Fine-tuning of lysine side chain modulates the activity of histone lysine methyltransferases

Histone lysine methyltransferases (KMTs) play an important role in epigenetic gene regulation and have emerged as promising targets for drug discovery. However, the scope and limitation of KMT catalysis on substrates possessing substituted lysine side chains remain insufficiently explored. Here, we identify new unnatural lysine analogues as substrates for human methyltransferases SETD7, SETD8, G9a and GLP. Two synthetic amino acids that possess a subtle modification on the lysine side chain, namely oxygen at the γ position (KO, oxalysine) and nitrogen at the γ position (KN, azalysine) were incorporated into histone peptides and tested as KMTs substrates. Our results demonstrate that these lysine analogues are mono-, di-, and trimethylated to a different extent by trimethyltransferases G9a and GLP. In contrast to monomethyltransferase SETD7, SETD8 exhibits high specificity for both lysine analogues. These findings are important to understand the substrate scope of KMTs and to develop new chemical probes for biomedical applications.


Results and discussion
The synthesis of γ-oxalysine (K O ) was envisioned by Lewis acid-catalyzed nucleophilic opening of serine-derived aziridine 8 by N-Fmoc-ethanolamine, yielding the desired building block 1 after protecting group modifications. This particular analogue has not been synthesized before, albeit that a similar threonine-derived building block was described by Gellman and co-workers, and by Vederas et al. 32,33 . Initially, we first protected the carboxylic acid of l-serine as an allyl ester followed by tritylation of the α-amino functionality. However, the initial Brønstedcatalyzed allylation under reflux conditions to produce compound 5 resulted in disappointingly low yields, mainly due to self-condensation of serine under the employed conditions. Alternatively, the first protection of the amino group of serine 3 with a trityl group to produce 4, and subsequent allylation afforded 5 in 42% yield over two steps, comparable to an earlier literature report 34 . The protected serine 5 was treated with MsCl under reflux conditions for 48 h to cleanly afford aziridine 6 in good yield comparable to the Thr analogue 32,33 . The trityl group was replaced with a benzyl carbamate (Cbz) in two steps to allow for a Lewis acid-catalyzed nucleophilic ring opening. This step was successfully achieved by employing BF 3 ·Et 2 O (0.5 equiv) under reflux conditions for 2 h in toluene, yielding 8. A one-pot replacement of the Fmoc group of 8 into Boc gave the penultimate protected amino acid 9. To allow for use in standard solid phase peptide synthesis (SPPS), both the allyl and Cbz protecting groups were removed by H 2 and Pd/C. A subsequent one-pot protection of the amine with Fmoc-OSu was performed to obtain the novel Fmoc-K O (Boc)-OH (1) in 75% yield (Fig. 2).
Next, the synthesis of Fmoc/Boc-protected γ-azalysine (K N ) 2 was achieved in 8 steps starting from Cbz-Dab-OH 10 (Fig. 3). The synthesis of this building block has been previously reported 35 . However, we designed a new synthetic route to produce this building block employing the readily available phthalimidoacetaldehyde thereby improving the overall yield (31%, 8 total steps) compared to a synthesis reported by Chhabra et al. (10%, 8 steps from 10) 35 . The synthesis started by protecting 10 as a methyl ester (11) in quantitative yield. After reductive amination of amine 11 with phthalimidoacetaldehyde, followed by protection of the secondary amine with a Boc group, intermediate 12 was obtained in a yield of 81%. Initially, selective deprotection of the phthalimide group in 12 was attempted with hydrazine, however, under the employed conditions the methyl ester was mainly converted into the corresponding acylhydrazide. To circumvent the formation of this byproduct, the ester was first hydrolysed under basic conditions. Subsequent phthalimide deprotection with hydrazine and protection of the resulting primary amine in the presence of Boc anhydride resulted in amino acid 13 in 54% yield over 3 steps. Final one-pot hydrogenolysis of the Cbz group, followed by Fmoc protection afforded the desired building block Fmoc-K N (Boc) 2 -OH (2) in 73% yield (Fig. 3).
With building blocks 1 and 2 in hand, we prepared histone peptide fragments containing the modified lysines for evaluation with human KMTs. To this end, six new peptides (Fig. 4) and three native peptides were synthesized by SPPS using Fmoc/t-butyl chemistry (Supplementary Schemes S1-S3; Supplementary Figs. S1-S3). The prepared histone peptides contained the N-terminal 15-residues of H3 (ARTKQTARK 9 STGGKA) in which K9 was replaced with either of the two analogues for evaluation with GLP and G9a. Additionally, the two analogues were also inserted at position 4 (ARTK 4 QTARKSTGGKA) to be examined with SETD7. To further expand the enzyme scope, we also prepared peptide fragment 13-27 of H4 (GGAKRHRK 20 VLRDNIQ), in which K20 was replaced with K O and K N for evaluation with SETD8. All synthetic histone peptides were produced in high purity using RP-HPLC; their purity was confirmed by analytical HPLC and their identity by LC-MS and MALDI-TOF MS (Supplementary Figs. S4-S6).
Histone peptides bearing the lysine analogues K O and K N were examined as substrates for the recombinantly expressed human KMTs using MALDI-TOF MS assays as a direct method to monitor histone methylation. We tested whether K O and K N can be methylated by di-and trimethyltransferases GLP and G9a, and monomethyltransferases SETD7 and SETD8. Samples containing the human KMT enzyme (2 µM), histone peptide (100 µM), and SAM (500 µM with GLP and G9a; 200 µM with SETD8 and SETD7) in Tris buffer (50 mM, pH 8.0) were incubated for 1 h at 37 °C. Following established enzymatic conditions [17][18][19]21,22 , all lysine-containing histone peptides were efficiently methylated by the four KMTs (Fig. 5a,d,g,j, respectively). Then the assessment of H3K O 9 and H3K N 9 peptides with GLP and G9a was carried out. Unlike trimethylation of lysine normally   (Fig. 5b,e). In the G9a assay, traces (< 5%) of the dimethylated species H3K O 9me2 were also observed. MALDI-MS assays with H3K N 9 showed that both enzymes catalyzed the formation of monomethylated H3K N 9me and dimethylated H3K N 9me2 (Fig. 5c,f). Omission of GLP/G9a resulted in no observable methylation, indicating an enzymatic process (black spectra in   . Similar results were observed with both enzymes upon screening the H3K N 9 peptide at high concentration, as the trimethylated product H3K N 9me3 was observed upon prolonged incubation (Supplementary Figs. S9-S13). These data indicate that the G9a/GLPcatalyzed first methylation reaction is the fastest, whereas the second and third methylations are comparatively slower, producing increasing amounts of higher methylation states products over time. It is worth noting that introduction of γ-oxalysine and γ-azalysine does not lead to complete alteration of the substrate specificity, but rather to modulation of activity by KMTs. Absence of the enzyme resulted in no observable methylation reaction (black spectra in the Supplementary Figs. S12, S13). Having established that GLP and G9a catalyzed efficient methylations of H3K O 9 and H3K N 9, we sought to investigate K O and K N with other members of the KMTs, monomethyltransferases SETD8 and SETD7. Applying the standard assay conditions, SETD8 notably catalyzed the monomethylation of H4K O 20 and H4K N 20 (Fig. 5h,i). Nearly complete monomethylation was observed with an increased concentration of SETD8 (10 µM) and SAM (1 mM) after 1 h at 37 °C ( Supplementary Fig. S14). On the other hand, examining these two analogues in the presence of SETD7 resulted in only traces (< 5%) of monomethylated species (Fig. 5k,l). The lack of methylation of these two analogues with SETD7 indicated a selectivity between KMTs. Increased amounts of SETD7 (10 µM) and SAM (1 mM), and prolonged incubation times (3 h) did not lead to formation of the monomethylated products either (only traces of H3K O 4me and H3K N 4me observed, Supplementary Fig. S15). Overall, the results show that histone lysine methyltransferases possess different activities towards lysine and simple lysine analogues; monomethyltransferase SETD7 is more restrictive towards K O and K N compared to SETD8 and trimethyltransferases GLP and G9a.
To determine the substrate specificity of SETD8 towards the two histone peptides bearing the unnatural lysine analogues K O and K N , a kinetic investigation was carried out employing a MALDI-TOF MS enzymatic assay 17,18,21 .
KMT-catalyzed methylation of the oxalysine-and azalysine-containing histone peptides was characterized by slower substrate conversion rates (k cat ) compared to their respective natural sequences, a result that we attribute to their poorer nucleophilic character due to the electron-withdrawing properties of the O and NH moieties. Furthermore, a highly hydrophobic channel of the methyltransferases is poorly tailored for significant electrostatic changes in the hydrophobic nature of the lysine side chain, likely leading to a larger penalty for desolvation of γ-azalysine and γ-oxalysine. Combined with the altered K m values for H4K O 20 and H4K N 20, this translated into a generally slightly worse catalytic efficiency for these two lysine analogues. Taken together, H4K O 20 and H4K N 20 were found as efficient substrates for SETD8 catalysis, with 2.1-and 2.8-fold decreases in k cat /K m values, www.nature.com/scientificreports/ respectively (Table 1; Supplementary Fig. S16). Comparisons of kinetics data for H4K20, H4K O 20, H4K N 20 with the highly related analogue H4K C 20 (ref. 18 ) shows that SETD8 very well tolerates subtle modifications at the γ position of the side chain, in the order H4K20 > H4K C 20 > H4K N 20 > H4K O 20. Following these trends, we anticipate that the high substrate specificity of SETD7 for H3K4 over analogues is a result of high K m values for the latter 18 ; more detailed kinetics analyses on H3K N 4 and H3K O 4 were therefore not possible, as very low levels of methylation were detected. Residual activity assays monitoring the KMT-catalyzed methylation of the histone peptides were then carried out, aimed at providing competitive data of these two substrates as compared to the 14-mer H3K9 peptide for the active site of G9a/GLP, and to determine whether the two lysine analogues inhibit G9a-and GLP-catalyzed methylation of H3K9. Both substrates appear to be very poor inhibitors of G9a and GLP (IC 50 > 100 µM), indicating that the H3K9 peptide outcompetes H3K O 9 and H3K N 9 peptides for binding in the active site of G9a/ GLP ( Supplementary Fig. S17). An examination of 15-mer H3K O 4 and H3K N 4 peptides as inhibitors of SETD7 using the 14-mer H3K4 substrate revealed that H3K O 4 exhibits weak inhibition activity (IC 50 = 38.9 µM), whereas H3K N 4 appears to be inactive (IC 50 > 100 µM) (Supplementary Figs. S18, S19).
To further validate the role of the lysine analogues as KMT substrates, we investigated the detection of GLPcatalyzed methylation of H3K O 9 and H3K N 9 by 1 H NMR and 1 H-13 C HSQC measurements. Before starting the enzymatic NMR studies, these histone peptides were analyzed by 1 H NMR and 1 H-13 C HSQC (Supplementary Figs. S20, S21). The enzymatic NMR sample contained GLP (8 µM), the H3K O 9 or H3K N 9 peptide (400 µM) and SAM (2 mM) in Tris-d 11 buffer at pD 8.0. After 1 h incubation at 37 °C, the samples were transferred to NMR tubes and then subjected to 1 H NMR analyses. GLP-catalyzed methylation of H3K9 resulted in a signal of H3K9me3 at 3.03 ppm ( 13 C 53.0 ppm, NMe 3 ) as shown in Fig. 6a,d, respectively, in line with recent work 21,36 . A concomitant conversion of SAM to SAH was evidenced by the 1 H NMR resonance at 2.62 ppm (corresponding to the methylene protons of SAH-CH 2 γ) for lysine and both lysine analogues (Fig. 6). H3K O 9 and H3K N 9 substrates showed the appearance of new singlet resonances at 3.09 ppm and 3.10 ppm, respectively (Fig. 6b,c). Subsequent 1 H-13 C HSQC measurements indicated that these signals were methyl groups of NMe 3 with correlated carbon resonance at 53.6 ppm and 53.1, respectively (Fig. 6e,f). It is noteworthy that trimethylation of lysine analogues as observed by NMR occurs at high concentration of GLP, in line with MALDI-TOF data using high concentrations of GLP/G9a.

Conclusion
Overall, our findings demonstrate that human KMTs in general have an ability to catalyze methylation of γ-azalysine and γ-oxalysine containing histones. Despite subtle chemical differences between lysine and its two simple analogues that differ in the γ position of the side chain, enzyme assays showed that different degrees of KMT-catalyzed methylation are observed. Methyltransferases G9a and GLP that catalyzed a complete trimethylation of H3K9 in vitro, predominantly catalyzed monomethylation of H3K O 9, and mono-and dimethylation of H3K N 9 at same conditions, demonstrating the fine-tuning of the lysine side chain leads to different methylation states. In addition, although SETD8 showed catalytic activity towards H4K O 20 and H4K N 20, the conversion of H3K O 4 and H3K N 4 by SETD7 was significantly reduced. These results highlight differences in activity of the human KMTs for methylation of simplest lysine analogues. Together with recent examinations of lysine analogues as substrates for KMTs, results presented here importantly contribute to a better understanding of biocatalysis of histone lysine methyltransferases that play essential roles in human health and disease.

Methods
Expression and purification of KMTs. Human KMTs enzymes were expressed and purified as previously described [36][37][38] . The methyltransferase plasmid (SETD8 residues 186-352, SETD7 residues 1-366, GLP residues 951-1235, G9a residues 913-1193) transformed into Escherichia coli Rosetta BL21 (DE3)pLysS cells. Cultures were grown at 37 °C in LB media containing kanamycin and chloramphenicol. Cells were grown to an OD 600 of 0.5-0.6 (approximately 2.5-3 h), at which point they induced by isopropyl β-D-1-thiogalactopyranoside (IPTG) and they were transferred to a temperature of 16 °C overnight. After letting the cells grow at this temperature, they were then harvested and lysed by sonication. The lysate was centrifuged at high-speed to remove unbroken cells. The supernatant was then centrifuged to further clean the lysate. Purification of the N-terminally his6tagged KMTs was performed using Ni-NTA affinity chromatography column, which was prewashed with lysis buffer. Target KMT enzyme was then eluted using a linear gradient concentration of imidazole. The elute was then concentrated with centrifugal concentrators (Millipore). All KMT enzymes were further purified by gel filtration on a Superdex 75 column (GE Healthcare) on an AKTA system. Purified proteins were concentrated employing Amicon Ultra Centrifugal Filter Units (Millipore) with suitable molecular weight cutoffs. Proteins concentrations were determined using the Nanodrop DeNovix DS-11 spectrophotometer and the purity was monitored by SDS-PAGE on a 4-15% gradient polyacrylamide gel (Bio-Rad). Enzyme kinetics. Kinetics studies were performed as described 17,18,21 . A solution of histone peptide NMR activity assays. NMR spectroscopy was carried out as described 21 . NMR spectra were recorded using a Bruker AVANCE III (500 MHz 1 H, 125 MHz 13 C) spectrometer equipped with a Prodigy BB cryoprobe. www.nature.com/scientificreports/ All samples were prepared in Eppendorf vials (1.5 mL volume) before being transferred to NMR tubes. The D 2 O solvent signal was used as internal lock signal and any residual HOD signal was suppressed. Chemical shifts are reported relative to the solvent water resonance (4.7 ppm). 33,39 . Ethanolamine (1.98 mL, 32.8 mmol) was dissolved in 2:1 acetone:sat. aq. NaHCO 3 (300 mL). Fmoc-OSu (11.04 g, 32.7 mmol) was added and the mixture was stirred at room temperature for 2 h. The acetone was removed in vacuo, and the residue was dissolved in water (200 mL) and EtOAc (200 mL (4) 34 . To a suspension of (S)-Serine (15.00 g, 142.7 mmol) in DCM (300 mL) under a nitrogen atmosphere, TMS-Cl (58 mL, 456.8 mmol) was added dropwise. The mixture was refluxed for 20 min, after which it was cooled to 0 °C. A solution of Et 3 N (66 mL, 473.5 mmol) in DCM (300 mL) was added slowly and the mixture was stirred at room temperature for 45 min. The mixture was cooled to 0 °C and MeOH (8 mL) was added dropwise. The mixture was warmed up to room temperature, after which Et 3 N (20 mL, 143.5 mmol) and Trt-Cl (39.79 g, 142.7 mmol) were added. The mixture was stirred at room temperature for 40 h, after which Et 3 N (100 mL) and MeOH (700 mL) were added. The solvents were removed in vacuo to yield an orange residue. This was dissolved in EtOAc (500 mL) and ice cold 5% aq. citric acid (300 mL). The organic layer was extracted with 2 M aq. NaOH (3 × 150 mL) and washed with water (3 × 150 mL). The combined aqueous phases was washed with EtOAc (150 mL) and neutralized with glacial acetic acid (20 mL). The aqueous phase was extracted with EtOAc (6 × 250 mL), after which the organic layer was washed with brine (400 mL). The organic layer was dried over Na 2 SO 4 , filtered, and the solvent was removed in vacuo to yield a light yellow solid (24.21 g, 49%  (5) 34 . To a solution of 4 (18.9 g, 54.4 mmol) in MeOH (190 mL) was added Cs 2 CO 3 (8.9 g, 27.3 mmol) in portions. The solution was stirred at room temperature for 10 min, after which the solvent was removed in vacuo to yield a beige solid. This was dissolved in DMF (38 mL), after which allyl bromide (4.7 mL, 54.3 mmol) was added dropwise. The mixture was stirred at room temperature for 18 h, after which the solvent was removed in vacuo. The residue was dissolved in EtOAc (190 mL) and washed with 5% aq. citric acid (625 mL). The combined aqueous layers were extracted with EtOAc (4 × 300 mL) and the combined organic layers were washed with water (12 × 625 mL). The organic layer was dried over Na 2 SO 4 and the solvent was removed in vacuo to yield yellow oil (18.11