S1). Consisting with the fact that both Asp80 and Asp81 of β-actin make hydrogen bonds with SETD3, D80A and D81A impair the binding to SETD3 by 20-and 11-folds, respectively (Supplementary table S1). In addition, truncation of Asp81-His88 of the peptide also reduces the SETD3 binding affinity by ~17 folds ( Figure 1B and Supplementary table S1). Taken together, flanking sequences of His73, rather than His73 itself on the β-actin peptide, play critical roles in binding to SETD3, suggesting that the substrate recognition by SETD3 is highly sequenceselective.
To gain further insight into how those mutants that weakened the substrate binding ability compromise catalysis ability, we used the mass spectrometry technique to compare the histidine methylation activity of SETD3 with its mutants in the presence of AdoMet. While we did detect the peak of methylated β-actin (66-88) when it was added with wild type SETD3 (2-502) and AdoMet, no such peak was detected for peptide only in the presence of AdoMet (Supplementary Figure S1C). All four SETD3 mutants (R215A, N256A, Y313F and R316A) displayed weaker histidine methylation activity towards the β-actin peptide since the peaks of products catalyzed by the SETD3 mutants were much lower than that observed for wild type SETD3 (Supplementary Figure S1E), owing to the weaker substrate binding affinities of the mutants (Supplementary table S1). Collectively, SETD3 mutants impairing the substrate binding also compromise subsequent catalysis.
To understand whether full-length β-actin is catalyzed in the same way as the β-actin fragment, we purified full-length recombinant β-actin (FL β-actin) in vitro and studied the enzyme activity of SETD3 (2-502) variants on β-actin protein. As expected, wild type SETD3 (2-502) displayed high activity towards FL β-actin (~1660.8 pmol/min/mg), and no activity towards the H73A mutant. The four SETD3 mutants that show reduced peptide binding affinities and peptide catalytic activity, also display much weaker activity towards the full length β-actin protein (164.4-245.3 pmol/min/mg) (Supplementary Figure S10). In addition, Arg75 and Asn278 of SETD3 are key AdoMet binding residues, with their replacement by alanine abolishing the catalytic activity (14.7-35.8 pmol/min/mg). Collectively, our data suggest that perturbation of βactin binding affinity significantly impacts on the SETD3-mediated methylation activity.

Catalytic mechanism of SETD3
Actins form microfilaments in vivo, which is one of the building blocks of cytoskeleton.
Different types of post translational modifications (PTM) were found in actins (Terman and Kashina, 2013). N 3 -methylhistidine was identified in β-actin decades ago (Johnson et al., 1967) and was reported to mediate the polymerization and hydrolysis of actin (Kabsch et al., 1990). However, how the mark is incorporated, remains enigmatic. Our identification of SETD3 as a histidine methyltransferase catalyzing the His73 methylation of β-actin, very well answers both of the unsolved questions.
Although SETD3 contains canonical SET domain and its AdoHcy binding mode is similar to those observed in other SET domain lysine methyltransferases. The predicted SET domain fold of SETD3 could not help to uncover the molecular mechanisms of the substrate recognition or histidine methylation, since SET domain itself, as a short motif of ~110 aa, always need iSET and C-SET domains to complete its function (Chang et al., 2011;Trievel et al., 2002). Therefore, two complex structures presented here not only for the first time uncover the molecular mechanism by which a SET domain protein acts as the N 3 -histidine methyltransferase, as revealed by the histidine specific binding pocket in both complexes, but also provide insights into the pre-and post-methyl transfer states of the histidine methylation by the same enzyme.
Our structural analysis, assisted by mutagenesis and biochemical experiments, suggested that the recognition of β-actin and subsequent catalysis by SETD3 is different from those observed from other SET domain lysine methyltransferases in following aspects: i) histidine recognition pocket of SETD3 is shallower and key residues of SET3, which form hydrogen bonds with the His73 imidazole ring, are not conserved in other SET domain members ( Figure 5A-5B and Supplementary Figure S4B). ii) In two solved SETD3 complex structures, the methylated histidine rotates its side chain by 90° to enable the release of the product (Figure 3), which is not reported in lysine methyltransferase complexes before. (iii) The recognition of β-actin by SETD3 is highly dependent on flanking sequences of His73 (Figure 4), consisting with the high binding affinity between H73A and SETD3 ( Figure 1B). Although N 3 -methylhistidine is also found in other proteins, such as mammalian myosin (Johnson et al., 1967) and Saccharomyces cerevisiae Rpl3 (Webb et al., 2010), the sequence preference suggests that SETD3 probably only work on His73 of β-actin, as well as on corresponding histidines in α-and γ-actins. One implication is that no SETD3 ortholog is found in baker's yeast.
Recently, the structures of CARNMT1 bound to analogs of N 1 -methylhistidine was also reported (Cao et al., 2018), which prompts us to compare the N 3 -histidine methylation by SETD3 with the N 1 -histidine methylation by CARNMT1 through comparing two complexes ( Figures 5A and 5C). We found the two histidine methyltransferases adopt distinct folds, which lead to different AdoHcy binding modes. Furthermore, SETD3 and CARNMT1 display different histidine methylation mechanism. While CARNMT1 utilizes Asp316 to form two hydrogen bonds with the imidazole ring of the histidine to facilitate the deprotonation of the histidine imidazole ring ( Figure 5C), while in the SETD3-substrate complex, histidine deprotonation is enabled by the Tyr313 acting as the general base and by Asp275 forming main chain hydrogen bond with the histidine imidazole ring (Figures 5A and 5D). The important role of Tyr313 in both binding and catalysis is further corroborated by the mass spectrum analysis that mutating Tyr313 to Phe markedly decreased the β-actin binding affinity, as well as the methylation activity (Supplementary Table S1 and Supplementary Figure S1E). Last but not least, unexpected 90-degree rotation of imidazole ring of His73me does not occur in CARNMT1-mediated histidine methylation.

Implications in conformational changes of actin upon binding to SETD3
In natively purified rabbit β-actin, His73 exists as the methylated form and it is localized close to the phosphate groups of ATP binding to actin (Kabsch et al., 1990). Although our complexes only contain a fragment of β-actin, they could still provide clues on the role of SETD3 in mediating actin functions. The residues 66-84 of β-actin consist of three β strands followed by an α-helice (Supplementary Figure S11), however, the βactin peptide adopts an extended conformation upon binding to SETD3 (Supplementary Figure S11), suggesting that β-actin would probably endure local structural remodeling during methylation, such as beta-sheet decomposition. To complete the methylation, remodeling of β-actin should occur at the very beginning to expose the partially buried His73 and make it accessible to SETD3, which is an energy-consumption step.
Our ITC binding data indicate that SETD3 binds to β-actin peptide with extreme high binding affinity (Kd = 170 nM) and the binding would provide energy to decompose the local secondary structures of β-actin. It is not unlikely that His73 is directly recognized by SETD3 as the first step considering the β-actin binding by SETD3 largely depends on the flanking sequences of His73. Of note, Leu67-Tyr69 of β-actin is exposed to the solvent, while Tyr69 of the peptide makes extensive hydrophobic contacts with SETD3, hence the decomposition and binding of β-actin to SETD3 might firstly occur at the N-terminal side of His73, which may help to "pull out" His73 to make it exposed to SETD3. However, we could not exclude the possibility that some factors, such as the CCT complex, may bind actin to facilitate the process (Martin-Benito et al., 2002). The structures of SETD3 with native actin may be required to provide more details on the dynamics of actin.
Since SETD3 is reported to be associated with cancer malignancy because of its higher level of expression in liver tumor (Cheng et al., 2017), uncovering the role of SETD3 in methylating actin may pave the way for understanding its role in carcinogenesis through influencing the functions of actin and cytoskeleton. Thus our results not only will facilitate the understanding the roles of actin methylation catalyzed by SETD3 in human cancer, but also would provide structural basis for designing SETD3 inhibitors in near future.

Cloning, protein expression and purification of SETD3
Gene encoding the core region of SETD3 (2-502) were synthesized by Sangon Biotech The SETD3 mutants were constructed by conventional PCR using the MutanBEST kit (TaKaRa) and further verified by DNA sequencing. The SETD3 mutants were expressed and purified in the same way as the wild type protein.

Protein
Plasmid pCOLD I encoding human β-actin (ACTB, GenBank: NM_001101.4) was a kind gift of Dr. Minoru Tamura (Ehime University, Japan) and was prepared as described (Tamura, 2018). For β-actin production, Escherichia coli BL21(DE3) (Agilent, USA) cells were transformed with the DNA construct and a single colony was selected to start an over-nigh pre-culture. 500 mL of LB broth (with 100 μg/mL ampicilin) was inoculated with 50 ml of the pre-culture and incubated at 37 °C and 200 rpm until an OD600 of 0.5 was reached. The culture was placed on ice for 20 min.  Fig. 1). The β-actin preparation was immediately desalted on PD-10 columns equilibrated with 20 mM Tris-HCl, pH 7.5, 1 mM DTT, 6% sucrose, 2 μg/ml leupeptin and 2 μg/ml antipain. The purified β-actin was stored at -70°C.

Isothermal titration calorimetry (ITC)
All peptides were synthesized by GL Biochem (Shanghai) Ltd. and were dissolved in water as a stock of 5-12 mM, with the pH of stock solutions adjusted to pH 7.5. Peptides

Crystallization, data collection and structure determination
All crystals were grown by using sitting drop vapor diffusion method at 18°C. For crystallization of SETD3 with methylated peptide, SETD3 (12 mg/ml) was preincubated with synthesized β-actin peptide (66-88) (GL Biochem Ltd.) and AdoMet at a molar ratio of 1:3:4, and mixed with the crystallization buffer containing 0.1 M Sodium cacodylate trihydrate pH 6.5, 0.2M Magnesium acetate tetrahydrate, and 20% v/v Polyethylene glycol 8000. For the crystallization of SETD3 with unmodified peptide, SETD3, in the concentration of 12 mg/ml, was pre-incubated with synthesized β-actin peptide (66-88) and AdoHcy at a molar ratio of 1:3:4, and mixed with the crystallization buffer containing 0.1M HEPES sodium pH 7.5, 2% v/v Polyethylene glycol 400, and 2.0 M Ammonium sulfate. Before flash-freezing crystals in liquid nitrogen, all crystals were soaked in a cryo-protectant consisting of 90% reservoir solution plus 10% glycerol. The diffraction data were collected on BL17U1 at the Shanghai Synchrotron Facility (28) at Shanghai Synchrotron Facility (SSRF). Data sets were collected at 0.9789 Å or 0.9789 Å, and were processed by using the HKL2000 program (29).
The initial structures of the SETD3-actin complexes were solved by molecular replacement in PHENIX (30) with previously solved SETD3 structure (PDB: 3SMT) as the search model. Then all the models were refined manually and built with Coot (31). The final structures were further refined by PHENIX (30). The statistics for data collection and refinement are summarized in Table 1.

Mass Spectrometry
Reversed-phase microcapillary/tandem mass spectrometry (LC/MS/MS) was performed using an Easy-nLC nanoflow HPLC (Proxeon Biosciences) with a selfpacked 75 μm id x 15 cm C18 column connected to either an QE-Plus (Thermo Scientific) in the data-dependent acquisition and positive ion mode at 300 nL/min.
Passing MS/MS spectra were manually inspected to be sure that all b-and y-fragment ions aligned with the assigned sequence and modification sites. A 25-ul reaction mixture containing 2 µM SETD3 or SETD3 mutants (final concentration), 20 µM peptide (final concentration) in a buffer containing 10mM Tris-HCl, pH 7.5, 20mM NaCl and 10µM AdoMet. The reaction was incubated at 37℃ for 2h before being quenched (at 70 ℃ for 10-15 mins). Then quenched reactions were analyzed by LC/MS/MS and Proteomics Browser Software, with the relative abundances of substrate and product reflecting the methylation activities of proteins.

Assay of the Actin-specific Histidine N-methyltransferases Activity
The enzyme activity was determined by measuring the incorporation of [ 3 H]methyl group from S-[methyl-3H]adenosyl-L-methionine ([ 3 H]SAM) into homogenous recombinant human (mammalian) β-actin or its mutated form in which histidine 73 was replaced by alanine residue (H73A). The standard incubation mixture (0.11 ml) contained 25 mM Tris-HCl, pH 7.2, 10 mM KCl, 1 mM DTT, 2 µM protein substrate and 1 μM [ 1 H+ 3 H] SAM ( 400 × 103 cpm). When appropriate, the incubation mixture was supplemented with recombinant S-adenosyl-L-homocysteine (SAH) nucleosidase and adenine deaminase as indicated in legends to figures and tables. The reaction was started by the addition of enzyme preparation and carried out at 37°C for 15 min unless otherwise described. Protein methylation was linear for at least 15 min under all conditions studied. By analogy to assays of nonribosomal peptide synthetase activity [Richardt et al., 2003;Drozak et al., 2014], the incubation was stopped by the addition of 0.1 ml of the reaction mixture to 0.025 ml of BSA (1 mg/ml) and 0.8 ml of ice-cold 10 % (w/v) trichloroacetic acid (TCA). After 10 min on ice, the precipitate was pelleted and washed twice with ice-cold 10 % TCA. The pellet was finally dissolved in pure formic acid.

ACCESSION NUMBERS
The coordinates and structure factors of AdoHcy bound SETD3 with unmodified βactin peptide and methylated β-actin peptide haven been deposited into Protein Data Bank (PDB) with accession numbers 6ICV and 6ICT, respectively.

SUPPLEMENTARY INFORMATION
Supplemental Information can be found online at XXXX

ACKNOWLEDGEMENT
We are grateful to the staff members at beam lines BL17U1, BL18U1 and BL19U1 at Shanghai Synchrotron Radiation Facility for assistance in data collection. This work was supported by National Natural Science Foundation of China Grants 31570737 and 31770806 (to C. X.) and 31500601 (to S. L.) and the "Thousand Young Talent program" (to C. X.). The Structural Genomics Consortium is a registered charity (no. 1097737) that receives funds from AbbVie; Bayer Pharma AG; BoehringerIngelheim; Canada            Figure 1A. The residues involved in binding to AdoHcy and His73 are labeled in cyan and yellow, respectively. (B) Sequence aligment of human SETD3, human SETD6 (NP_001153777.1) and Pisum sativum LSMT (AAA69903.1). The secondary structures, AdoHcy binding residues, and His73 binding residues, are labeled in the same way as shown in Supplementary Figure S4A  with AdoHcy bound SETD3-actin. In the structure of AdoMet bound SETD3, the protein is shown in grey ribbon and AdoMet is shown in grey sticks. In the structure of AdeHcy bound SETD3 in the presence of actin, protein is shown in blue ribbon and AdoHcy is shown in yellow sticks. β4-β9-β10 and β3-β11 changed conformations significantly upon binding to actin peptide, which are shown in cartoon representation. Leu67-Ile71 of β-actin and their interaction residues in SETD3, are shown in sticks. Three loops that precede β4, β9 and β11 shift 4.2Å, 4.4Å, and 6.8Å, respectively.  Asn256 of SETD3 forms one main chain hydrogen bond with Cα of β-actin His73 via its carboxyl group. (b) Asn256 flips side chain by 180 degree, making one side chain hydrogen bond with N 1 atom of His73me via its side chain amide. The distance between the side chain NH of Asn256 and the Hα of His73me, is only 1.7 Å, which would destabilize the complex.
Supplementary Figure S9 A B Supplementary Figure S9. Schematic of the detailed interactions (A) between SETD3 and unmodified β-actin (B) and between SETD3 and methylated β-actin. The SETD3 residues are colored as shown in Figure 1A, while the peptide residues are colored in yellow. The hydrophobic interactions and hydrogen bonds between protein and peptide are shown in black solid arrows and black dash arrows, respectively.
Supplemental Figure S10. SETD3 (2-502) displayed high enzyme activity towards FL wild type β-actin, but not the H73A mutant. The mutants exhibiting weaker peptide binding affinies also displayed weaker activity towards β-actin protein.  Figure S11. Actin endures conformational changes upon binding to SETD3. β-actin peptide (66-88, yellow cartoon) is over laid with the same fragment in the structure of native β-actin (red cartoon) (PDB id: 1HLU). native β-actin decomposes its local secondary structures upon SETD3 binding. Native β-actin and SETD3 are shown in red and blue ribbon, respectively.   Supplementary Table S1.