We and others had previously shown that SETD3 acts as an actin-specific histidine N-methyltransferase, but the molecular basis for the selective histidine methylation catalyzed by SETD3 remained unknown (Kwiatkowski et al., 2018; Wilkinson et al., 2019). We measured the binding affinity of the SETD3 core region (aa 2–502) to a His73-containing fragment of β-actin (aa 66–88) by ITC and found that SETD3 binds to the β-actin peptide with a Kd of 0.17 μM (Figure 1A–B and Table 1). Given that SETD3 has also been reported to be a putative lysine methyltransferase that methylates Lys4 and Lys36 of histone H3 (Eom et al., 2011), we also examined the binding of SETD3 to two different histone peptides, H3K4(1–23) and H3K36(25–47), and found that neither of these peptides binds to SETD3 (Table 1).

Figure 1 with 8 supplements see all

Download asset Open asset

Furthermore, we detected the activity of SETD3 on β-actin(66–88), H3K4(1–23), and H3K36(25–47) by mass spectrometry and found that SETD3 methylates the β-actin peptide (Figure 1—figure supplement 1A) but does not modify either H3K4 or H3K36 (Figure 1—figure supplement 1B–1C). No methylated product was detected for any of the three peptides in the presence of AdoMet without the addition of SETD3 (Figure 1—figure supplement 1A–1C). Moreover, SETD3 did not methylate the β-actin(66–88) H73A mutant, although binding of the mutant peptide to SETD3 is only ~2.5 fold weaker than that of the wild-type peptide (0.45 μM vs. 0.17 μM) (Figure 1B and Figure 1—figure supplement 1D). Collectively, the binding experiments and mass spectrometry data indicate that SETD3 specifically binds to the β-actin peptide and methylates it at His73.

To uncover the mechanisms underlying the recognition and methylation of β-actin by SETD3, we attempted to crystallize AdoMet-bound SETD3 with full-length β-actin, but failed to obtain diffractable crystals. By using the core region of SETD3(2–502) with the β-actin peptide(66–88), we succeeded in crystallizing the complex and obtained a 1.95 Å structure of AdoHcy-bound SETD3(2–502) with β-actin(66–88) (Table 2, Figure 1—figure supplement 2A, Figure 1—figure supplement 3). There are four molecules in a crystallographic asymmetric unit and residues 22–501 of SETD3 and 66–84 of β-actin are visible in the complex structures (Figure 1—figure supplement 4A). On the basis of the density map, we found that His73 of β-actin had been methylated in the structure (Figure 1—figure supplement 5A), although we used an unmethylated peptide. The reason for this could be that SETD3 bound to the methyl donor AdoMet from Escherichia coli, which we used to express our protein, and that the AdoMet-bound SETD3 methylated the peptide at His73 during crystallization. Therefore, this complex represents a snapshot of the post- methyl transfer state, which prompted us to crystallize the substrate-bound enzyme.

To avoid the methylation reaction, SETD3 was purified by including a 5-fold excess of AdoHcy in the buffer that would compete with AdoMet for binding with SETD3. Then, the non-methylated 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 2). There are two molecules in a crystallographic asymmetric unit (Figure 1—figure supplement 4B). In this structure, the density map of the peptide indicates that it is unmodified, suggesting that methylation did not occur (Figure 1—figure supplement 5B). The complex structure therefore reflects the image of the pre-methyl transfer state.

In both substrate-bound and product-bound structures, the SETD3 core region adopts a V-shape architecture and is composed of N-SET (N-terminal to the SET domain), SET, iSET (an insertion in the SET domain), and C-SET (C-terminal to the SET domain) domains (Figure 1C). The N-terminal lobe contains the N-SET (α1–α3), SET (β1–β12), and iSET (α4–α11) domains (Figure 1—figure supplement 2A, Figure 1—figure supplement 3). The N-SET domain consists of three α helices (α1–α3), and encompasses the SET and iSET domains (Figure 1—figure supplement 2A, Figure 1—figure supplement 3). α1 and α2 of N-SET form a four-helix bundle with α10–α11 of iSET, whereas α3 of N-SET is localized at the SET-iSET interface and comes into contact with the anti-parallel β strands (β1–β2) of SET and α8 of iSET (Figure 1C, Figure 1—figure supplement 3). The SET domain adopts a canonical SET domain fold similar to that of LSMT (Trievel et al., 2002), with twelve β strands arranged into four anti-parallel β sheets (β1-β2, β3-β11, β4-β10-β9, β5-β7-β6) and one parallel β sheet (β8-β12) (Figure 1—figure supplement 2A, Figure 1—figure supplement 3). The iSET domain is a helical region (α4–α11) inserted between β5 and β6 of the SET domain (Figure 1—figure supplement 2A, Figure 1—figure supplement 3). The C-terminal lobe, consisting of just the C-SET domain, is almost entirely helical (α12–α19) except for two anti-parallel strands (β13–β14), with α12 and the N-terminal end of α19 packed against α10 and α9 of iSET, respectively (Figure 1—figure supplement 2A, Figure 1—figure supplement 3). Consistently, a database search using the DALI server (Holm and Rosenström, 2010) revealed that the overall fold of SETD3 is highly similar to those of LSMT (Z-core: 31.4, RMSD: 3.8 Å for 425 Cα atoms) and SETD6 (Z-core: 28.7, RMSD: 2.8 Å for 421 Cα atoms), despite the low sequence identities (24–25%).

In both substrate-bound and product-bound structures, AdoHcy is located at a cleft in SETD3 formed by the SET and C-SETD domains, and is 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 with Glu104 and Phe327, respectively (Figure 1—figure supplement 6A). The N⁶ and N⁷ atoms of AdoHcy are hydrogen-bonded to the main chain carbonyl and amide groups of His279, respectively, while the AdoHcy C⁸ atom forms one C-H...O hydrogen bond with the hydroxyl group of Tyr313 (Figure 1—figure supplement 6A).

In addition to the adenine-specific interactions, the ribose O3′ and O4′ atoms of AdoHcy form hydrogen bonds with the main chain carbonyl of Ser325 and the side chain carboxyl of Asn278, respectively (Figure 1—figure supplement 6A). The side chain carboxyl of Asn278 makes another hydrogen bond with the AdoHcy amide, and the latter is also H-bonded to the main chain carbonyl of Phe106 (Figure 1—figure supplement 6A). The AdoHcy carboxylate group forms four hydrogen bonds with SETD3 residues: one with the main chain amide of Phe106, one with the guanidino group of Arg254, and the other two with the guanidino group of Arg75 (Figure 1—figure supplement 6A). All AdoHcy-binding residues are conserved in other SETD3 orthologs, suggesting a conserved function of SETD3 (Figure 1—figure supplement 2A).

The AdoHcy-binding mode of SETD3 is similar to those observed in some other SET domain structures, such as those of LSMT (Trievel et al., 2002) and SETD6 (Chang et al., 2011) (Figure 1—figure supplement 6B–6C). Among these AdoHcy-binding residues, Asn278, His279 and Phe327 are conserved in LSMT and SETD6, whereas Glu104 and Arg254 are conserved only in LSMT and not in SETD6 (Figure 1—figure supplement 2B, Figure 1—figure supplement 6B–6C). Arg75 is an AdoHcy-binding residue that is found only in SETD3 (Figure 1—figure supplement 2B).

In both substrate-bound and product-bound structures, the β-actin peptide lies in a narrow groove formed by SET, iSET and C-SET, with the unmodified or methylated His73 residue snugly accommodated in a pocket composed of α11 of iSET, β6 and β12 of SET, and α12 of C-SET, and η5 of SET (Figure 1C, Figure 2A). When superimposing our substrate-bound complex structure with the previously determined AdoMet–SETD3 binary structure (PDB id: 3SMT), we found that, despite the overall structural similarity between the two SETD3 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) adopts a conformational change upon peptide binding. Specifically, peptide binding induces the three loops preceding β4, β9 and β11 to shift toward the peptide by 4.2 Å, 4.4 Å and 6.8 Å, respectively (Figure 1—figure supplement 7). The detailed interactions between SETD3 and the N-terminal end of the peptide will be described in the following section.

Figure 2 with 3 supplements see all

Download asset Open asset

In the substrate-bound structure, residues Leu67-Glu83 of the β-actin peptide are visible, with His73 occupying a hydrophobic pocket (Figure 2A). 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 2A). The imidazole ring of His73 is parallel to the aromatic ring of Tyr313, with its orientation determined by two hydrogen bonds: one between the N¹ atom of the ring and the guanidino of Arg316 mediated by a water molecule, and the other between the N³ atom of the ring and the main chain carbonyl of Asp275 (Figure 2A). In addition to the stacking interactions with Tyr313, His73 also makes hydrophobic contacts with Trp274 and Ile311. The distance between the N³ atom of His73 and the sulfur atom of AdoHcy is 3.7 Å, suggesting that this structure represents the pre-methyl transfer state well (Figure 2A).

In the product complex, the residues Leu67-Lys84 and the main chain of Ile85 of the β-actin peptide are visible, with His73 methylated to His73me (Figure 1—figure supplement 8). Despite the fact that the methylated peptide adopts a 3¹⁰ helix at its C-terminal end, the substrate-bound and product complexes are highly similar, with an RMSD of 0.19 Å over protein Cα atoms and an RMSD of 0.32 Å over peptide Cα atoms, respectively. In the product-bound structure, His73me is inserted into the same pocket of SETD3 as in the substrate complex but the distance between the installed methyl group of His73 and the sulfur atom of AdoHcy is 3.2 Å, with its density map distinguishable from that of unmodified histidine (Figure 2B).

Superposition of the substrate-bound and product complexes clearly indicates that the hydrogen bonds between the main chain of His73me and the main chain carbonyl of Tyr313 and the guanidino of Arg316 are intact after methylation (Figure 2B). However, the imidazole ring of His73me is rotated by ~90° relative to that of the unmodified His73, with its C⁴H pointing toward the ring of Tyr313 (Figure 2B–2C) and its N¹ atom hydrogen-bonded to the side chain carboxyl group of Asn256 (Figure 2B).

When comparing SETD3 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 a lysine. Replacement of His73 with a lysine would not only disrupt the water-mediated hydrogen bond between N¹ of His73 and Arg316, but could also lead to a steric clash between the aliphatic side chain of lysine and the main chain of the bottom carbonyl cage residue, Asp275. Second, the phenylalanine of LSMT and SETD6, which creates a cation-π interaction with the lysine, is replaced by an asparagine (Asn256) in SETD3, and Asn256 is critical for substrate binding and histidine methylation (Figure 2A–2B and Figure 1—figure supplement 2A). Therefore, this substitution establishes an imidazole ring-specific interaction with histidine by impairing the lysine-specific interaction.

In both substrate-bound and product complexes, peptide residues flanking His73 or His73me of β-actin also interact extensively with the N-terminal lobe of SETD3, with Leu67-Glu72 and Gly74-Glu83 mainly contacting the SET and iSET motifs, respectively (Figure 2—figure supplement 1, Figure 2—figure supplement 2). Leu67 and Pro70 of β-actin make hydrophobic contacts with Ile284 of SETD3 and there are additional hydrophobic interactions between Tyr69 of β-actin and Pro259, Tyr288 and Leu290 of SETD3 (Figure 2—figure supplement 1, Figure 2—figure supplement 2). Two main chain hydrogen bonds are formed between Tyr69 of β-actin and Tyr288 of SETD3, and between Ile71 of β-actin and Thr286 of SETD3, inducing the shift of the two sheets (β4-β10-β9 and β3-β11) in SETD3, as mentioned above (Figure 1—figure supplement 7). Ile71 of actin is positioned into a hydrophobic pocket of SETD3 that is formed by Ile258, Ile271, Trp274, Thr286, Tyr288, and Cys295 (Figure 2—figure supplement 1, Figure 2—figure supplement 2). Notably, Leu283-Thr285 of SETD3 adopts a β strand in the structure of AdoMet-bound SETD3 without a peptide (PDB: 3SMT), but decomposes upon binding to the N-terminal side of β-actin, which shortens the β9 strand of SETD3 in the peptide-bound structure (Figure 1—figure supplement 7). The side chain carboxylate group of Glu72 forms one hydrogen bond with the guanidino group of Arg316, which is the only C-SET domain residue that contacts the N-terminal side of the β-actin peptide (Figure 2—figure supplement 1, Figure 2—figure supplement 2).

Regarding the residues of the C-terminal to His73 of β-actin, the main chain amide group of Gly74 is hydrogen-bonded to the main chain carbonyl group of Gln255, while the side chain of Val76 stacks with the imidazole ring of His324 (Figure 2—figure supplement 1, Figure 2—figure supplement 2). In contrast to Gly74 and Val76, the Ile75 side chain is solvent-exposed and is therefore not involved in the interactions with SETD3 (Figure 2—figure supplement 1). Thr77 not only makes hydrophobic contact with Leu268, but also forms two hydrogen bonds through its hydroxyl group with the side chain amide groups of Asn154 and Gln255 (Figure 2—figure supplement 1, Figure 2—figure supplement 2). The side chain amide group of Asn154 makes another hydrogen bond with the main chain carbonyl of Asn78. The indole ring of Trp79 is inserted into an open hydrophobic pocket composed of Ile155, Val248, Val251 and Met252, with its main chain carbonyl H-bonded to the side chain amide of Gln216, while the carboxyl group of Asp80 is hydrogen-bonded to the side chain amide of Arg215 (Figure 2—figure supplement 1, Figure 2—figure supplement 2). Asp81 of β-actin forms three hydrogen bonds with SETD3: one with the side chain amide of Gln216 and two with the guanidino group of Arg215 (Figure 2—figure supplement 1, Figure 2—figure supplement 2). Notably, the β-actin peptide sequence is also conserved in α-and γ-actin proteins, suggesting that SETD3 might also recognize and methylate other actin isotypes.

To evaluate the roles of the SETD3 residues around the active site in β-actin methylation, we made several single SETD3 mutants and quantitatively compared their affinities for binding to the β-actin peptide by ITC. All of the SETD3 mutants displayed weaker binding affinities toward the peptide, but R215A and R316A exhibited the most significantly reduced binding affinities: 21-and 42-fold, respectively (Table 1). Our complex structures showed that both Arg215 and Arg316 make several hydrogen bonds with β-actin (Figure 2B, Figure 2—figure supplement 1), implying that they play a critical role in substrate recognition (Table 1). N256A binds the peptide with a 12-fold lower affinity than wildtype SETD3, (Kd = 2.1 µM) (Table 1); this is because the mutation disrupts not only the hydrogen bond to the main chain of the His73 Cα atom (Figure 2A) but also the one to the N¹ atom of His73me (Figure 2B). In the substrate-bound structure, the imidazole ring of His73 stacks with the aromatic ring of Tyr313, with the N³ atom of His73 in proximity to the hydroxyl group of Tyr313 (Figure 2A). The hydroxyl group of Tyr313 partially neutralizes the charge of N³, as is verified by the fact that mutating Tyr313 to phenylalanine diminishes binding affinity 17-fold (Kd = 3.0 µM) (Table 1). A similar favorable charge–charge interaction has also been observed in the structure of Gemin5 bound to m⁷G base (Xu et al., 2016).

Given that the β-actin peptide interacts with SETD3 through the sequences that flank His73, we performed β-actin mutagenesis and ITC studies to corroborate the roles of Tyr69, Ile71, Asp80, Asp81, and Met82. The Tyr69, Ile71 and Met82 of β-actin are involved in binding to SETD3 via hydrophobic interactions, and the replacement of any of them with alanine leads to a 5–90 fold reduction in the SETD3 binding affinity (Table 1). Consistent with the fact that Asp80 and Asp81 of β-actin both form hydrogen bonds with SETD3, D80A and D81A cause a 20- and 11-fold decrease in SETD3 binding affinity, respectively (Table 1). The interactions between Asp81 and Met82 of β-actin and SETD3 also explain why truncation of the Asp81–His88 of the peptide reduces the binding to SETD3 ~17 fold (Figure 2—figure supplement 1, Table 1). Taken together, the flanking sequences of His73 on the β-actin peptide, rather than His73 itself, are critical for binding to SETD3, suggesting that the recognition of substrate by SETD3 is highly sequence-selective.

To obtain further insights into how substrate recognition by SETD3 affects subsequent histidine methylation, we used mass spectrometry to compare the activity of the above-mentioned SETD3 mutants with that of the wildtype protein in the presence of AdoMet. The mass spectrometry data indicate that the peaks of products generated by the four SETD3 mutants (R215A, N256A, Y313F, and R316A) were much lower than that observed for wildtype SETD3 (Figure 1—figure supplement 1E), suggesting that these mutants displayed weaker histidine methylation activity.

To verify the identification of the key catalytic residues of SETD3, and to analyze the roles of substrate recognition residues in catalysis, we determined the kinetic properties of wildtype and mutant variants of purified recombinant SETD3 in the presence of homogenous full-length recombinant human β-actin. All enzyme assays were performed in the presence of E. coli AdoHcy nucleosidase and Bacillus subtilis adenine deaminase to prevent the accumulation of AdoHcy in the reaction mixture (Dorgan et al., 2006). As shown in Table 3, the wildtype SETD3 has a very high affinity toward both AdoMet (K_M ≈ 0.1 µM) and β-actin (K_M ≈ 0.5 µM), although it appeared to be a sluggish catalyst, with k_cat equal to about 0.7–0.8 min⁻¹. All mutant enzymes exhibited a reduced catalytic efficiency as indicated by k_cat/K_M ratios, which were 2–400-fold and 1.6–2000-fold for β-actin and AdoMet, respectively (Table 1), confirming the importance of the examined amino-acid residues for efficient catalysis.

Arg75 and Asn278 are both AdoMet-binding residues and the mutation of either of these to alanine severely impaired the k_cat of the enzyme toward β-actin and AdoMet (c.f. Table 3). Interestingly, although N256A and Y313F weakened binding to the β-actin peptide (by ~12 and~17 fold, respectively (Table 1)), both mutants displayed K_M values (0.449 µM and 0.581 µM, respectively) comparable to that of wildtype SETD3 (0.502 µM), probably because their k_cat values are much lower (~26- and ~36-fold, respectively) than that of wildtype SETD3 (c.f. Table 3). For AdoMet, N256A and Y313F both have a higher K_M (~6- and ~58-fold, respectively) and lower k_cat (~9- and ~6-fold, respectively) than the wildtype enzyme. We also examined the kinetic properties of N256D and N256Q, and found that, for both β-actin and AdoMet, the kinetic parameters of N256Q are comparable to those of wildtype SETD3, while N256D behaves similarly to N256A (Table 3). Taking into account the fact that Asn256 and Tyr313 of SETD3 interact mainly with His73 of β-actin, and not with AdoMet (Figure 2A–2B), these results may imply that Asn256 and Tyr313 of SETD3 plausibly bind to His73 of β-actin by favorably orienting its imidazole ring to facilitate the efficient transfer of the methyl group from AdoMet to His73. The consequence of the N256A and Y313F mutations is thus a decrease in the rate of ‘successful’ methylation events, which could be partially overcome by an increase in AdoMet concentration.

In contrast to Asn256 and Tyr313, the Arg215 and Arg316 residues of SETD3 contact with the flanking sequences of His73. R215A and R316A, which impaired the binding of SETD3 to the β-actin peptide, also displayed higher K_M values (~2-fold for both) and lower k_cat values (~3.5- and~4.5-fold, respectively) for the β-actin protein (Table 3), confirming that the SETD3 residues that are involved in β-actin peptide recognition are also important for efficient His73 methylation in the context of the full-length β-actin.

We further investigated the pH-dependent activity of SETD3 toward full-length β-actin and found that SETD3 catalyzes the methylation of β-actin at a wide range of pH (6.0–9.0) environments, with its activity increasing with pH and reaching its maximum at pH 9.0 (Figure 3). The detectable activity of SETD3 at neutral pH is consistent with the fact that histidine deprotonation occurs at neutral pH, which facilitates the transfer of the methyl group from AdoMet to histidine. The increasing rate of the reaction with increasing pH to pH 9.0 could suggest, however, that the imidazolium pKa is perturbed (elevated) in the active site due to the active site environment or that proton transfer from the imidazole/imidazolium may not contribute to the rate-limiting step.

Figure 3

Download asset Open asset

Figure 3—source data 1

Radiochemical measurements of SETD3-dependent methylation of either human recombinant β-actin or its mutated form (H73A) in the presence of increasing pH values of the reaction mixture.

https://doi.org/10.7554/eLife.43676.021

Download elife-43676-fig3-data1-v3.xlsx

As the major component of microfilaments, actin participates in diverse cellular processes. Different types of PTM have been identified in actin, and these modifications play important roles in mediating actin functions in vivo (Terman and Kashina, 2013). N³-methylhistidine was identified in β-actin decades ago (Johnson et al., 1967) and was postulated to mediate the polymerization and hydrolysis of actin filaments (Kabsch et al., 1990), but how the histidine methylation is incorporated remains enigmatic. Our identification of SETD3 as a histidine methyltransferase that catalyzes the His73 methylation of β-actin answers this unsolved question (Kwiatkowski et al., 2018).

Although SETD3 contains a canonical SET domain and its AdoHcy-binding mode is similar to those observed for other SET domain lysine methyltransferases, the predicted SET-domain fold of SETD3 does not help to disclose the molecular mechanisms of the enzyme's substrate recognition and histidine methylation activities, because the SET domain itself, as a short motif of ~110 aa, always requires the iSET and C-SET domains to complete its function (Chang et al., 2011; Trievel et al., 2002). Our solved substrate- and product-bound structures of SETD3 provide an insight into the pre-and post- methyl transfer states of this enzyme as it catalyses the histidine methylation of its substrate. For the first time, these SETD3 structures reveal the molecular mechanism by which a SET domain protein acts as the actin His73-specific methyltransferase.

Our structural analyses, assisted by mutagenesis and biochemical experiments, suggest that the recognition of β-actin and its subsequent catalysis by SETD3 are different from those observed for other SET-domain-containing lysine methyltransferases in the following respects: (i) the histidine recognition pocket of SETD3 is shallower than that of other SET lysine methyltransferases, and key residues of SETD3 that form hydrogen bonds with the His73 imidazole ring are not conserved in other SET proteins (Figure 2A–2B, Figure 1—figure supplement 2B); (ii) in two solved SETD3 complex structures, the methylated histidine rotates its side chain by 90° (Figure 2C); and (iii) the recognition of β-actin by SETD3 is highly dependent on the sequences flanking His73 of β-actin (Figure 2—figure supplement 1), consistent with the high binding affinity between SETD3 and a β-actin H73A peptide (66–88) (Figure 1B). Although N³-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 works only on His73 of β-actin, as well as on the corresponding histidines in α- and γ-actins. One implication is that no SETD3 ortholog has been found in S. cerevisiae.

Previously, we identified Carnosine N-methyltransferase 1 (CARNMT1) as the gene that encodes carnosine N-methyltransferase in mammals (Drozak et al., 2015). This enzyme catalyzes N¹-methylation of the histidine imidazole ring of carnosine (β-alanyl-L-histidine), an abundant dipeptide in the skeletal muscle of vertebrates. Recently, the structures of CARNMT1 bound to analogs of N¹-methylhistidine were also reported (Cao et al., 2018), allowing us to compare N³-histidine methylation by SETD3 with N¹-histidine methylation by CARNMT1 by studying the two complexes (Figure 2A and E). We found that SETD3 and CARNMT1 not only adopt different folds and AdoHcy-binding modes, but also display distinct histidine recognition and catalysis mechanisms. Asp316 of CARNMT1 forms two hydrogen bonds with the imidazole ring of the histidine to facilitate its deprotonation (Figure 2E) (Cao et al., 2018). In the SETD3–substrate complex, His73 stacks with Tyr313 and is hydrogen-bonded to the main chain carbonyl group of Asp275 (Figure 2A), which orients the His73 imidazole ring for efficient catalysis. As determined by the mutagenesis, ITC binding experiments and enzymatic assays (Table 1, Table 3), Tyr313 of SETD3 plays important roles in substrate recognition and in subsequent histidine methylation. Finally, the 90° rotation of the His73me imidazole ring was not found in CARNMT1-mediated histidine methylation.

In natively purified rabbit β-actin, His73 exists in the methylated form and it is localized close to the phosphate groups of ATP binding to actin (Kabsch et al., 1990). Although our complexes contain only a fragment of β-actin, they could still provide clues about the role of SETD3 in mediating actin functions. The residues 66–84 of β-actin consist of three β strands followed by an α-helix; however, the β-actin peptide adopts an extended conformation upon binding to SETD3 (Figure 2—figure supplement 3), suggesting that β-actin probably endures local structural remodeling during methylation. Given that SETD3 is found to be co-purified with other factors, such as the peptide release factor ERF3A (Kwiatkowski et al., 2018), SETD3 might require the assistance of other protein(s) to achieve the methylation of actin His73.

In the current investigation, we have identified the molecular mechanism of β-actin methylation by SETD3 and have shown that the enzyme displays a very high substrate specificity toward β-actin His73. The overexpression of SETD3 was recently reported to be associated with malignant cancer (Cheng et al., 2017), so the high-resolution complex structures of SETD3 might be helpful in providing a better understanding of the role of actin histidine methylation in human cancer, as well as in providing a structural basis for the design of specific SETD3 inhibitors in the near future.

During the peer review process for our manuscript, a manuscript was published that reveals the function of mammalian SETD3 in preventing primary dystocia and that presents the structures of SETD3 bound to an unmodified actin peptide (aa 66–80) (Wilkinson et al., 2019). The structural findings reported in this manuscript are similar to ours, except that our complex structures contain a longer actin fragment (aa 66–88) and that the histidine is methylated in one of our complexes.

Genes encoding the core region of SETD3 (aa 2–502) were synthesized by Sangon Biotech (Shanghai) and cloned into pET28-MHL (Genbank accession number: EF456735). Then, the plasmid was transformed into E. coli BL21 (DE3) and the recombinant protein was overexpressed at 16°C for 20 hr in the presence of 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Recombinant SETD3 proteins were purified with a fast flow Ni-NTA column (GE Healthcare). N-terminal 6 × His tags of recombinant proteins were removed with Tobacco Etch Virus (TEV) protease. Gel filtration and ion-exchange were employed for further purification. Purified SETD3 proteins, with their purities exceeding 90% as assessed by SDS-PAGE, were concentrated to 15 mg/ml and stored at −80°C before further use. The SETD3 mutants were constructed by conventional PCR using a MutanBEST kit (TaKaRa) and further verified by DNA sequencing. The SETD3 mutants were expressed and purified in the same way as the wildtype protein.

For kinetic studies, the enzyme and its mutant forms were produced as fusion proteins with the N-terminal 6 × His tag and purified by Ni-NTA affinity chromatography. Briefly, the E. coli cell paste was resuspended in 11 ml lysis buffer consisting of 25 mM Hepes (pH 7.5), 300 mM NaCl, 10 mM KCl, 1 mM dithiothreitol (DTT), 2 mM MgCl₂, 1 mM phenylmethanesulfonylfluoride (PMSF), 0.25 mg/ml hen egg white lysozyme (BioShop, Canada) and 250 U Viscolase (A and A Biotechnology, Poland). The cells were lysed by freezing in liquid nitrogen and, after thawing and vortexing, the extracts were centrifuged at 4°C (20,000 × g for 20 min). For the purification, the supernatant of E. coli lysate (11 ml) was diluted 3-fold with buffer A (50 mM Tris-HCl (pH 7.4), 400 mM NaCl, 10 mM KCl and 1 mM DTT) and applied onto a HisTrap FF crude column (1 ml) equilibrated with the 90% buffer A and 10% buffer B (30 mM imidazole). The column was then washed with 12 ml 90% buffer A and 10% buffer B, and the retained proteins were eluted with a stepwise gradient of imidazole (5 ml of 60 mM, 5 ml of 150 mM and 5 ml of 300 mM) in buffer A. The recombinant proteins eluted at the highest concentration of imidazole were further processed. The enzyme preparation was desalted onto PD-10 columns equilibrated with 20 mM Tris-HCl (pH 7.2), 50 mM KCl, 1 mM DTT and 6% sucrose. The yield of recombinant proteins ranged from 1.6 mg to 0.1 mg per 200 ml of culture for the wildtype and the N278A mutant, respectively. The purified proteins were aliquoted and stored at –70°C.

Recombinant human β-actin (ACTB, GenBank: NM_001101.4) was prepared as described previously (Kwiatkowski et al., 2018). The plasmid pCOLD I that encodes the human protein was a kind gift from Dr. Minoru Tamura (Ehime University, Japan) and was prepared as described by Tamura (2018).

For β-actin production, E. coli BL21(DE3) (Agilent, USA) cells were transformed with the DNA construct and cultured in 500 ml of LB broth (with 100 μg/ml ampicillin) at 37°C and 200 rpm until an OD₆₀₀ of 0.5 was reached. Protein production was induced by cold-shock (20 min on ice) and IPTG addition to a final concentration of 0.2 mM. β-Actin production was carried out for 20 hr at 15°C, 200 rpm, and harvested by centrifugation (6,000 × g for 10 min). The cell paste was resuspended in 27.5 ml lysis buffer consisting of 20 mM Hepes (pH 7.5), 1 mM DTT, 1 mM ADP, 0.5 mM PMSF, 2 μg/ml leupeptin, 2 μg/ml antipain, 0.2 mg/ml hen egg white lysozyme (BioShop), and 1,000 U Viscolase (A and A Biotechnology, Poland). The cells were lysed by freezing in liquid nitrogen and, after thawing and vortexing, the extracts were centrifuged at 4°C (20,000 × g for 30 min).

The pellet containing inclusion bodies was completely resuspended in buffer A (20 mM Hepes (pH 7.5), 2 M urea, 0.5 M NaCl, 5 mM DTT, 2 mM EDTA) with the use of a Potter-Elvehjem homogenizer and centrifuged at 4°C (20,000 × g for 10 min). The resulting pellet was then subjected to a further two rounds of sequential wash in buffer B (20 mM Hepes (pH 7.5), 0.5 M NaCl, 5 mM DTT, 2 mM EDTA) and buffer C (20 mM Hepes (pH 7.5), 0.5 M NaCl). The washed inclusion bodies were finally solubilized in loading buffer (20 mM Tris-HCl (pH 7.5), 6 M guanidine HCl, 0.5 M NaCl, 10 mM imidazole) and applied on a HisTrap FF column (5 ml) equilibrated with the same buffer.

The column was washed with 20 ml of loading buffer and the bound protein was refolded with the use of a linear gradient of 6–0 M guanidine HCl in loading buffer (40 ml for 20 min). Next, the column was washed with 15 ml of loading buffer without guanidine HCl and the retained proteins were eluted with a stepwise gradient of imidazole (25 ml of 40 mM, 25 ml of 60 mM and 20 ml of 500 mM). The recombinant proteins were eluted with 500 mM imidazole in homogeneous form as confirmed by SDS-PAGE (not shown). The β-actin preparation (10 ml) was immediately dialyzed against 400 ml of the buffer consisting of 20 mM Tris-HCl (pH 7.5), 1 mM DTT, 6% sucrose, 2 μg/ml leupeptin and 2 μg/ml antipain. The dialysis buffer was exchanged three times with the following dialysis times: 2, 2 and 12 hr. The purified β-actin was stored at −70°C.

Escherichia coli DNA was extracted by heating 50 μl of over-night cultured BL21(DE3) cells at 95°C for 15 min, whereas B. subtilis (ATCC 6633, Sigma-Aldrich) genomic DNA was purified from 100 mg of bacterial cells with the use of TriPure reagent (Roche, Swizerland) according to the manufacturer’s instructions.

The open reading frames encoding E. coli AdoHcy nucleosidase and B. subtilis adenine deaminase (NCBI Reference Sequence: NC_000913.3 and NC_000964.3, respectively) were PCR-amplified using either Pfu DNA polymerase alone or a mixture of Taq:Pfu polymerases (1:0.2), respectively, in the presence of 1 M betaine. The AdoHcy nucleosidase ORF was amplified using a 50-nucleotide primer containing the initiator codon preceded by an NdeI site and a 30-nucleotide primer with a HindIII site, whereas adenine deaminase DNA was amplified using a 50-nucleotide primer with the initiator codon preceded by a KpnI site and a 30-nucleotide primer with a BamHI site (for primer sequences, see Kwiatkowski et al., 2018). Amplified DNA products of expected size were digested with the appropriate restriction enzymes, cloned into the pCOLD I expression vector (pCOLD I/AdoHcy nucleosidase and pCOLD I/adenine deaminase) and verified by DNA sequencing (Macrogen, The Netherlands).

For protein production, E. coli BL21(DE3) cells were transformed with the appropriate DNA construct and a single colony was selected to start an over-night pre-culture. 100 mL of LB broth (with 100 mg/mL ampicillin) was inoculated with 10 ml of the pre-culture and incubated at 37°C and 200 rpm until an OD600 of 0.6 was reached. The culture was placed on ice for 20 min (cold-shock) and IPTG was added to a final concentration of 0.25 mM to induce protein expression. Cells were incubated for 16 hr at 13°C, 200 rpm, and harvested by centrifugation (6,000 × g for 10 min). The cell pellet was resuspended in 10 ml lysis buffer consisting of 25 mM Hepes (pH 7.5), 300 mM NaCl, 50 mM KCl, 1 mM DTT, 2 mM MgCl2, 1 mM PMSF, 5 mg/ml leupeptin and 5 mg/ml antipain, together with 0.25 mg/ml hen egg white lysozyme (BioShop) and 25 U DNase I (Roche). The cells were lysed by freezing in liquid nitrogen and, after thawing and vortexing, the extracts were centrifuged at 4°C (20,000 × g for 30 min).

For protein purification, the supernatant of the E. coli lysate (10 ml) was diluted 3-fold with buffer A (50 mM Tris-HCl (pH 7.2), 400 mM NaCl, 10 mM KCl, 30 mM imidazole, 1 mM DTT, 3 mg/ml leupeptin and 3 mg/ml antipain) and applied onto a HisTrap FF column (5 ml) equilibrated with the same buffer. The column was washed with 20–30 ml buffer A and the retained protein was eluted with a stepwise gradient of imidazole (25 ml of 60 mM, 20 ml of 150 mM and 20 ml of 300 mM) in buffer A. The recombinant proteins were eluted with 150–300 mM imidazole in homogeneous form, as confirmed by SDS-PAGE (not shown). The enzyme preparations were desalted onto PD-10 columns equilibrated with 20 mM Tris-HCl (pH 7.2), 50 mM KCl, 1 mM DTT, 6% sucrose, 2 μg/ml leupeptin and 2 μg/ml antipain. The yield of recombinant proteins was 1.2 mg and 3.1 mg of homogenous adenine deaminase and AdoHcy nucleosidase, respectively, per 200 ml of E. coli culture. The purified enzymes were aliquoted and stored at −70°C.

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 and concentrated proteins were diluted with ITC buffer (20 mM Tris, pH 7.5 and 150 mM NaCl). ITC experiments were performed by titrating 2 μl of peptide (1.2–1.5 mM) into cells containing 50 µM proteins on MicroCal (Malvern Panalytical, UK) at 25°C, with a spacing time of 160 s and a reference power of 10 µCal/s. Control experiments were performed by injection of peptides into buffer. Binding isotherms were plotted, analyzed and fitted in a one-site binding model by MicroCal PEAQ-ITC Analysis Software (Malvern Panalytical, UK) after subtraction of the respective controls. The dissociation constants (Kds) were determined from a minimum of two experiments (mean ± SD). The ITC binding curves are shown in Supplementary file 1.

All crystals were grown using the sitting drop vapor diffusion method at 18°C. For crystallization of SETD3 with methylated peptide, SETD3 (12 mg/ml) was pre-incubated with synthesized β-actin peptides (aa 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), 0.2 M magnesium acetate tetrahydrate, and 20% v/v polyethylene glycol 8000. For the crystallization of SETD3 with unmodified peptide, SETD3, at the concentration of 12 mg/ml, was pre-incubated with synthesized β-actin peptides (aa 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 (Wang et al., 2016) (SSRF). Data sets were collected at 0.9789 Å or 0.9789 Å, and were processed using the HKL2000 program (Otwinowski and Minor, 1997).

The initial structures of the SETD3-actin complexes were solved by molecular replacement using Phaser (McCoy et al., 2007) with a previously solved SETD3 structure (PDB: 3SMT) as the search model. Then, all of the models were refined manually and built with Coot (Emsley and Cowtan, 2004). The final structures were further refined by PHENIX (Adams et al., 2010). The statistics for data collection and refinement are summarized in Table 2.

Reversed-phase microcapillary/tandem mass spectrometry (LC/MS/MS) was performed using an Easy-nLC nanoflow HPLC (Proxeon Biosciences) with a self-packed 75 μm × 15 cm C18 column connected to a QE-Plus (Thermo Scientific) in data-dependent acquisition and positive ion mode at 300 nL/min. Passing MS/MS spectra were manually inspected to ensure that all b-and y-fragment ions aligned with the assigned sequence and modification sites. A 25 ul reaction mixture contained 2 µM SETD3 or SETD3 mutants (final concentration) and 20 µM peptide (final concentration) in a buffer containing 10 mM Tris-HCl, (pH 7.5), 20 mM NaCl and 10 µM AdoMet. The reaction was incubated at 37°C for 2 hr before being quenched (at 70°C for 10–15 mins). Then, 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.

Enzyme activity was determined by measuring the incorporation of the [³H]methyl group from S-[methyl-³H]adenosyl-L-methionine ([³H]AdoMet) into homogenous recombinant human (mammalian) β-actin or its mutated form in which histidine 73 was replaced by an alanine residue (H73A). The standard incubation mixture (0.06–0.11 ml) contained 25 mM Tris-HCl (pH 7.2), 10 mM KCl, 1 mM DTT and various concentrations of protein substrate and [¹H+³H] AdoMet (≈300–700 × 10³ cpm), as indicated in the legends to the figures and tables. When the effect of pH on recombinant enzyme activity was investigated, 25 mM Tris-HCl (pH 7.2) was replaced by 40 mM MES (pH 6.0 or 6.5), 40 mM Hepes (pH 7.0 and 7.5), or 40 mM Tris (pH 8.0, 8.5 or 9.0).

The incubation mixture was supplemented with recombinant AdoHcy nucleosidase and adenine deaminase to prevent AdoHcy accumulation, as indicated in the legends to the figures and tables. The reaction was started by the addition of an enzyme preparation and carried out at 37°C for 10 min. Protein methylation was linear for at least 15 min under all conditions studied. By analogy to assays of non-ribosomal peptide synthetase activity (Drozak et al., 2014; Richardt et al., 2003), the incubation was stopped by the addition of 0.05–0.11 ml of the reaction mixture to 0.025 ml of bovine serum albumin (BSA) (1 mg) 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.

V_max, K_M and k_cat were calculated for the methyltransferase activity of the studied enzymes with Prism 8.0 (GraphPad Software, La Jolla, USA) using a nonlinear regression.

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

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We are grateful to the staff members at beam lines BL17U1, BL18U1 and BL19U1 at the Shanghai Synchrotron Radiation Facility for their assistance in data collection. This work was supported by the National Natural Science Foundation of China Grants 31770806 (to CX), 31570737 (to CX), 31500601 (to SL), and 31501093 (to HY). JD is supported by Narodowe Centrum Nauki (Opus Grant UMO-2017/27/B/NZ1/00161). HY is supported by a China Postdoctoral Science Foundation Grant (No. 2015M580547). CX is supported by the Major/Innovative Program of the Development Foundation of the Hefei Center for Physical Science and Technology (2018CXFX007), the ‘Thousand Young Talent program’ and the Fundamental Research Funds for the Central Universities (WK2070080001)." The Structural Genomics Consortium is a registered charity (no. 1097737) that receives funds from: AbbVie; Bayer Pharma AG; BoehringerIngelheim; the Canada Foundation for Innovation; the Eshelman Institute for Innovation; Genome Canada through the Ontario Genomics Institute; the Innovative Medicines Initiative (European Union/European Federation of Pharmaceutical Industries and Associations; Unrestricted Leveraging of Targets for Research Advancement and Drug Discovery [ULTRA-DD] grant no. 115766); Janssen, Merck, and Company; Novartis Pharma AG; the Ontario Ministry of Economic Development and Innovation; Pfizer; the São Paulo Research Foundation (FAPESP); Takeda; and the Wellcome Trust (to JM).

© 2019, Guo et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.