Substrate Specificity of the HEMK2 Protein Glutamine Methyltransferase and Identification of Novel Substrates*

Bacterial HEMK2 homologs initially had been proposed to be involved in heme biogenesis or to function as adenine DNA methyltransferase. Later it was shown that this family of enzymes has protein glutamine methyltransferase activity, and they methylate the glutamine residue in the GGQ motif of ribosomal translation termination factors. The murine HEMK2 enzyme methylates Gln185 of the eukaryotic translation termination factor eRF1. We have employed peptide array libraries to investigate the peptide sequence recognition specificity of murine HEMK2. Our data show that HEMK2 requires a GQX3R motif for methylation activity. In addition, amino acid preferences were observed between the −3 and +7 positions of the peptide substrate (considering the target glutamine as 0), including a preference for Ser, Arg, and Gly at the +1 and a preference for Arg at the +7 position. Based on our specificity profile, we identified several human proteins that contain putative HEMK2 methylation sites and show that HEMK2 methylates 58 novel peptide substrates. After cloning, expression, and purification of the corresponding protein domains, we confirmed methylation for 11 of them at the protein level. Transfected CHD5 (chromodomain helicase DNA-binding protein 5) and NUT (nuclear protein in testis) were also demonstrated to be methylated by HEMK2 in human HEK293 cells. Our data expand the range of proteins potentially subjected to glutamine methylation significantly, but further investigation will be required to understand the function of HEMK2-mediated methylation in proteins other than eRF1.

Indirect effects may be mediated by an influence on other posttranslational modifications (2)(3)(4). Although protein methylation on lysine or arginine residues has been intensively studied (2)(3)(4)(5)(6)(7), the investigation of methylation of other amino acids including glutamine is still at the inception. Glutamine methylation was first described almost 40 years ago as a modification present in the Escherichia coli ribosomal protein L3 (8). Later, ribosomal polypeptide release factors (RFs), which recognize stop codons in mRNA in the A site of the ribosome were shown to contain the same modification as well (9). Recently, glutamine methylation was identified in the histone H2A protein at Gln 104 (10).
The HemK enzyme was first identified in E. coli in a genetic screen designed to isolate new heme biosynthesis mutants (11). The HemK protein at that time was thought to be involved in the oxidation of protoporphyrinogen to protoporphyrin IX. Despite the mutant phenotype, subsequent biochemical and genetic studies revealed that the hemK gene product appeared to have no direct involvement in the heme biosynthetic pathway (12). It is still unclear what exactly caused the defects in heme metabolism observed in the hemK mutants. Later, HemK homologs were found in various species including bacteria and lower and higher eukaryotes, and sequence alignments of HemK proteins revealed that all of them share a common (D/N/S)PP(Y/FW) amino acid sequence motif, which is characteristic for adenine-N6 and cytosine-N4 specific DNA MTases (PROSITE accession no. PS00092) (13,14). This led to the suggestion that HemK was itself an AdoMet-dependent DNA MTase (15), and its classification in data bases as "probable MTase modifying N6-adenine or N4-cytidine in DNA" including the renaming of many entries to N6AMT. However, no evidence could be found that HemK is able to methylate DNA (16,17). Later it was found that E. coli HemK methylates a glutamine residue in the ribosomal release factors (RF1 and RF2) at the universally conserved tripeptide GGQ motif (18,19), indicating that the alignment based assignment of HemK as an AdoMet dependent MTase was correct although the predicted substrate was wrong. This novel activity triggered renaming of HEMK2 to PrmC (for protein methyltransferase C). Glutamine methylation of the polypeptide release factor stimulates the hydrolysis of the peptidyl-tRNA in the ribosomal P-site, leading to the release of the nascent polypeptide chain and ensuring efficient translational termination (18). The HemK enzyme has two homologs in many eukaryotes, HEMK1 in mitochondria and HEMK2 in the cytosol. The human HEMK2, also called N6AMT1, and its yeast homolog called YDR140w or Mtq2p were shown to methylate the glutamine residue in the GGQ motif of the eukaryotic release factor eRF1 (17,20,21). The human HEMK1 (HMPrmc) and its yeast homolog Mtq1p methylate the corresponding site in the mitochondrial release factor and regulate the mitochondrial translational activity (22).
The first structures of HemK MTases have been determined from Thermotoga maritima (23) and E. coli (16). Closely similar, the structures revealed two domains: an N-terminal domain with no significant similarity to MTase sequences other than close HemK homologs and a C-terminal domain that contains the (D/N/S)PP(Y/FW) motif and shows strong similarity to members of the family 7 ␤-strand methyltransferases (also referred as class I methyltransferases). This group of enzymes contains the majority of AdoMet-dependent methyltransferases in most organisms, including DNA adenine and DNA cytosine MTases, RNA MTases, small molecule MTases, and protein arginine and some protein lysine methyltransferases (24,25). The structure of E. coli HemK in complex with the bacterial RF1 and the methyl donor product S-adenosyl-Lhomocysteine was also solved (26), showing that both the GGQ part and the central region of RF1 interact with the enzyme.
In eukaryotes, HEMK2 alone is not sufficient to methylate eRF1; it requires the binding partner TRM112 to exhibit high methyltransferase activity (21). Structural studies of the yeast homologs, Mtq2 and Trm112, revealed that Mtq2 binds to the zinc binding domain of Trm112 forming a heterodimer. This interaction triggers a conformational change of Mtq2, which increases the binding of AdoMet (27). In addition to that, Trm112 binding masks a hydrophobic region of Mtq2, and this prevents Mtq2 homodimer formation and enhances the expression of the enzyme in soluble form (27). In mice, HEMK2 knock-out heavily impaired postimplantation development of mutant embryos, and it led to early embryonic lethality indicating that HEMK2 is an essential protein in mammals (17). However, the mechanistic basis of the interaction of HEMK2 with its eRF1 substrate is not fully understood. In this context, the structure of the E. coli enzyme in complex with its cognate release factor is not of help, because the amino acid sequences of E. coli and mammalian release factors are distinct outside of the conserved GGQ motif. Therefore, the aim of this study was to characterize the substrate specificity of murine HEMK2. Based on this, we also examined whether HEMK2 has additional substrates beyond eRF1, similarly to other protein MTases, which often have several methylation substrates. We identified several new substrates of HEMK2 at the peptide and protein level in vitro and confirmed methylation of two of them in human calls.

Experimental Procedures
Cloning of Proteins-A bacterial expression pRSF-Duet1 vector that encodes the His-tagged murine HEMK2 (N6AMT1) and untagged TRM112, a pRSET vector encoding human eRF1, and mammalian expression constructs of HA-tagged HEMK2 and Myc-tagged TRM112 were kindly provided by Dr. G. L. Xu (17). The coding sequences of other putative substrate protein domains (see Table 2) were amplified from cDNA prepared from HEK293 cells and cloned into the pGEX-6P-2 vector as GST fusion proteins. The coding sequences of CHD5 (kindly provided by Dr. A. A. Mills), protein NUT (kindly provided by Dr. C. A. French), and eRF1 were subcloned into pEYFP-C1 vector (Clontech).
Expression and Purification of Proteins-For bacterial expression, E. coli BL21-CodonPlus (DE3) cells (Novagen) were transformed with the corresponding plasmid, and the cells were grown in Luria-Bertani medium at 37°C until an A 600 of 0.6 -0.8 was reached. The bacterial cultures were then either shifted to 30°C and induced with 1 mM isopropyl ␤-D-thiogalactopyranoside for 4 h or to 20°C and induced overnight (10 -12 h) with 1 mM isopropyl ␤-D-thiogalactopyranoside. Later, the cells were collected by centrifugation at 5,000 ϫ g.
For purification of HEMK2 and TRM112, the cell pellet was resuspended in sonication buffer (30 mM potassium P i buffer, pH 7.4, 500 mM KCl, 0.2 mM DTT, 1 mM EDTA, 20 mM imidazole, and 10% glycerol), and the cells were disrupted by sonication. The samples were centrifuged at 20,000 ϫ g for 90 min, and supernatants were passed through nickel-nitrilotriacetic acid-agarose beads (Qiagen) that were preincubated with the sonication buffer. The beads were washed with sonication buffer. The bound proteins were eluted with the sonication buffer containing 220 mM imidazole. Proteins were dialyzed against low glycerol dialysis 1 buffer (20 mM HEPES, pH 7.2, 200 mM KCl, 0.2 mM DTT, 1 mM EDTA, and 10% glycerol) for 3 h and then overnight against high glycerol dialysis 2 buffer containing 65% glycerol.
For purification of GST-tagged substrate proteins, the cell pellet was resuspended in sonication buffer (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1 mM DTT, and 5% glycerol), and the cells were disrupted by sonication. The samples were centrifuged at 20,000 ϫ g for 90 min, and supernatants were passed through glutathione-Sepharose 4B resin (GE Healthcare) which were preincubated with the sonication buffer. The beads were washed with wash buffer (50 mM Tris/HCl, pH 8.0, 500 mM NaCl, 1 mM DTT, and 5% glycerol). The bound proteins were eluted with the wash buffer containing 40 mM glutathione. Proteins were dialyzed against low glycerol dialysis 1 buffer (20 mM Tris/HCl, pH 7.4, 100 mM KCl, 0.5 mM DTT, and 10% glycerol) for 3 h and then overnight against high glycerol dialysis 2 buffer (20 mM Tris/HCl, pH 7.4, 100 mM KCl, 0.5 mM DTT, and 60% glycerol).
Cell Culture, Transfection, and Immunoprecipitation-HEK293 cells were grown in Dulbecco's modified Eagle's medium (Sigma) supplemented with 5% fetal bovine serum, penicillin/ streptomycin, and L-glutamine (Sigma). The mammalian expression plasmids of HA-HEMK2, myc-TRM112, and YFP-tagged substrate proteins were transfected into cells using polyethylenimine (Promega) according to the manufacturer's instructions. In transfections without HEMK2 and TRM112 expression plasmids, corresponding amounts of empty pcDNA3.1 vector were added. After 72 h of transfections, the cells were washed with PBS buffer and harvested by centrifugation at 500 ϫ g for 5 min. The YFP-fused substrate proteins were immunoprecipitated from the cell extract using GFP-Trap A (Chromotek) following the manufacturer's instructions.
Synthesis of Peptide SPOT Arrays-Peptide arrays were synthetized using the SPOT synthesis method using an Autospot peptide array synthesizer (Intavis AG) (28). Each peptide spot had a diameter of 2 mm and contained ϳ9 nmol of peptide (Autospot Reference Handbook; Intavis AG). Successful synthesis of peptide arrays were confirmed by bromphenol blue staining of the membranes. Data analysis and calculation of discrimination factors were conducted as described (7,29).
Peptide Methylation Assays-Peptide SPOT arrays were preincubated for 10 min in methylation buffer containing 10 mM Tris/HCl, pH 7.6, 50 mM KCl, 10 mM Mg(OAc) 2 , and 1 mM DTT. Afterward the SPOT arrays were incubated for 60 min in methylation buffer containing 1.3 M HEMK2 and 0.76 M radioactively labeled [methyl-3 H]AdoMet (PerkinElmer Life Sciences) at room temperature (25°C). The arrays were then washed five times for 5 min with 50 mM NH 4 HCO 3 and 1% SDS. Later the SPOT arrays were incubated for 5 min in Amplify NAMP100V solution (GE Healthcare). Finally, the arrays were exposed to Hyperfilm TM high performance autoradiography films (GE Healthcare) in the dark for 1-3 days at Ϫ80°C. The films were developed using an Optimus TR developing machine.
In Vitro Methylation of Protein Domains-Protein methylation reactions were performed in methylation buffer containing 10 mM Tris/HCl, pH 7.6, 50 mM KCl, 10 mM Mg(OAc) 2 , 1 mM DTT supplemented with 0.76 M radioactively labeled [methyl-3 H]AdoMet and 4,5 M HEMK2 overnight at 25°C. The methylation reaction was stopped by incubating the sample in SDS loading buffer at 95°C for 5 min. Afterward, the samples were separated by SDS-PAGE, and the methylation signal was detected either by autoradiography or using the methylglutamine specific antibody.
Generation of the Methylglutamine Antibody-A methylglutamine antibody was generated by the Biotem (Apprieu, France) following standard procedures. In short, a GQ(me)G tripeptide was synthesized and conjugated to the keyhole limpet hemocyanin protein as carrier. New Zealand White rabbits were immunized with the conjugated protein in four subsequent injections at days 0, 7, 14, and 42. Preimmune serum and serum after days 49 and 63 were collected. The quality of unpurified anti-methylglutamine serum was initially assessed by ELISA using the modified and unmodified peptides using preimmune serum as control. After the initial screening of the serum in ELISA, day 63 serum was collected. Because we still observed some residual binding to unmethylated CHD5 and NUT proteins, it was further purified by passing through GST-Sepharose (GE Healthcare) beads bound with the unmethylated recombinant proteins. The flowthrough was collected, and specific binding to glutamine-methylated eRF1, CHD5, and NUT proteins was validated as described in the text with the methylated and unmethylated proteins. The final dilution of the methylglutamine antibody used in the Western blots was 1:10,000.

Results
Purification and Activity of HEMK2-The His-tagged mouse HEMK2 protein and the mouse TRM112 homolog were coexpressed in E. coli BL21-CodonPlus cells and purified with good yield (Fig. 1A). Because the eukaryotic translation termination factor eRF1 was the only known HEMK2 substrate, eRF1 peptides were used to test the activity of the purified enzyme. Peptide SPOT arrays were synthesized on cellulose membrane with 15-amino acid peptides comprising the eRF1 residues 178 -192 surrounding the methylation target site Gln 185 . Control peptides were used in which Gln 185 was mutated to alanine. The peptide arrays were then incubated with the HEMK2-TRM112 complex and radioactively labeled [methyl-3 H]AdoMet. The transfer of the methyl groups to the peptide was detected by autoradiography. As shown in Fig. 1B, a clear methylation signal was observed with the wild type but not with the mutant peptides, which confirms the specific methylation of Gln 185 in the eRF1 peptide.
Specificity Analysis of HEMK2-To examine the specificity of HEMK2 in peptide recognition, mutational scanning peptide SPOT arrays were synthesized using the sequence of eRF1 (179 -192) as a template. In these arrays, each peptide contained an exchange of one amino acid of the original sequence against one of the 20 proteinogenic amino acids, such that all possible single exchange variants of the original sequence are investigated. In total, 300 peptides were synthetized (20 possible amino acids ϫ 14 residues plus one wild type peptide at the beginning of each row). The peptide arrays were then incubated with the HEMK2-TRM112 complex and radioactively labeled [methyl-3 H]AdoMet, and the methylation of each peptide was detected by autoradiography as described above (Fig. 2A). The advantage of this method is that all peptides are methylated in competition, so that during the linear phase of the methylation reaction, the relative methylation rate of each peptide is proportional to the catalytic single turnover rate constant divided by the dissociation constant (k st /K D ) of the enzyme for the respective peptide substrate (7,29). As a result, the amount of radioactivity transferred to each peptide spot is directly correlated with the enzymatic activity toward that particular peptide. The experiment was repeated three times, and the results of the individual experiments were normalized and averaged (Fig.  2B). Standard deviations indicate that the obtained data are of good quality, because 70% of the peptide substrates exhibited S.D. less than 10%, and more than 95% of the peptides exhibited S.D. less than 20% (Fig. 2D). Next, to assess the contribution of each amino acid of the peptide substrate, the discrimination factor for the recognition of each amino acid residue at each position was calculated (Fig. 2C) (7,29). These factors represent the preference of the enzyme for one particular amino acid over all other amino acids at a given site. The results show that HEMK2 has long recognition motif encompassing the residues from Arg 182 (Ϫ3) to Arg 192 (ϩ7) of eRF1 (defining the target Gln 185 as position 0). In addition to the target residue Gln 185 , HEMK2 exhibits absolute specificity at the Ϫ1 (Gly 184 ) and ϩ4 positions (Arg 189 ). At these sites, an exchange of the cognate amino acid residue with other amino acids abolished the methylation activity. Hence, the minimal recognition element of HEMK2 is a GQX 3 R motif. In addition, stronger preferences were also seen at ϩ1 (Ser 186 , preference for Ser, Arg, and Gly) and Methylation of Putative Peptide Substrates-Lysine and arginine methyltransferases were initially identified as histone modifying enzymes, but some of them later were found to methylate other proteins and play an important role in various cellular functions (2-4, 7). The HEMK2 substrate specificity profile shows that it accepts several other residues on both sides of the target glutamine apart from the cognate residues, which indicates that it could also methylate other proteins apart from eRF1. To identify putative candidate substrates, Scansite (30) searches were performed with search profiles as indicated in Table 1. In total, 138 putative novel substrates were identified in the human proteome with substrate specificity profile 1. A second search with a more relaxed specificity profile (profile 2) identified 164 additional substrates candidates. Methylation of these putative substrates was initially tested on peptide arrays. For this, 15-amino acid peptides were synthesized containing the target glutamine in the center. Two independent SPOT arrays with the novel substrates from the corresponding profiles were synthesized always including eRF1 and eRF1 Q185A as positive and negative controls. The peptide arrays were methylated with HEMK2-TRM112 as described above (Fig. 3, A  and B). Quantitative analysis of the results revealed that 49 of the putative substrates were methylated with an activity comparable with eRF1, and 76 peptide substrates showed a reduced but clearly detectable activity. The high frequency of methylated peptides in these libraries confirms the validity of the search motifs. Moreover, most of the strongly methylated peptides contained an Arg at the ϩ7 position (corresponding to Arg 192 ), although this was not included in the search parameters. Furthermore, most of the highly methylated peptides contained an Ser, Arg, or Gly at the ϩ1 position, although other residues were allowed at this position in the search profiles. In summary, HEMK2 methylates several novel peptide substrates in vitro, and the sequences of the strongly methylated peptides confirmed the specificity data obtained in the mutational scanning approach.
Methylation of Putative Protein Substrates-Although several novel peptide substrates were methylated in par with the eRF1 peptide, this cannot be extrapolated to protein methylation, because the target glutamine may not be accessible to the enzyme in a folded protein. Therefore, we next aimed to investigate the methylation of these substrates at the protein level. 58 proteins were selected that were strongly methylated at the peptide level, and the corresponding protein domains containing the target glutamine residue were cloned as GST fusion proteins. The candidate proteins were overexpressed and purified by affinity chromatography. Finally, 20 protein domains were successfully purified with sufficient yield for a methylation assay. Methylation reactions were performed by incubating the purified protein domains with the HEMK2-TRM112 complex in methylation buffer containing radioactively labeled [methyl-3 H]AdoMet. The samples were later separated by polyacrylamide gel electrophoresis, and the transfer of methyl groups to the substrate proteins was detected by autoradiography. As shown in Fig. 4A, methylation was detected on 11 of the 20 protein domains (Table 2). Four protein domains (CHD5, NUT, ANKRD34A, and ABCA2) were methylated in par with the eRF1, and six additional domains were methylated at lower level (GHDC, ZSCAN10, RPP1, ASH1L, ARGEF10, and TGFB3). In addition, strong methylation was observed on AMPD2, but it could only be obtained in a highly degraded form. Site-directed mutagenesis was performed to confirm the methylation of the predicted target glutamine. For this, the target glutamine was exchanged to arginine in all 11 protein domains. The mutated protein domains were expressed, purified, and analyzed for methylation using equal amounts of wild type and the target glutamine mutant protein domains (Fig.   4B). HEMK2 exhibited no activity on all the target glutamine mutant proteins, confirming methylation at the predicted target glutamine residues.
HEMK2 Methylates Novel Protein Substrate in HEK293 Cells-After confirming target glutamine methylation of several protein domains in vitro, we next aimed to investigate their methylation in cells. For this purpose, we chose two proteins that were strongly methylated in vitro (CHD5 and NUT). For detecting cellular methylation, a methylglutamine specific antibody was generated. Validation of the antibody was performed using in vitro methylated and unmethylated proteins and protein domains of eRF1, CHD5, and NUT. Methylated substrate proteins were generated by incubating the proteins with the HEMK2-TRM112 complex in methylation buffer in presence of unlabeled AdoMet. Equal amounts of unmethylated substrate proteins were used as negative control. The proteins were separated on SDS-PAGE, transferred to nitrocellulose membrane, and probed with the antibody. After purification with unmethylated proteins, the antibody specifically recognized the methylated substrate proteins, which indicates that the antibody can be used to detect the methylated glutamine (Fig. 5A).  Table S1). A, the potential substrates identified using specificity profile 1. B, the potential substrates identified using specificity profile 2 ( Table 1). The eRF1 wild type and the Q185A mutant peptides were included as controls and marked with black and gray arrowheads. To investigate the methylation of substrate proteins by HEMK2 in human cells, the YFP-fused substrate protein domains were transiently expressed in HEK293 cells either alone or together with HEMK2 and TRM112. We first confirmed the expression of the HA-tagged HEMK2 (Fig. 5B) and Myc-tagged TRM112 (Fig. 5C). Next, the YFP fused substrate protein domains were immunoprecipitated from cells using GFP trap beads. Substrate proteins purified from the cells with and without HEMK2-TRM112 were separated on SDS-PAGE and analyzed by Western blot with anti-YFP antibodies to confirm their identity and adjust the protein concentrations of the samples (Fig.  5D). Next, methylation of the target proteins was analyzed using the methylglutamine antibody. Increasing concentrations of recombinant unmethylated protein were loaded as negative con-trols to exclude any cross-reactivity of the antibody with unmethylated proteins in these immunoblots (Fig. 5, E and F). In both experiments, the methylglutamine antibody specifically bound to the substrate proteins purified from the HEK293 cells, and no binding to recombinant unmethylated substrates was detected even at high protein concentrations. With CHD5, methylation could only be detected after coexpression of HEMK2 and TRM112, indicating that a HEMK2-dependent methylation occurred in the human cells. In the case of NUT, weak methylation was observed in the absence of ectopically expressed HEMK2 and TRM112, which likely was due to the endogenous enzymes. The methylation signal increased strongly after coexpression of NUT with HEMK2 and TRM112, again indicating that a HEMK2-dependent methylation took place.  Table 2. A, detection of protein methylation. Purified GST-fused protein domains were incubated with the HEMK2-TRM112 complex in the presence of radioactively labeled [methyl-3 H]AdoMet, and the transfer of methyl groups to the substrate proteins was detected by autoradiography. The Coomassie-stained gel image shows the amount of substrate proteins used in the methylation assays. As a positive control, methylation of eRF1 was included. B, confirmation of methylation of the predicted target glutamine. Purified GST-fused WT and target Glu to Arg mutant proteins were incubated with the HEMK2-TRM112 complex in the presence of radioactively labeled [methyl-3 H]AdoMet, and the methylation signal was detected by autoradiography. The Coomassie-stained images show the equal loading of wild type and corresponding Gln to Arg mutants.

Discussion
Methylation on lysine and arginine residues of histone proteins has been intensively studied, and its significant role in chromatin regulation has been well established (31,32), but methylation of these residues also occurs on non-histone proteins, and it has emerged as an important post-translational modification affecting several protein functions like stability, activity, protein-protein interactions (2, 3). However, protein methylation is not limited to lysine and arginine residues; other amino acid residues like cysteine, histidine, and glutamine have also been reported to carry this modification. The first discovered mammalian protein glutamine methyltransferase was HEMK2, which methylates the glutamine residue in the GGQ motif of the eRF1 protein. Recently, Kouzarides and co-workers (10) showed that the Nop1 methyltransferase in yeast, and its human homolog Fibrillarin, methylate H2A protein at Gln 105 at the 35S ribosomal DNA transcriptional unit. Hence, overall so far only two proteins with a methylglutamine modification are described in the human proteome. Herein, we investigated the substrate specificity profile of the HEMK2 MTase. For this purpose, we used a strategy employing peptide libraries that was highly successful with different protein lysine methyltransferases (33)(34)(35). Our results describe the specificity of HEMK2, and we identified several novel protein substrates of this enzyme.
Methylation of peptides with point exchanges in the sequence of eRF1 (179 -192) showed that HEMK2 recognizes a long peptide stretch ranging from the Ϫ3 to the ϩ7 position. At some sites, clear preferences for the canonical residues could be detected most prominently at the Ϫ1 (Gly 184 ) and ϩ4 (Arg 189 ) positions, but also at ϩ1 (Ser 186 ) and ϩ7 (Arg 192 ). Further structural and mechanistic interpretation of these data awaits the successful determination of a HEMK2-TRM112 structure in complex with eRF1 or an eRF1 peptide. However, our data also show that HEMK2 accepts other residues apart from the native sequence at certain positions. Using the derived specificity profile, several potential novel substrates were identified in the human proteome, and we showed that 58 peptides derived from them were efficiently methylated by HEMK2. Most of the highly methylated substrates had an Arg at the ϩ7 position and the preferred residues at the ϩ1 position, although this infor-mation was not included in the search filter. This result shows that the sequence motif of HEMK2 remains similar irrespective of the template sequence used for the specificity analysis (eRF1 in our case). This finding indicates that our peptide array data generated by variation of the eRF1 sequence can be extrapolated to other sequences.
We next tested methylation of several of the identified candidate substrates at the protein level; however, many of them failed at different stages of cloning, expression, and purification. Finally, we were able to purify 20 protein domains and detected methylation signals on 11 of them; strong methylation comparable with eRF1 was observed with four proteins (CHD5, NUT, ANKRD34A, and ABCA2); and weaker methylation signals were observed on 7 other proteins. Based on this, one may extrapolate that approximately 8 additional strong methylation substrates should be among the 38 additional highly methylated peptides that could not be tested. Furthermore, we tested the methylation of two proteins (CHD5 and NUT protein) in HEK293 cells and could confirm methylation of both of them, suggesting that strong in vitro protein domain methylation is indicative of cellular methylation in this system. Hence, our findings expand the range of human proteins potentially subjected to glutamine methylation considerably. While so far only two human proteins were known to be glutamine-methylated, we now identified 2 more HEMK2 substrate proteins in cellular studies, 9 additional ones in vitro at the protein level, and more than 50 additional peptide substrates. Based on the extrapolation described above, our data suggest that HEMK2 has at least approximately 13 strong methylation substrates in human cells (eRF1, 4 candidates identified here at protein level and 8 more candidates expected among the methylated peptides), which suggests that glutamine methylation is more widespread than currently known.
However, the biological significance of the methylation of the novel HEMK2 substrates still needs to be studied. Of note, glutamine methylation is a subtle post-translational modification. In case of eRF1, glutamine methylation changes the properties of the catalytic pocket directly; in case of methylation of H2A, protein-protein interaction is modulated. It is not known at present, if reading domains exist that specifically interact with glutamine-methylated proteins similarly as the well known  The cells were transfected with YFP-tagged substrate proteins either alone or together with HEMK2 and TRM112. Cell lysates were harvested, and the target proteins were enriched by GFP trap. The purified target proteins were analyzed by Western blots with anti-YFP antibodies, and the concentrations of the samples obtained with and without HEMK2/TRM112 overexpression were adjusted. E, cellular methylation of YFP-tagged CHD5 overexpressed in HEK293 cells in the presence and absence of ectopically expressed HEMK2 and TRM112. The substrate protein was purified using GFP trap as shown in D. Protein methylation was tested by Western blot with the methylglutamine antibody. The Ponceau S-stained image indicates that the loading control, recombinant unmethylated CHD5 protein was used as negative control for methylglutamine specificity. F, cellular methylation of YFP-tagged NUT overexpressed in HEK293 cells. The experiment was conducted as described in E, with the exception that recombinant unmethylated NUT domain protein was used as negative control in the Western blot with the methylglutamine antibody.
domains for readout of methyl-lysine, methylarginine, acetyllysine, and phosphoserine (36). However, it is clear that CHD5 and NUT (which were methylated in cells by HEMK2) both have very important biological functions. They are involved in chromatin regulation and also play important roles in tumor formation. CHD5 belongs to the family of chromatin remodeling proteins, it contains multiple PHD and Chromo domains and functions as tumor suppressor in variety of tumor types (37). NUT has been reported to undergo chromosomal translocations with BRD4, resulting in the formation of a BRD4-NUT fusion oncoprotein in 75% of the NUT midline carcinoma, which is an aggressive human cancer (38). Interestingly, the NLS of NUT resides in the C-terminal sequence (positions 1017-1042) (39), very close to the HEMK2 methylated site Gln 1046 . However, we did not observe any changes in the localization of methylated and unmethylated YFP-tagged NUT protein in HEK293 cells (data not shown). Nevertheless, it would be interesting to see whether the methylation affects the targeting of BRD4-NUT to chromatin.

Conclusions
Enzymes of the HemK family have been subjected to a long search full of twists and turns to find their biological function. Starting as factors believed to be involved in heme biogenesis, they changed to putative DNA methyltransferases but finally were shown to function as protein glutamine methyltransferases, which methylate translation termination factors, eRF1, in case of the human and mouse HEMK2 enzyme. Here, we identify several novel protein substrates of HEMK2 in vitro and in vivo, and our data expand the range of proteins potentially subjected to glutamine methylation significantly. The strong phenotype of HEMK2 deletion in mice (17) and the important biological role of the novel HEMK2 substrates described here may suggest that the biological function of HEMK2 is not restricted to its role in translational termination mediated by eRF1. Hence, the search for the biological role of HEMK2 must be continued beyond eRF1 methylation. However, the investigation of the signaling functions of glutamine methylation is complicated by the fact, that the addition of a methyl group on a glutamine residue is a very subtle modification, which only slightly affects its chemical properties. Our results indicate that additional dedicated research will be needed to understand the role of glutamine methylation in signaling networks and other cellular processes.
Author Contributions-S. K. and A. J. devised the research. D. K. and S. K. conducted the experiments. D. K., S. K., and A. J. analyzed the data. All authors contributed to the writing of the manuscript and approved the final version of the manuscript.