While the oxygen-dependent reversal of lysine Nɛ-methylation is well established, the existence of bona fide Nω-methylarginine demethylases (RDMs) is controversial. Lysine demethylation, as catalysed by two families of lysine demethylases (the flavin-dependent KDM1 enzymes and the 2-oxoglutarate- and oxygen-dependent JmjC KDMs, respectively), proceeds via oxidation of the N-methyl group, resulting in the release of formaldehyde. Here we report detailed biochemical studies clearly demonstrating that, in purified form, a subset of JmjC KDMs can also act as RDMs, both on histone and non-histone fragments, resulting in formaldehyde release. RDM catalysis is studied using peptides of wild-type sequences known to be arginine-methylated and sequences in which the KDM’s methylated target lysine is substituted for a methylated arginine. Notably, the preferred sequence requirements for KDM and RDM activity vary even with the same JmjC enzymes. The demonstration of RDM activity by isolated JmjC enzymes will stimulate efforts to detect biologically relevant RDM activity.
Post-translational modifications, which vastly increase the size of the functional proteome, play central roles in numerous cellular processes, including enzyme catalysis, protein–protein interactions and gene regulation1. Modifications to histones, which are especially prevalent on the N-terminal tail of histone H3, include Nɛ-acetylation and ubiquitination of lysine residues, phosphorylation of serine and threonine residues, and Nɛ/ω-methylation of lysine and arginine residues2. Coupled with other factors, these modifications regulate the sets of genes that are transcribed in a time- and context-dependent manner. While lysine acetylation is usually transcriptionally activating, arginine- and lysine methylation can be either activating or repressive, depending on the site and extent of methylation.
Arginine/lysine methylation is catalysed by S-adenosyl-methionine-dependent methyltransferases (protein arginine methyltransferases, PRMTs, and lysine methyltransferases, KMTs, respectively)3. Although for many years N-methylation was thought to be a stable, irreversible modification, the reversal of lysine methylation is now well established. Two families of histone Nɛ-methyllysine demethylases (KDMs) have been identified; the flavin-dependent KDM1s (lysine-specific demethylase) and the Fe(II)- and 2-oxoglutarate (2OG)-dependent JmjC-domain-containing enzymes. Both families of KDMs operate via methyl-group oxidizing mechanisms that result in the co-production of formaldehyde4. Lysine methylation/demethylation can thus occur on a timescale of only two enzyme-catalysed steps, that is, has the potential to be highly dynamic.
In contrast to lysine methylation, the extent to which arginine methylation is dynamic is much less clear. Arginine residues undergo PRMT-catalysed methylation on their terminal guanidine nitrogens to give monomethyl (MMA/Rme), and symmetric (SDMA/Rme2s) and asymmetric (ADMA/Rme2a) dimethylated forms5. The peptidyl-arginine deiminases (PADs) catalyse the hydrolysis of the guanidino arginine side chain to the urea group of citrulline. Both PAD2 and PAD4 catalyse citrulline formation at multiple positions on histone tails6. In the case of PAD4, this activity has also been reported on monomethylarginine residues7, although evidence that this activity occurs in an in vivo setting remains limited8. In either case, demethylimination is not a ‘true’ demethylation reaction—it removes but does not reverse Nω-methylation. Neutral citrulline is produced, which has considerably different chemical properties compared to unmethylated arginine. One 2OG oxygenase, JMJD6, has been reported to catalyse methylarginine demethylation of histone H3/H4 residues9. Some subsequent reports have not reproduced these observations, and JMJD6 has also been assigned as a lysine C-5 hydroxylase acting on mRNA splicing-regulatory proteins, and potentially histone proteins10,11,12. Other recent reports have identified further arginine demethylation activity by JMJD6 on histone and non-histone substrates13,14,15,16; however, the biochemical evidence for methylarginine demethylation by JMJD6 is not unequivocal17.
Despite the controversial history as potential methylarginine demethylases (RDMs), the 2OG oxygenases represent good candidates for such reactions because of the broad spectrum of oxidative reactions that they catalyse. Some JmjC oxygenases catalyse hydroxylation of lysine residues to give stable hydroxylated products18. Arginine demethylation could occur via oxidation of the arginine Nω-methyl group and concomitant loss of formaldehyde in a reaction analogous to that of lysine demethylation (Fig. 1a,b). In a recent study we found that the JmjC KDMs can catalyse oxidation of Nɛ-alkyl groups other than methyl groups, including oxidation reactions on isopropyl groups at a position equivalent to that required for oxidation of a Nω-methylarginine group (Fig. 1c)19. Here we report systematic studies on the potential of the catalytic domains of representative human JmjC enzymes to catalyse Nω-methylarginine demethylation. The results clearly demonstrate that some, but not all, purified JmjC KDMs can act as RDMs.
JmjC enzymes have RDM activity
To test whether the catalytic domains of the JmjC KDMs can act as RDMs, we produced truncated recombinant proteins containing the catalytic domains of representatives of the six identified human JmjC KDM subfamilies (KDM2A, KDM3A, KDM4E, KDM5C, KDM6B and PHF8 (also known as KDM7B), Fig. 2a)20,21. The representative JmjC KDMs were tested for in vitro demethylation activity with synthetic fragment sequences of the histone H3 tail with mono- and asymmetric/symmetric dimethylated arginine residues replacing the Nɛ-methylated lysine residues known to be demethylated in established substrates for each JmjC KDM (Supplementary Table 1). The corresponding ‘wild-type’ peptides with the naturally positioned Nɛ-methylated lysine residues were used as positive controls. Reactions were initially analysed using MALDI-TOF mass spectrometry (MS) to test for demethylation (observed as a −14 Da mass shift).
For KDM2A and PHF8, demethylation was observed only for the characterized lysine substrates (Supplementary Figs 1 and 2), that is, no RDM activity was detected. Interestingly, however, in addition to demethylation of the characterized lysine substrates, evidence for demethylation of both asymmetric and symmetric dimethyl- and monomethylarginine residues was accrued by MS for KDM3A, KDM4E, KDM5C and KDM6B (Fig. 2b and Supplementary Figs 3–6), that is, product peptide spectra manifested mass shifts of −14 Da (and in some cases −28 Da, corresponding to the loss of two methyl groups) relative to the methylated substrate residue. Under our incubation conditions, such mass shifts were only observed in the mass spectra of reactions in the presence of enzyme. KDM3A, KDM4E and KDM5C catalysed demethylation of both symmetric and asymmetric dimethylarginine residues to give a mixture of both the monomethylated and unmethylated arginine residues. At least under the tested conditions, however, we observed that KDM6B only catalyses a single demethylation of asymmetric dimethylarginine to the monomethylated species. Within our limits of detection, no activity was observed for KDM6B and the tested peptides containing monomethylarginine or symmetric dimethylarginine (Supplementary Fig. 6a). Interestingly, when the KDM6B substrate was extended from 15 to 21 residues, a double demethylation reaction to the unmethylated arginine residue was observed in the MALDI MS of reactions containing KDM6B (Supplementary Fig. 6b). As expected for a 2OG oxygenase-catalysed reaction, demethylation by KDM6B required the presence of the Fe(II) cofactor and 2OG co-substrate (Fig. 3a).
In order to validate the MS-based results, we used NMR spectroscopy to follow KDM6B-catalysed demethylation of the 15- and 21-residue H3K27Rme2a peptides. Immediately after initiation of the reaction, the samples containing KDM6B, with either the 15- or 21-residue peptide substrate and cofactors/co-substrates, were transferred to NMR tubes and then subjected to 1H NMR (700 MHz) analyses (Fig. 3b and Supplementary Fig. 7a). In both samples, conversion of 2OG to succinate was evidenced by reduction of the 1H resonance at δH 2.4 p.p.m. (corresponding to the protons attached to C4 of 2OG) with concomitant emergence of a 1H resonance at δH 2.3 p.p.m. (corresponding to the methylene protons of succinate). An additional singlet resonance at δH 2.8 p.p.m. was also observed to increase in intensity during the analyses; this was assigned to the N-methyl protons of the monomethylarginine product based on 1H and 13C chemical shift analyses (δC 27.5 p.p.m., as determined by 1H-13C-HSQC analysis of the sample with the 15-residue peptide, Fig. 3b and Supplementary Fig. 7b). Quantitative analyses on the sample with the 21-residue peptide revealed a faster rate of succinate production relative to the rate of peptide demethylation, indicating some uncoupling of 2OG and peptide oxidation (that is, some conversion of 2OG to succinate without concomitant oxidation of the peptide methyl group). The degree of 2OG oxidation was greater in the presence of peptide than without peptide, suggesting that binding of the peptide to KDM6B stimulates 2OG oxidation but does not always necessarily lead to methyl group oxidation (Fig. 3c and Supplementary Fig. 7c). This ‘uncoupling’ of 2OG oxidation and peptide demethylation is well precedented for 2OG oxygenases, although it is less pronounced during KDM4E-catalysed demethylation of methyllysine peptides22. Addition of the formaldehyde-trapping reagent dimedone to the NMR reaction mixture in the presence of the 15 residue peptide resulted in the emergence of new 1H resonances, which were assigned to two dimedone–formaldehyde adducts (Fig. 3d). These adducts have been previously observed to form upon incubation of dimedone with formaldehyde released during JmjC KDM-catalysed demethylation, and therefore imply that KDM6B-catalysed methylarginine demethylation follows a similar mechanism to that for methyllysine demethylation (that is, methyl group oxidation to release formaldehyde)22. The combined MS and NMR results thus reveal that some JmjC oxygenases can catalyse Nω-methyl arginine demethylation.
Preliminary kinetic parameters for selected methylarginine variant peptides were determined using a formaldehyde dehydrogenase-coupled demethylation assay (Table 1)23. In the case of KDM6B, the methylated arginine peptide probably binds more weakly (as judged by KM values) in the active site than its methylated lysine analogue (KM 676.3±118.3 versus 6.7±0.6 μM). Interestingly, however, the kcat value for the arginine peptide was higher than that for the analogous lysine peptide (kcat 28.8 × 10−3±3.6 × 10−3 versus 11.7 × 10−3±0.1 × 10−3s−1), suggesting that in the context of a tighter binding substrate (for example, a protein rather than the peptides tested here) it is possible that arginine demethylation by KDM6B is as efficient as, or favoured relative to, lysine demethylation. For KDM5C, the rate of asymmetric dimethylarginine demethylation was also found to be of similar levels to that of lysine demethylation (kcat 41.2±3.5 versus 67.1±10.3 s−1), while for KDM3A and KDM4E both the rate of demethylation and binding constants were found to be considerably lower for the arginine peptides than the analogous lysine peptides (Table 1). KDM4E appears to bind to the peptide containing symmetric arginine less well than the asymmetric residue (KM 101.1±11.0 versus 66.0±5.0 μM), but has a faster maximal turnover rate for the symmetric peptide (kcat 8.7 × 10−3±0.3 × 10−3 versus 4.6 × 10−3±0.7 × 10−3s−1), resulting in similar kcat/KM values. All kcat values were lower and KM were higher than those for the established H3K9me3 peptide substrate for KDM4E. The monomethylated H3K9Rme peptide had a higher KM value than both the dimethylated peptides, and displayed the lowest turnover rate (kcat 2.7 × 10−3±0.2 × 10−3s−1. Note that monomethylated lysine residues are poor substrates for the KDM4 subfamily20). For KDM5C, although both the asymmetric and symmetric dimethylarginine-containing peptides had similar KM values (KM 29.0±0.5 versus 21.1±5.8 μM), the turnover rate for the asymmetric peptide was an order of magnitude higher (41.2 × 10−3±3.5 × 10−3 versus 4.6 × 10−3±0.7 × 10−3s−1).
For KDM3A, the monomethylated arginine peptide was a better substrate than the corresponding dimethylated arginine peptides (Table 1). These results reflect the MS screening results, where in the context of overall incomplete substrate peptide demethylation, only conversion to the unmethylated residue was observed, that is, no monomethylated peptide intermediate was detected (Supplementary Fig. 3). Although the monomethylated arginine-methylated substrate showed a lower kcat value than the lysine-methylated substrate (11.0±1.2 versus 355±95.3 s−1), the KM values for all methylated arginine peptides were found to be smaller than those for H3K9me2 (H3K9Rme=12.1±3.1 μM, H3K9Rme2a=11.6±1.4 μM, H3K9Rme2s=23.7±1.9 μM and H3K9me2=44.5±4.8 μM). The kcat value for K9Rme2a and K9Rme2s were lower than that for K9Rme. The combined kinetic results described above for each enzyme imply that the relative efficiency of RDM versus KDM substrate binding and rate of catalysis is sequence and context-dependent.
RDM activity is observed on ‘natural’ histone peptides
Given the ability of JmjC enzymes to catalyse demethylation of arginine residues in synthetic peptide sequences, we were interested to investigate whether these enzymes could also catalyse arginine demethylation in ‘natural’ positions in histone peptides. Peptides representing known methylated arginine sites within the canonical histone tail sequences were synthesized and screened as substrates for the four enzymes identified as RDMs (Supplementary Table 1). Interestingly, arginine demethylation activity on the ‘natural’ histone peptide sequences was observed for some, but not all of the KDMs (Table 2 and Fig. 4). Both KDM4E and KDM5C catalysed demethylation of some of the screened peptides, including peptides methylated at H3R2 and H4R3 (Fig. 4a,b and Supplementary Figs 8 and 9), while KDM3A and KDM6B displayed no activity on any of the new peptides tested (Supplementary Figs 10 and 11). The observation of a lack of RDM activity for KDM6B with these substrates is notable, given that KDM6B displayed such high RDM activity with the unnatural H3K27Rme2a substrate peptide (Table 1). Thus, these results demonstrate that RDM activity is dependent on the sequence context.
Further kinetic characterization was carried out on some of the ‘biologically relevant’ peptides identified to be substrates for the KDMs (Table 1). Interestingly, despite the different neighbouring peptide sequences to the characterized lysine H3K9me3 substrate of KDM4E, the H3R2me2a peptide was found to be a better substrate for KDM4E than H3K9Rme2a or H3K9Rme2s (kcat/KM 114.0 × 10−6 versus 70.4 × 10−6μM−1s−1/85.9 × 10−6μM−1s−1). The H4R3me2a peptide was a less efficient substrate for KDM4E, with a turnover rate almost five times lower than that of the H3R2me2a peptide, similar to that observed for H3K9Rme. For KDM5C, the H3R2me2a peptide had a lower KM value than the H3K4Rme2a peptide, but was an overall slightly less efficient substrate for the enzyme (kcat/KM 1,420 × 10−6 versus 630 × 10−6μM−1s−1). Nevertheless, with a kcat/KM value only 14 times lower than that for H3K4me3 it appears possible that this reaction could also occur in cells in some contexts, particularly if promoted by auxiliary binding domains or binding partners (as is known to be important for some JmjC KDM activities (for example, PHF8))24.
Because some JmjC KDMs were found to catalyse demethylation of methylated arginine peptides in sequences substantially different from their characterized lysine substrates (for example, we observed low levels of demethylation of H4R3me2a (SG-Rme2a-GKGGKGLGKGGAK) by both KDM4E and KDM5C, Fig. 4), we were interested to see whether substitution of an arginine residue in the ‘natural’ histone sequence with a methylated lysine residue would produce a KDM substrate. Interestingly, despite a peptide containing H3R2me2a being a relatively good RDM substrate for both KDM4E and KDM5C, the corresponding H3R2Kme3 peptide (A-Kme3-TKQTARKSTGGKA) was not observed to be a KDM substrate for either enzyme within our limits of detection (Fig. 4c). Overall, the combined results reveal that, at least for some JmjC demethylases, the preferred sequence contexts can differ for RDM and KDM activities.
To investigate whether methylarginine demethylation can be catalysed by full-length JmjC KDMs, as well as by the truncated constructs, full-length proteins were exogenously overexpressed in HEK293T cells either as N-terminally Flag-tagged wild-type proteins or active-site variants known to be catalytically inactive25,26,27. The resultant full-length proteins were immunoprecipitated by their Flag-tags and used in MS assays with methylated histone peptides. As with the catalytic domain constructs, arginine demethylation was also observed with all four of the full-length proteins (KDM3A, KDM4A, KDM5C and KDM6B; note KDM4A was used in place of KDM4E as KDM4E is a pseudogene) tested on both ‘unnatural’ and ‘natural’ peptide substrates (Fig. 5 and Supplementary Figs 12–16). No RDM activity was observed with the catalytically inactive enzyme variants. Note that, with KDM3A, arginine demethylation activity was observed for the monomethylated peptides, but not for the symmetrically or asymmetrically methylated peptides; this result is consistent with the observed relative reaction efficiencies of these peptides with the truncated construct (Supplementary Fig 13 and Table 1).
JmjC enzymes bind arginine and lysine in a similar manner
Crystallographic studies were then undertaken to investigate the binding modes of methylarginine-containing peptides with JmjC enzymes. KDM4A, a close structural homologue of KDM4E that can also catalyse demethylation of methylarginine peptides (albeit less efficiently in our current assays, Fig. 5), was selected for the crystallographic work because of its relatively robust crystallization properties and because it has previously been crystallized with Nɛ-methylated lysine peptide substrates20,28,29,30. A suitable crystal was obtained using KDM4A and an H4R3me2s fragment sequence (residues 1–15) with nickel substituting for iron and N-oxalylglycine (NOG) acting as a 2OG mimetic). The structure was solved by molecular replacement using PDB ID: 2OX0 as the search model; two molecules of KDM4A (chains A and B) are present in the asymmetric unit. In the initial electron density maps, other than for the methylated arginine itself, only weak difference in electron densities were observed for residues in the substrate peptide, suggesting that the mode of methylated arginine peptide binding could be flexible. We generated an Fo–Fc OMIT map (where different densities were calculated omitting H4R3me2s) as a guide to locate any unmodelled difference peak around the KDM4A (A/B) active sites. This unweighted omit map revealed more than one likely conformation of the methylated arginine side chains and traces of H4R3me2s peptide backbones; only major conformations were modelled and refined in the final structure.
Although care should be taken in over-interpreting ‘static’ crystal structures of substrate fragments, it is of interest that the arginine side chain was observed to adopt two different extended conformations after refinement. One orientation positions one of the methyl groups of the symmetrically dimethylated arginine residue sufficiently close to the metal centre to enable catalysis (within 4.5 Å; Fig. 6a). The other side chain orientation is a likely catalytically unproductive conformation, wherein the side chain is oriented away from the metal (Fig. 6b; for overlay, see Fig. 6c). The presence of such an unproductive binding conformation might in part account for the higher degree of uncoupled 2OG turnover stimulated by methylated arginine compared with methylated lysine peptide binding, as observed for KDM6B with H3(14–34)K27Rme2a (for overlay of arginine and lysine binding, see Fig. 6d).
Potential for JmjC RDM activity on non-histone substrates
Arginine methylation has been reported for numerous non-histone proteins, as observed by antibody- and/or MS-based approaches, where it is proposed to regulate many cellular processes including alteration of cellular localization, protein stability and protein–protein/DNA interactions31,32. Although limited at present, there are some reports providing at least preliminary evidence for non-histone protein lysine demethylation catalysed by JmjC KDMs33,34,35. Given the ability of the JmjC KDMs to act as RDMs for peptides with sequences substantially different from their characterized lysine sequences, we were interested to investigate whether removal of non-histone arginine methylation could also be catalysed by 2OG oxygenases. As KDM6B is known to interact with tumour protein p53 (p53), peptide sequences containing known arginine methylation sites were synthesized and tested as JmjC RDM substrates36; however, no demethylation was observed (Fig. 7 and Supplementary Fig. 17). A selection of peptides from some other nuclear proteins known to be abundantly methylated on arginine residues were also synthesized and tested as RDM substrates by MS. While the majority of these peptides were found not to be substrates, in some cases clear arginine demethylation was observed (Fig. 7 and Supplementary Fig. 17). Demethylation of asymmetric dimethylated arginine residues within peptides from heterogeneous nuclear ribonucleoprotein K (hnRNP K) and tumour suppressor p53-binding protein 1 (53BP1) was observed, as catalysed by KDM4E and KDM5C, respectively. If replicable in cells, these results raise the possibility that arginine demethylation could be a widespread phenomenon extending well beyond the canonical histone tails.
Since the observation of Nɛ-lysine methylation in 1959 (ref. 37), multiple protein residues have been found to undergo N- and O-methylation31,38. Within the context of gene regulation, the N-methylation of lysine and arginine residues on histone tails plays critical roles. Multiple methyltransferases have been identified, many of which are apparently highly promiscuous, acting to catalyse methylation on numerous proteins31,33,39,40. Despite the pioneering observations in 1973 of the oxidative demethylation of Nɛ-methylated lysine residues in calf thymus histones41, the identification of the sequences of Nɛ-methyl-group demethylases took three decades. Following the identification of KDM1A in 2004 (ref. 42), the subsequent discovery of the JmjC KDMs validated the early studies in identifying oxygenases that catalyse Nɛ-methyllysine demethylation to release formaldehyde4,43,44. Given the plethora of methylated proteins, it appears possible that some KDMs might be found to be equally promiscuous as KMTs.
The possibility of RDM activities related to those of the KDMs is readily envisaged. While no lysine-specific demethylase-type flavin-dependent RDMs have been reported (although are feasible), the JmjC oxygenase JMJD6 has been reported as an RDM acting on histones and other non-histone proteins9,15,16. However, in our laboratory’s work with JMJD6 we have observed only lysine hydroxylation, not RDM activity10,45. Various reports have appeared in the recent literature describing both hydroxylation and RDM activities for JMJD6 (refs 13, 14, 15, 16, 17, 46). Other JmjC oxygenases have also been the subject of controversial assignments47,48,49. In most cases the cell-based work has principally employed antibody-based methods, in some cases supported by ‘proteomic’ MS evidence, and at present biochemical evidence for the RDM activity of JMJD6 remains limited. Identification of both RDM and KDM activities in a cellular context using antibodies can be complicated by difficulties in accurately quantifying absolute levels of both methylated substrates and the demethylated products and (along with other factors) by a lack of appropriately specific antibodies for substrates and products50. MS-based methods, while rapidly improving, are also not normally sufficiently well optimized for measuring absolute levels in cells, particularly for low abundance post-translational modifications32,50,51,52. Thus, while we appreciate that enzyme activity may differ in the isolated protein form from cellular contexts (for example, as observed for the 2OG-dependent protein hydroxylase factor inhibiting hypoxia-inducible factor53), given the complexity and controversy in JmjC assignments, in particular, in the arginine demethylation field, we believe that, where possible, it is important to define the ‘biochemical reactions’ catalysed by potential RDMs (and KDMs) using isolated enzymes.
The assays applied here directly measure, either by mass shifts or by the appearance of new NMR signals, the product(s) of KDM/RDM activity. It should be noted that in the context of MS analyses we undertook to carry out careful non-enzyme controls to ensure that the conditions of MS analysis did not invoke apparent demethylation, or that we were not observing contamination of the putative product in the starting material, which might not be apparent before incubation.
The results clearly show that, at least under our assay conditions, some, but not all, recombinant forms of JmjC KDMs do have RDM activity. Thus, we observed that KDM3A, KDM4E, KDM5C and KDM6B can catalyse demethylation of methylated arginine residues when substituted for the methylated lysine residue in their characterized peptide substrates. In addition, KDM4E (and full-length KDM4A) and KDM5C were found to catalyse demethylation of histone peptides methylated at H3R2, H3R8, H3R26 (KDM4E, but not KDM5C) and H4R3, and also some non-histone peptide sequences.
One interesting result to emerge from the studies on both histone and non-histone RDM and KDM substrates for the JmjC enzymes is that the apparent sequence selectivities can differ for RDM and KDM activities, even with the same enzyme. Thus, for example, H3R2me2a is a better substrate for KDM4E than H3K9Rme2a (Table 1), and the analogous H3R2Kme3 peptide is not a substrate for KDM4E or KDM5C (Fig. 4c). Although the biological significance of these results is presently unclear, they can be rationalized by the likelihood that in order for productive demethylation to occur, the methylarginine and methyllysine side chains are likely accommodated differently by dual-function KDM/RDM enzymes in order to ensure a catalytically productive relationship between the reactive Fe(IV)=O intermediate and the methyl group at the active site. The differential orientation of the lysine and arginine side chains may in turn have ‘knock-on’ effects with respect to optimal binding of the surrounding substrate residues. Support for this proposal comes from work revealing that in the case of KDM4A-C, aside from the methylated lysine residues, the H3K9 and H3K36 substrates are accommodated differently at the entrance to the active site30,54. Given that for some KDMs post-translational modifications distinct from the residue undergoing demethylation can substantially promote activity, it is possible that RDM activity is similarly promoted. The partial decoupling of 2OG oxidation and demethylation observed for some RDM substrates may reflect non-optimal substrate sequences, or use of a truncated substrate, as precedented with other 2OG-dependent protein hydroxylases55. However, it is important to note that, at least in the case of KDM6B, some of the RDM activities we observe are of similar levels to the KDM activity, implying the likelihood of biological relevance of the dual-function KDM/RDM results in some situations (including by competition between substrates).
We appreciate that further work is required on protein substrates in cells for the full biological relevance of these results to be established. Despite considerable efforts, we are not yet able to conclusively determine RDM activity in cells, in part because of the lack of suitably useful antibodies. Nonetheless, the biochemical assignment of RDM activity is interesting and relevant to future work on the functional studies on JmjC enzymes in cells. The JmjC KDMs have been linked to a plethora of human developmental diseases and cancers, and there is growing interest in them as pharmaceutical targets56,57,58. In addition to helping to elucidate the regulatory mechanisms involved in arginine methylation cycles, these results reveal new enzymatic activity and function of the JmjC KDMs, an understanding of which may be crucial to their effective therapeutic targeting. In this regard, it is important to note that, at least some of the roles of JmjC KDMs in disease may involve their production of formaldehyde, which is a product of both KDM and RDM catalysis, and which, at least above threshold levels, is toxic59. Now that enzymes that have RDM activity have been conclusively identified, efforts can focus on the demonstration of their biological relevance.
Full-length Flag-tagged KDMs were expressed in HEK293T cells (ATCC) using plasmids encoding full-length KDM4A1–1,064 (ref. 26), H188A KDM4A1–1,064 (ref. 26), KDM6B1–1,636 (ref. 27), H1390A KDM6B1–1,636 (ref. 27), KDM3A1–1,321 (ref. 25) and H1120Y KDM3A1–1,321 (ref. 25) as described previously. Plasmids encoding Flag-tagged KDM5C1–4,680 and H513A/E515A KDM5C1–4,680 were kindly provided by Pavel Savitsky (Structural Genomics Consortium, Oxford). Recombinant constructs of KDM2A1–557 (ref. 60), KDM4A1–359 (ref. 30), KDM4E1–337 (ref. 61), KDM6B1,141–1,590 (ref. 62) and PHF81–447 (ref. 63) were produced as N-terminally His6-tagged proteins in Escherichia coli cells. KDM3A515–1,317 was produced as an N-terminally His6-tagged protein in Sf9 cells62. KDM5C1–765 was produced as an N-terminally His6-tagged protein in Sf9 cells and was kindly provided by Aleksandra Szykowska (Structural Genomics Consortium). Proteins were purified by Ni-affinity chromatography and size exclusion chromatography as previously described.
Peptides were produced in house using a Liberty Blue automated microwave peptide synthesizer (CEM Corporation) for kinetic and crystallographic studies or either a Multipep RSi (Intavis Bioanalytical Instruments AG) or an S336X peptide synthesizer (CS Bio) for screening purposes. Fluorenylmethyloxycarbonyl-mediated solid-phase chemistry (on MBHA resin) was used to produce the peptides with C-terminal amides. After synthesis, peptides were cleaved from the resin by incubation with 92.5% trifluoroacetic acid/2.5% water/2.5% triisopropylsilane/2.5% dimethoxybenzene (3 h) followed by precipitation with ice-cold diethyl ether. Lyophilized peptides produced on the Liberty Blue or S336X synthesizers were purified by reverse-phase high-performance liquid chromatography using a Vydac C18 column (Solvent A=0.1% trifluoroacetic acid in H2O, Solvent B=0.1% trifluoroacetic acid in acetonitrile) to >95% purity as determined by MS. Peptides used for screening were assayed unpurified. Peptide sequences are given in Supplementary Table 1. Methylated arginines suitably protected for solid-phase peptide synthesis were purchased from Novabiochem. In all cases the observed masses for peptides were within 2 Da of the predicted values by MALDI MS.
MALDI MS activity assays. Recombinant proteins (2 μM) were incubated with peptides (10 μM, sequences are given in Supplementary Table 1) and cofactors/substrates for 1 h at 37 °C, unless otherwise specified. Reactions were quenched with 1:1 (v/v) methanol and analysed with MALDI MS. Assay conditions for each enzyme are given in Supplementary Table 2. Negative controls without enzyme were included. Note that in some cases small variations (≤2 Da) in the absolute experimentally observed values compared with the calculated values for peptides were observed. However, in all cases where demethylation is assigned the mass shift was 14 Da.
Formaldehyde dehydrogenase-coupled assays. Preliminary kinetic parameters were determined by detection of the formaldehyde by-product of the demethylation reaction using a FDH (formaldehyde dehydrogenase-coupled)/NAD+-coupled assay as previously described61. Assays were carried out in buffer (enzyme-dependent) with the addition of Fe(II) ammonium sulfate (10 μM), sodium ascorbate (100 μM), 2OG (200 μM), NAD+ (500 μM), enzyme (0.2–4 μM) and FDH enzyme (0.025 U per assay, Sigma-Aldrich; total volume 30 μl) and were conducted at 37 °C in clear-bottom black 384-well plates. Reactions were monitored over 25 min using a PHERAstar FS (BMG Labtech) plate reader with 355 nm excitation and 460 nm emission. The Km and Vmax values were calculated from the reaction rate during the linear phase of formaldehyde production, as determined by the slope of the fluorescence signal recorded.
NMR assays. NMR spectroscopy was carried out as described22. NMR spectra were recorded using a Bruker Avance AVIII 700 MHz spectrometer equipped with an inverse TCI cryoprobe, optimized for 1H observation and installed with the Topspin software. All samples were prepared in Eppendorf tubes (75 μl volume) before being transferred to 2-mm MATCH NMR tubes (Wilgenberg); time course data were then collected over a period of 20 min at 89-s intervals using an automated routine. The solvent deuterium signal was used as an internal lock signal, and the solvent signal was reduced by excitation sculpting64. Enzyme stocks were in protiated HEPES/NaCl buffer, pH 7.5, which was diluted with dAFN buffer (prepared by diluting 500 mM ammonium formate, 5 M NaCl in H20 pH 7.5 to 10% v/v in D2O) when added to the samples. Chemical shifts are reported relative to the solvent water resonance (δH 4.7 p.p.m.).
HEK293T cells (ATCC) were cultured at 37 °C and 5% CO2 in high-glucose DMEM (Sigma) supplemented with 10% fetal bovine serum (Sigma) and 1% Glutamax (Sigma). Full-length KDMs with Flag-tags were transiently transfected into HEK293T cells using linear polyethylenimine (PEI) (10 μg DNA, 60 μl PEI per T75, 60% confluency, PAA Laboratories) and were harvested after 24–48 h.
On-bead demethylation assays
Cells transiently transfected with the relevant KDM were washed in PBS buffer and lysed in 50 mM HEPES, pH 7.5, 150 mM NaCl, 0.5% NP40, 1% EDTA-free protease inhibitor cocktail and DNAse (10 μg ml−1). Lysate was diluted fivefold with buffer containing 50 mM HEPES, pH 7.5, 150 mM NaCl before being incubated with 20 μl anti-FLAG M2 Magnetic Beads (Sigma-Aldrich) for 3 h at 4 °C. Beads were washed four times with lysis buffer (500 μl) and two times with 50 mM HEPES (500 μl; or 50 mM HEPES, 150 mM NaCl where NaCl was used in the reaction buffer) before MS analysis. Beads were resuspended in a final volume of 30 μl for use in activity assays.
Demethylation reactions were conducted with 2 μl of the immobilized anti-FLAG bead suspension with peptide (10 μM) and cofactors/substrates (conditions given in Supplementary Table 3). Reactions were incubated for 3 h at 37 °C and quenched with 1:1 MeOH by volume. The demethylation products were detected by MALDI MS. Western blotting of lysate samples was carried out to confirm exogenous overexpression of all enzymes using mouse anti-Flag primary (F3165 Sigma, 1:1,000, overnight, 4 °C) and goat anti-mouse-HRP (W4021 Promega, 1:10,000, 1 h, room temperature, Supplementary Fig. 12).
Co-crystals of KDM4A1–359.Ni(II).H4(1–15)R3me2s.NOG were obtained in sitting drops grown at 4 °C. Drops contained 10 μg ml−1 KDM4A, 5 mM H4(1–15)R3me2s, 5 mM NOG and 4 mM NiCl2. Crystals were cryoprotected with 25% glycerol before being flash-frozen in liquid N2. Data were collected using a single crystal at 100 K at the Diamond I04-1 MX beam line and were processed with HKL2000 (ref. 65). The structure was solved by molecular replacement using PHASER66 (search model PDB ID 2OX0) and was refined by alternative cycles of CNS67 and PHENIX68, with iterative rebuilding of the refined model using COOT69. All residues were in the allowed regions of the Ramachandran plot as calculated by PROCHECK70. Data collection and refinement statistics are given in Supplementary Table 4. A stereo image of a portion of the electron density map is shown in Supplementary Fig. 18.
Each result shows the mean of three independent experiments with error bars representing the s.e.m. Each experiment was carried out (at least) in technical triplicate.
The data that support the findings of this study are available from the corresponding author upon request. The crystal structure has been deposited under PDB accession code 5FWE.
How to cite this article: Walport, L. J. et al. Arginine demethylation is catalysed by a subset of JmjC histone lysine demethylases. Nat. Commun. 7:11974 doi: 10.1038/ncomms11974 (2016).
Walsh, C. T., Garneau-Tsodikova, S. & Gatto, G. J. Jr Protein posttranslational modifications: the chemistry of proteome diversifications. Angew. Chem. Int. Ed. Engl. 44, 7342–7372 (2005).
Rothbart, S. B. & Strahl, B. D. Interpreting the language of histone and DNA modifications. Biochim. Biophys. Acta 1839, 627–643 (2014).
Le, D. D. & Fujimori, D. G. Protein and nucleic acid methylating enzymes: mechanisms and regulation. Curr. Opin. Chem. Biol. 16, 507–515 (2012).
Walport, L. J., Hopkinson, R. J. & Schofield, C. J. Mechanisms of human histone and nucleic acid demethylases. Curr. Opin. Chem. Biol. 16, 525–534 (2012).
Di Lorenzo, A. & Bedford, M. T. Histone arginine methylation. FEBS Lett. 585, 2024–2031 (2011).
Zhang, X. et al. Peptidylarginine deiminase 2-catalyzed histone H3 arginine 26 citrullination facilitates estrogen receptor alpha target gene activation. Proc. Natl Acad. Sci. USA 109, 13331–13336 (2012).
Wang, Y. et al. Human PAD4 regulates histone arginine methylation levels via demethylimination. Science 306, 279–283 (2004).
Thompson, P. R. & Fast, W. Histone citrullination by protein arginine deiminase: is arginine methylation a green light or a roadblock? ACS Chem. Biol. 1, 433–441 (2006).
Chang, B., Chen, Y., Zhao, Y. & Bruick, R. K. JMJD6 is a histone arginine demethylase. Science 318, 444–447 (2007).
Webby, C. J. et al. Jmjd6 catalyses lysyl-hydroxylation of U2AF65, a protein associated with RNA splicing. Science 325, 90–93 (2009).
Han, G. et al. The hydroxylation activity of Jmjd6 is required for its homo-oligomerization. J. Cell. Biochem. 113, 1663–1670 (2012).
Mantri, M. et al. The 2-Oxoglutarate-dependent oxygenase JMJD6 catalyses oxidation of lysine residues to give 5S-hydroxylysine residues. ChemBioChem 12, 531–534 (2011).
Gao, W. W. et al. Arginine methylation of HSP70 regulates retinoid acid-mediated RARbeta2 gene activation. Proc. Natl Acad. Sci. USA 112, E3327–E3336 (2015).
Lawrence, P., Conderino, J. S. & Rieder, E. Redistribution of demethylated RNA helicase A during foot-and-mouth disease virus infection: role of Jumonji C-domain containing protein 6 in RHA demethylation. Virology 452-453, 1–11 (2014).
Poulard, C., Rambaud, J., Hussein, N., Corbo, L. & Le Romancer, M. JMJD6 regulates ERalpha methylation on arginine. PLoS ONE 9, e87982 (2014).
Liu, W. et al. Brd4 and JMJD6-associated anti-pause enhancers in regulation of transcriptional pause release. Cell 155, 1581–1595 (2013).
Bottger, A., Islam, M. S., Chowdhury, R., Schofield, C. J. & Wolf, A. The oxygenase Jmjd6--a case study in conflicting assignments. Biochem. J. 468, 191–202 (2015).
Markolovic, S., Wilkins, S. E. & Schofield, C. J. Protein hydroxylation catalyzed by 2-oxoglutarate-dependent oxygenases. J. Biol. Chem. 290, 20712–20722 (2015).
Hopkinson, R. J. et al. Is JmjC oxygenase catalysis limited to demethylation? Angew. Chem. Int. Ed. Engl. 52, 7709–7713 (2013).
Williams, S. T. et al. Studies on the catalytic domains of multiple JmjC oxygenases using peptide substrates. Epigenetics 9, 1596–1603 (2014).
Klose, R. J., Kallin, E. M. & Zhang, Y. JmjC-domain-containing proteins and histone demethylation. Nat. Rev. Genet. 7, 715–727 (2006).
Hopkinson, R. J., Hamed, R. B., Rose, N. R., Claridge, T. D. & Schofield, C. J. Monitoring the activity of 2-oxoglutarate dependent histone demethylases by NMR spectroscopy: direct observation of formaldehyde. ChemBioChem 11, 506–510 (2010).
Sakurai, M. et al. A miniaturized screen for inhibitors of Jumonji histone demethylases. Mol. Biosyst. 6, 357–364 (2010).
Feng, W., Yonezawa, M., Jung, Y., Jenuwein, T. & Grummt, I. PHF8 activates transcription of rRNA genes through H3K4me3 binding and H3K9me1/2 demethylation. Nat. Struct. Mol. Biol. 17, 445–450 (2010).
Yamane, K. et al. JHDM2A, a JmjC-Containing H3K9 Demethylase, FacilitatesTranscription Activation by Androgen Receptor. Cell 125, 483–495 (2006).
Hopkinson, R. J. et al. 5-Carboxy-8-hydroxyquinoline is a broad spectrum 2-oxoglutarate oxygenase inhibitor which causes iron translocation. Chemical Science 4, 3110–3117 (2013).
Kruidenier, L. et al. A selective jumonji H3K27 demethylase inhibitor modulates the proinflammatory macrophage response. Nature 488, 404–408 (2012).
Couture, J. F., Collazo, E., Ortiz-Tello, P. A., Brunzelle, J. S. & Trievel, R. C. Specificity and mechanism of JMJD2A, a trimethyllysine-specific histone demethylase. Nat. Struct. Mol. Biol. 14, 689–695 (2007).
Chen, Z. et al. Structural basis of the recognition of a methylated histone tail by JMJD2A. Proc. Natl Acad. Sci. USA 104, 10818–10823 (2007).
Ng, S. S. et al. Crystal structures of histone demethylase JMJD2A reveal basis for substrate specificity. Nature 448, 87–91 (2007).
Bedford, M. T. & Clarke, S. G. Protein arginine methylation in mammals: who, what, and why. Mol. Cell 33, 1–13 (2009).
Plank, M. et al. Expanding the yeast protein arginine methylome. Proteomics 15, 3232–3243 (2015).
Zhang, X., Wen, H. & Shi, X. Lysine methylation: beyond histones. Acta Biochim. Biophys. Sin. 44, 14–27 (2012).
Ponnaluri, V. K., Vavilala, D. T., Putty, S., Gutheil, W. G. & Mukherji, M. Identification of non-histone substrates for JMJD2A-C histone demethylases. Biochem. Biophys. Res. Commun. 390, 280–284 (2009).
Lu, T. et al. Regulation of NF-kappaB by NSD1/FBXL11-dependent reversible lysine methylation of p65. Proc. Natl Acad. Sci. USA 107, 46–51 (2010).
Sola, S. et al. p53 interaction with JMJD3 results in its nuclear distribution during mouse neural stem cell differentiation. PLoS ONE 6, e18421 (2011).
Ambler, R. P. & Rees, M. W. Epsilon-N-Methyl-lysine in bacterial flagellar protein. Nature 184, 56–57 (1959).
Lanouette, S., Mongeon, V., Figeys, D. & Couture, J. F. The functional diversity of protein lysine methylation. Mol. Syst. Biol. 10, 724 (2014).
Moore, K. E. & Gozani, O. An unexpected journey: lysine methylation across the proteome. Biochim. Biophys. Acta 1839, 1395–1403 (2014).
Kudithipudi, S., Kusevic, D., Weirich, S. & Jeltsch, A. Specificity analysis of protein lysine methyltransferases using SPOT peptide arrays. J. Vis. Exp. e52203 (2014).
Paik, W. K. & Kim, S. Enzymatic demethylation of calf thymus histones. Biochem. Biophys. Res. Commun. 51, 781–788 (1973).
Shi, Y. et al. Histone demethylation mediated by the nuclear amine oxidase homolog LSD1. Cell 119, 941–953 (2004).
Whetstine, J. R. et al. Reversal of histone lysine trimethylation by the JMJD2 family of histone demethylases. Cell 125, 467–481 (2006).
Tsukada, Y. et al. Histone demethylation by a family of JmjC domain-containing proteins. Nature 439, 811–816 (2006).
Hahn, P. et al. Analysis of Jmjd6 cellular localization and testing for its involvement in histone demethylation. PLoS ONE 5, e13769 (2010).
Unoki, M. et al. Lysyl 5-hydroxylation, a novel histone modification, by Jumonji domain containing 6 (JMJD6). J. Biol. Chem. 288, 6053–6062 (2013).
Sinha, K. M., Yasuda, H., Coombes, M. M., Dent, S. Y. R. & de Crombrugghe, B. Regulation of the osteoblast‐specific transcription factor Osterix by NO66, a Jumonji family histone demethylase. EMBO J. 29, 68–79 (2010).
Ishimura, A. et al. Jmjd5, an H3K36 histone demethylase, modulates embryonic cell proliferation through the regulation of Cdkn1a expression. Development 139, 749–759 (2012).
Liu, L. et al. Hairless is a histone H3K9 demethylase. FASEB J. 28, 1534–1542 (2014).
Carlson, S. M. & Gozani, O. Emerging technologies to map the protein methylome. J. Mol. Biol. 426, 3350–3362 (2014).
Guo, A. et al. Immunoaffinity enrichment and mass spectrometry analysis of protein methylation. Mol. Cell. Proteomics 13, 372–387 (2014).
Bremang, M. et al. Mass spectrometry-based identification and characterisation of lysine and arginine methylation in the human proteome. Mol. Biosyst. 9, 2231–2247 (2013).
Yang, M. et al. Asparagine and aspartate hydroxylation of the cytoskeletal ankyrin family is catalyzed by factor-inhibiting hypoxia-inducible factor. J. Biol. Chem. 286, 7648–7660 (2011).
Hillringhaus, L. et al. Structural and evolutionary basis for the dual substrate electivity of the human KDM4 histone demethylase family. J. Biol. Chem. 286, 41616–41625 (2011).
Ehrismann, D. et al. Studies on the activity of the hypoxia-inducible-factor hydroxylases using an oxygen consumption assay. Biochem. J. 401, 227–234 (2007).
Johansson, C. et al. The roles of Jumonji-type oxygenases in human disease. Epigenomics 6, 89–120 (2014).
Greer, E. L. & Shi, Y. Histone methylation: a dynamic mark in health, disease and inheritance. Nat. Rev. Genet. 13, 343–357 (2012).
Helin, K. & Dhanak, D. Chromatin proteins and modifications as drug targets. Nature 502, 480–488 (2013).
Bosetti, C., McLaughlin, C., Tarone, R. E., Pira, E. & LaVecchia, C. Formaldehyde and cancer risk: a quantitative review of cohort studies through 2006. Ann. Oncol. 19, 29–43 (2008).
Rotili, D. et al. A photoreactive small-molecule probe for 2-oxoglutarate oxygenases. Chem. Biol. 18, 642–654 (2011).
Rose, N. R. et al. Inhibitor scaffolds for 2-oxoglutarate-dependent histone lysine demethylases. J. Med. Chem. 51, 7053–7056 (2008).
Rose, N. R. et al. Plant growth regulator daminozide is a selective inhibitor of human KDM2/7 histone demethylases. J. Med. Chem. 55, 6639–6643 (2012).
Loenarz, C. et al. PHF8, a gene associated with cleft lip/palate and mental retardation, encodes for an Nepsilon-dimethyl lysine demethylase. Hum. Mol. Genet. 19, 217–222 (2010).
Hwang, T. L. & Shaka, A. J. Water suppression that works. excitation sculpting using arbitrary wave-forms and pulsed-field gradients. J. Magn. Reson. 112, 275–279 (1995).
Otwinowski, Z. & Minor, W. Jr in Methods in Enzymology (eds. Carter Jr C. W., Sweet R. M. Part A, 307–326Academic Press (1997).
McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).
Brunger, A. T. et al. Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921 (1998).
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010).
Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132 (2004).
Laskowski, R. A., MacArthur, M. W., Moss, D. S. & Thornton, J. M. PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291 (1993).
This work was supported by grants from Cancer Research UK (C8717/A18245) and the Wellcome Trust (091857/7/10/7), the Biotechnology and Biological Sciences Research Council (studentships to L.J.W. and R.J.H.) and the Engineering and Physical Sciences Research Council (EP/L003376/1). R.J.H. acknowledges a William R. Miller Junior Research Fellowship, St Edmund Hall. A.K. acknowledges a Royal Society Dorothy Hodgkin Fellowship (DH120028) and the BHF Centre of Research Excellence Oxford (RE/13/1/30181). We thank Tristan Smart for purification of KDM2A, the Structural Genomics Consortium (Aleksandra Szykowska) for providing KDM5C protein and the Flag-tagged full-length KDM5C plasmids (Pavel Savitsky), Sarah Madden for KDM4A protein, Matthew Smith for assistance with the NMR experiments and Alexandra Brooks for the H3(14–34)K27Rme2a peptide.
The authors declare no competing financial interests.
About this article
Cite this article
Walport, L., Hopkinson, R., Chowdhury, R. et al. Arginine demethylation is catalysed by a subset of JmjC histone lysine demethylases. Nat Commun 7, 11974 (2016). https://doi.org/10.1038/ncomms11974
This article is cited by
The phosphorylation to acetylation/methylation cascade in transcriptional regulation: how kinases regulate transcriptional activities of DNA/histone-modifying enzymes
Cell & Bioscience (2022)
Communications Chemistry (2022)
Scientific Reports (2022)
Nature Communications (2022)
Signal Transduction and Targeted Therapy (2022)