Posttranslational modifications (PTMs) of histone proteins, such as acetylation, methylation, phosphorylation, and ubiquitylation, play essential roles in regulating chromatin dynamics. Combinations of different modifications on the histone proteins, termed ‘histone code’ in many cases, extend the information potential of the genetic code by regulating DNA at the epigenetic level. Many PTMs occur on non-histone proteins as well as histones, regulating protein–protein interactions, stability, localization, and/or enzymatic activities of proteins involved in diverse cellular processes. Although protein phosphorylation, ubiquitylation, and acetylation have been extensively studied, only a few proteins other than histones have been reported that can be modified by lysine methylation. This review summarizes the current progress on lysine methylation of non-histone proteins, and we propose that lysine methylation, like phosphorylation and acetylation, is a common PTM that regulates proteins in diverse cellular processes.
Eukaryotic DNA is packaged by histone proteins into the higher-order structure of chromatin. The basic unit of chromatin is the nucleosome, which contains ∼146 bp of DNA wound around an octamer of four core histones: H2A, H2B, H3, and H4A. A key molecular mechanism for regulating accessibility of the underlying DNA—the genetic information—is achieved via covalent posttranslational modifications (PTMs) of histone proteins, especially on their N-terminal unstructured tails. Examples of such modifications are acetylation, methylation, phosphorylation, and ubiquitylation. Histone acetylation is, in general, associated with gene activation. In contrast, methylation on a specific lysine (K) residue is correlated with either an active or a silent state of gene expression, depending on the residues being methylated. Moreover, each methylated lysine residue can exist in a mono-, di-, or tri-methylated state, further extending the indexing potential of this modification. In general, methylation on histone H3 lysine 4 (H3K4), H3K36, and H3K79 is linked to active gene expression, whereas di- and tri-methylation on H3K9, H3K27, and H4K20 are associated with gene silencing. Histone lysine methylation has also been shown to modulate other chromatin-related processes such as replication, recombination, and DNA repair.
Dynamic PTMs serve as a means to regulate protein–protein interactions, protein stability, protein localization, and/or enzymatic activities. Many PTMs, such as phosphorylation, ubiquitylation, and acetylation, occur on numerous proteins involved in diverse cellular processes. Although ∼80 enzymes have been shown to dynamically regulate histone lysine methylation, only a few non-histone proteins have been reported as substrates of these enzymes. In the past few years, more attention has been drawn to methylation of non-histone proteins. In this review, we briefly review all the enzymes known to regulate histone lysine methylation and summarize the current knowledge of lysine methylation of non-histone proteins and its role in regulating proteins involved in diverse cellular processes.
Modifying Enzymes for Lysine Methylation
The enzymes involved in lysine methylation were first found to target histone and thus were initially named histone methyltransferases and histone demethylases, following the naming model for histone acetyltransferase and histone deacetylase. With accumulating evidence that these modifications are not histone specific, a new nomenclature has been advocated for more generic names for these enzymes . The enzymes that add or remove the methylation mark on lysine residues are now named lysine methyltransferases (KMTs) and lysine demethylases (KDMs), respectively, for their broad or potentially broad spectra of protein substrates. We follow the new nomenclature throughout this review.
KMTs catalyze mono-, di-, or tri-methylation by transfering one, two, or three methyl groups, respectively, from S-adenosyl-L-methionine to the ε-amino group of a lysine residue. Except for KMT4/DOT1L, all known KMTs contain a conserved SET (Su(var)3-9, Enhancer of Zeste, Trithorax) domain harboring the enzymatic activity . Besides the SET domain, most KMTs also contain some other defined protein domain or homologous sequence that is used to classify KMTs into distinct subfamilies [1,3]. The enzymatic activities and substrate specificities of the eight KMT subfamilies are summarized here (Table 1).
aReference for non-histone substrates only.
Most KMT1 group proteins are members of the SUV39 family that specifically methylate H3K9. This family includes KMT1A/SUV39H1, KMT1B/SUV39H2, KMT1C/G9a, KMT1D/EHMT1/GLP, KMT1E/SETDB1, and KMT1F/SETDB2. The methylation activity of KMT1A–KMT1D requires both the SET domain and two adjacent regions called pre-SET and post-SET, whereas KMT1E and 1F have a large insertion within their SET domains, which are fully functional .
KMT1A and 1B are the first SET domain-containing KMTs identified . They are the dominant enzymes generating H3K9 trimethylation (H3K9me3) at pericentric heterochromatin, a highly compacted, transcriptionally silent chromatin domain . A structural characteristic of KMT1A and 1B is a SET domain-adjacent chromodomain that recognizes methylated histone lysines [34,35], whereas KMT1C and 1E contain an ankyrin repeat domain (ANK) that is also capable of binding to methylated histones . KMT1C and 1D specifically mono- and di-methylate H3K9 at euchromatin, and therefore are also known as euchromatic histone methyltransferases [33,37]. These two proteins form homomeric or heteromeric dimers in vivo via their SET domains, and the stability of the KMT1C protein depends on KMT1D. KMT1E and 1F trimethylate histone H3K9 and facilitate transcriptional repression [38,39]. They both contain a methyl-CpG-binding domain with affinity to methylated DNA, and KMT1E contains an extra Tudor domain, which also functions as a reader for methyl-lysines [40,41].
The KMT2 family includes Drosophila Trithorax homologs MLL family proteins (MLL1–MLL5, named KMT2A–KMT2E, respectively), two proteins with great similarity to yeast Set1 proteins SET1A/KMT2F and SET1B/KMT2G and ASH1/KMT2H protein. All KMT2 family proteins are H3K4 methyltransferases, and they share a distinct SET domain with an essential post-SET region at the C-terminus. KMT2A/MLL1 is a major regulator of hematopoiesis and embryonic development through regulation of HOX gene expression, and the MLL gene is frequently rearranged in human acute leukemias . The H3K4 trimethylation activity of KDM2A requires components of its core complex, including WDR5, RbBP5, and Ash2L, despite the isolated KMT2A SET domain having weak H3K4 mono-methylation activity . KMT2D/MLL4 is closely related to KMT2A in protein complex composition and target gene regulation [44,45]. KMT2B and 2C associate with H3K27 demethylase KDM6A/UTX, which also occupies the promoters of HOX gene clusters and regulates their transcriptional output .
The KMT3 family includes yeast Set2 homolog SETD2/KMT3A, nuclear receptor binding SET domain protein 1, NSD1/KMT3B, and SET and MYND domain-containing proteins SMYD2/KMT3C, SMYD1/KMT3D, and SMYD3/KMT3E. This group of proteins methylates mainly histone H3K36, but their KMT activities are not restricted to H3K36. KMT3A is a Huntingtin-interacting protein, and it is responsible for global transcription-dependent H3K36 trimethylation, a specific epigenetic mark for transcriptional activation enriched at the gene body [47,48]. KMT3B is involved in regulation of gene activation through its dimethylation activity at H3K36 , but it was also shown to methylate H4K20 . A recurrent translocation of KMT3B fused to nucleoporin-98 has been reported in childhood acute myeloid leukemia, indicating the oncogenic property of KMT3B [49,51]. KMT3B/NSD1 has two homologs, NSD2/MMSET/WHSC1 and NSD3/WHSC1L1, which are also associated with multiple cancers. Although NSD2 and NSD3 have not been classified in the KMT3 family, both were shown to possess strong H3K36 di-/tri-methylation activity in addition to their activities toward other histone lysine residues [52–54]. KMT3C has methylation activity toward both H3K4 and H3K36 [55,56], whereas KMT3D and KMT3E were shown to methylate H3K4 [57,58].
KMT4/DOT1L is a H3K79-specific methyltransferase and represents the only class of KMT without a SET domain . KMT4-mediated H3K79 di- and tri-methylation are evolutionarily conserved from yeast to human and ubiquitously correlated with active transcription . It is the sole enzyme responsible for H3K79 methylation, as knockout of Dot1 in yeast, flies, and mice results in complete loss of H3K79 methylation [61–63]. In mammals, KMT4 is an essential gene for embryogenesis, hematopoiesis, and cardiac function. Its H3K79 methylation activity is essential for MLL-AF4 and MLL-AF9 fusion-induced leukemias, suggesting that KMT4 could serve as a potential therapeutic target for MLL-rearranged leukemias [64–66]. Indeed, a recent study has demonstrated that a potent KMT4 inhibitor selectively kills cells bearing the MLL gene translocation in vitro, and leads to extension of survival in a mouse MLL xenograft model .
The KMT5 family consists of enzymes that methylate H4K20. KMT5A/SET8 specifically mono-methylates H4K20, which has been shown to associate with both gene repression and activation, depending on its chromatin context [68–71]. KMT5A and H4K20me1 are also essential for multiple other chromatin-associated processes, such as cell cycle progression, DNA replication, and DNA damage response [72–77]. KMT5B/SUV420H1 and its homolog KMT5C/SUV420H2 catalyze di- and tri-methylation of H4K20, which, along with H3K9me2/3, is essential for maintenance of repressive heterochromatin at pericentric and telomeric regions [78,79]. Dimethylation of H4K20 is recognized by the Tudor domain of 53BP1 in mammals and by its homolog Crb2 in Schizosaccharomyces pombe, and targets these proteins to DNA damage sites [80,81].
Two H3K27-specific KMTs, KMT6A/EZH2, and KMT6B/EZH1 belong to the KMT6 family. The methylation activity of KMT6A and KMT6B requires other core components of their associated protein complex, polycomb repressive complex 2 (PRC2) [82,83]. As the catalytic subunit of PRC2, KMT6A participates in maintaining the transcriptional repressive state of chromatin and is upregulated in a broad range of human cancers [84,85]. Rapidly dividing cells such as embryonic stem cells and cancer stem cells exhibit high levels of KMT6A expression and H3K27me3 . Interestingly, PTMs such as phosphorylation of enhancer of Zeste homolog 2 (EZH2) proteins appear to be essential mechanisms in regulating EZH2 enzymatic activity and transcriptional repression function [87–89]. KMT6B is present in a non-canonical PRC2 complex as an H3K27 methyltransferase, and is functionally redundant with EZH2 [90,91].
This family contains only one protein, SET7/9, which mono-methylates histone at H3K4 . KMT7 was found to methylate, besides histone, a number of non-histone proteins, including p53, DNA methyltransferase 1 (DNMT1), estrogen receptor alpha (ERα), nuclear factor kappaB (NFκB), and components of the TATA binding protein (TBP) complex, TBP-associated factors TAF10 and TAF7 . The consensus recognition sequence in substrates for SET7-mediated lysine methylation, K/R–S/T–K (target lysine is bold), is also recognized by the H3K4 demethylase KDM1A/LSD1, which is capable of removing the methyl mark on most of these substrates [17,94]. Therefore KMT7/KDM1A has emerged as a classic model for dynamic lysine methylation of both histone and non-histone proteins.
Currently this family comprises only one member, PRDM2/RIZ1. KMT8 was identified as retinoblastoma (RB) protein-interacting zinc-finger protein (RIZ1), and it belongs to the PRDM family of proteins which are characterized by the presence of a N-terminal positive regulatory (PR) domain (PRDI-BF1 and RIZ) . The PR domain is a homolog of the SET domain and shares 20%–30% identity with the SET module. Some PR domains show intrinsic methyltransferase activity, whereas the methylation activity of most other PRDM proteins has not been identified . KMT8 possesses H3K9 methylation activity and functions as corepressor for gene regulation . Many other PRDM proteins, although lacking methylation activity, also play a role in regulating chromatin dynamics during stem cell self-renewal, differentiation, and development .
Histone lysine methylation was regarded as enzymatically irreversible for decades until the recent discovery of the first histone KDM, LSD1/KDM1A . Soon after, Jumonji (JmjC) domain was identified as another module that possesses enzymatic activity in removing methyl groups from lysine residues . As a large number of proteins in the human genome contain the JmjC domain, numerous JmjC domain-containing KDMs were discovered in the past few years, which in turn provide novel insights into the mechanisms of histone modification and epigenetic regulation [100–102]. Like the KMTs, KDMs are classified according to the new nomenclature into several distinct groups based on their substrate specificities and protein domain organization (Table 2).
aReference for non-histone substrates only.
KDM1A/LSD1 is a flavin adenine dinucleotide-dependent monoamine oxidase that can remove mono- and di-methyl, but not tri-methyl, groups from methylated lysines such as H3K4 . Other core components of its protein complex, such as CoREST and BHC80, are essential for its demethylating activity on nucleosomes [105,106]. KDM1 has been shown to alter its substrates' specificity toward H3K9 methylation when associated with nuclear receptors, thereby acting as a transcriptional coactivator [107–109]. Furthermore, KMT1 has a broad spectrum of non-histone substrates, such as p53 and DNMT1 . As a homolog of KDM1A, KDM1B/LSD2 was recently identified and characterized as another H3K4 demethylase [111–113]. In contrast to KDM1A, which functions at promoters, KDM1B removes intragenic H3K4 methylation for gene activation . Like KDM1A, KDM1B was reported to be able to remove mono- and di-methylation at histone H3K9 .
KDM2A/JHDM1A is the first JmjC domain-containing histone demethylase identified . KDM2A and KDM2B demethylate mono- and di-methylation from H3K36, while KDM2B is also implicated in demethylation of H3K4 . This family of proteins has been shown to function as transcriptional corepressors for regulation of several tumor-associated genes, including c-Jun and p15Ink4b [116–119]. The F-Box present in KDM2 family members suggests that KDM2 proteins might also be involved in additional mechanisms of chromatin regulation, such as ubiquitylation .
Soon after the discovery of the JmjC domain as a signature demethylation motif, the KDM3 family of proteins was identified as the second family of JmjC histone demethylases (JHDM2). KDM3A/JHDM2A and KDM3B/JHDM2B have specific action toward mono- and di-methylation of H3K9, and they therefore function as transcriptional coactivators for gene expression. KDM3A and 3B have been shown to be involved in multiple biological processes such as androgen receptor (AR) signaling and spermatogenesis [121–124]. Another JHDM2 family member, JHDM2C/TRIP8, has not yet been shown to have enzymatic activity.
The KDM4 family encompasses four homologous demethylases, KDM4A–4D (JMJD2A–2D, respectively). KDM4 proteins are the first demethylases that show demethylation activity on trimethylation. All KDM4 family members are able to remove di- and tri-methylation from H3K9 and/or H3K36 [125–127]. Besides the JmjC domain, they share a highly conserved JmjN domain, and three of them (all except KDM4D) contain tandem PHD fingers and Tudor domains that read distinct histone methylation [128,129]. KDM4 family proteins function in hormone response. KDM4A and KDM4D were shown to be AR coactivators , and KDM4B demethylates H3K9me3 for transcription activation by ER [131,132]. Recent in vitro studies demonstrated that this family of proteins can also act on non-histone proteins [133,134].
KDM5 family members KDM5A–5D/JARID1A–1D specifically remove di- and tri-methylation from H3K4 [135–141]. They are multi domain-containing proteins characterized by a combination of JmjC and JmjN catalytic domains with an ARID DNA-binding domain, a C5HC2 zinc finger, and two to three PHD fingers. Distinct PHD fingers of KDM5A and KDM5C were shown to bind methylated H3K4 or H3K9, respectively [137,142]. This family of proteins shows oncogenic activity, as KDM5A is overexpressed in gastric cancer and KDM5B is highly induced in breast cancer. Inhibition of these proteins can suppress cell growth and/or tumorigenesis, indicating that these enzymes may have potential as therapeutic targets .
This family comprises two H3K27-specific histone demethylases, KDM6A/UTX and KDM6B/JMJD3, which are capable of removing di- and tri-methylation from H3K27 [46,144–148]. As H3K27 methylation is a repressive epigenetic mark elevated in multiple cancers, both KDM6A and KDM6B function as tumor suppressors, implicated in gene transcriptional activation, epigenetic reprogramming, and RB-dependent cell fate control [147,149,150]. KDM6A associates with the KMT2/MLL H3K4 methylation protein complexes in regulating HOX gene expression, and therefore is essential for proper development .
The KDM7/PHF2 family consists of three members: KDM7A/JHDM1D, KDM7B/PHF8, and KDM7C/PHF2, which are involved in regulation of the expression of ribosomal RNA and genes involved in X-linked mental retardation. This family of proteins possesses strong demethylation activity toward H3K9 and H3K27 mono- and di-methylation [151–157]. Furthermore, PHF8 is able to remove mono-methylation from H4K20 [154,155,158]. All three KDM7 family proteins contain a PHD finger that binds to histone H3K4me3, and this recognition is essential for their substrate specificity, genomic occupancy, and regulation of target gene expression [151,152,159–161].
So far, KDMs have been identified for nearly all known histone methyl-lysines, with the exception of residue H3K79. However, evidence suggests that H3K79 methylation is also reversible . It is possible that methylated H3K79 may be targeted by other yet undefined JmjC types of KDMs, or by novel KDMs different from KDM1 and the JmjC demethylases.
Lysine Methylation of Non-histone Proteins
Although lysine methylation of histones has been studied extensively as an essential mechanism for epigenetic regulation of chromatin, our understanding of this modification on non-histone proteins has just begun. Methylation of p53 protein by KMT7/SET7 is the first reported KMT-mediated methylation event on a non-histone protein . Since that discovery, several KMTs, as well as KDM1A, have been identified as p53 regulators that methylate or demethylate p53 protein, making p53 the most extensively studied non-histone protein undergoing lysine methylation [4,13,16,18,162]. Here we use p53 as a model to discuss the functions of lysine methylation of non-histone proteins. We also summarize the current knowledge of this modification on other proteins in diverse cellular processes (Table 1).
Lysine methylation of p53
The p53 gene is the most frequently mutated tumor suppressor gene in human cancers . One critical mechanism by which p53 protein is regulated upon genotoxic stresses is through PTMs, including ubiquitylation, phosphorylation, acetylation, methylation, sumoylation, and neddylation . Under unstressed growth conditions, p53 is tightly controlled at low level through ubiquitylation-dependent degradation, modulated mainly by MDM2 E3 ubiquitin ligase. Upon genotoxic stresses such as DNA damage, p53 proteins are quickly accumulated and activated in the setting of multiple PTMs such as phosphorylation and acetylation for full activation. Accumulating evidence suggests that lysine methylation of p53, especially on its C-terminus, also plays a key role in regulating protein stability, protein–protein interactions, and transactivation activity of p53 (Fig. 1; Table 1) [4,13,16,18,162].
Crosstalk between methylation and other PTMs of p53
Crosstalk (or cross-regulation) between histone PTMs provides a wealth of potential for epigenetic regulation of chromatin dynamics [165,166]. Accumulating evidence suggests that crosstalk between lysine methylation and other modifications is also a frequent mechanism of regulating non-histone proteins, such as p53. Reinberg and colleagues reported the first lysine methylation event on p53 in 2004 . They found that KMT7 specifically methylates p53 at K372 within the C-terminus regulatory region. This methylation increases upon DNA damage, stabilizes p53 proteins, and promotes p53 transactivation activity on its target genes . Subsequent studies suggest that the role of p53 K372 methylation is, at least in part, to stimulate p53 protein acetylation, as knockdown of KMT7 impairs K373/382 acetylation of p53 . In an independent study, Kurash and colleagues also showed that this methylation (K369 in mouse) is essential for p53 acetylation, and it may serve as an anchor site for a lysine acetyltransferase, Tip60, providing a molecular mechanism for the crosstalk between different PTMs (see discussion below) .
Acetylation at K382 serves as an activation mark for p53's transcriptional regulation activity. This acetylation is directly inhibited by methylation on the same residue by KMT5A/SET8, an enzyme previously reported to mono-methylate H4K20 . K382 methylation robustly suppresses p53-mediated transcription activation of highly responsive target genes, and depletion of KMT5A augments the proapoptotic and checkpoint activation functions of p53 upon DNA damage . It would be interesting to know whether p53K382 methylation works in concert with other inhibitory marks to directly counteract other p53 activation marks, such as K373 acetylation and K372 methylation.
The crosstalk between different PTMs in the vicinity of the p53 C-terminus seems a frequent mechanism for modulating p53 activities. In addition to acetylation, KMT7-mediated p53K372 methylation also interplays with methylation at nearby K residues. p53 can be methylated by another methyltransferase, KMT3C/SMYD2, at K370, two amino acids apart from the KMT7 methylation site . Knockdown of KMT3C enhances p53-dependent apoptosis upon DNA damage, indicating that this methylation is a repressive mark. Methylation at K372 and K370 is directly inhibitory of each other. Inhibition of p53K370 methylation by KMT7-mediated K372 methylation is thought to directly block the interaction between p53 and KMT3C . In contrast, however, p53 peptide with K370 methylation is still accessible to KMT7 .
That KMT7-mediated p53 K372 methylation stabilizes p53 protein suggests that this methylation event may also modulate p53 ubiquitylation, which also occurs at the C-terminus of p53. Although direct evidence is still lacking, a recent study showed that besides ubiquitylation of p53, MDM2 recruits two H3K9 KMTs, KMT1A and KMT1D, to modulate p53 activity . KMT1D and its closely related homolog KMT1C are able to methylate p53 specifically at K373 and repress p53 transactivation activity [4,162]. Knockdown of each p53-associated KMT1 protein increases p53 activity during the stress response, whereas their overexpression inhibits p53 activity. Although modification of p53 by these KMT1 proteins is enhanced by MDM2, changes in total p53 protein levels were not observed in these cells, indicating a more complicated regulation .
Regulation of protein–protein interactions by p53 methylation
One main function of histone lysine methylation is modulating interactions of chromatin with proteins containing ‘reader’ domains that specifically recognize this modification . Methylation of non-histone proteins is likely to function through a similar mechanism. As we already discussed, Tip60 associates with and acetylates p53 upon DNA damage. This interaction seems to be dependent on K372 methylation, as Tip60 is able to pull down more p53 proteins premethylated by KMT7, and addition of methylated p53 peptide specifically competes off interactions between p53 and Tip60 . Therefore, KMT7-mediated methylation of p53K372 may serve as an anchor for Tip60. Interestingly, Tip60 contains a chromodomain, a reader domain of methyl-histones. Although there is no direct evidence, it is likely that the Tip60 chromodomain specifically recognizes methylation of p53K372 to facilitate Tip60–p53 protein–protein interactions.
Two conserved domains, the MBT domain of L3MBTL1 and Tudor domain of 53BP1, both of which function as readers for histone methylation, have been shown to also read p53 methylation. Gozani and colleagues reported that KMT5A-mediated p53K382 methylation promotes the interaction between p53 and L3MBTL1, a general transcriptional corepressor. Compelling evidence from biochemical analyses and crystallographic studies showed that the MBT domain of L3MBTL1 specifically recognizes p53K382me1, with a KD value comparable with that of L3MBTL1 MBT-H3K20me1 . This interaction is essential for the chromatin occupancy of L3MBTL1 at p53 target promoters, as either knockdown of L3MBTL1 or mutation of the MBT domains that abolishes L3MBTL1–p53 interaction results in a p53-dependent increase in p21 and PUMA transcript levels in the absence of DNA damage .
In a study of the regulation of p53 by the lysine demethylase KDM1A/LSD1, Berger and colleagues identified 53BP1 as a reader for p53K370 dimethylation (p53K370me2) . In addition to mono-methylation (by KMT3C), p53 K370 can also be dimethylated in vivo. Although the primary KMT putting on this methylation mark has not been identified, KDM1A was shown to specifically remove p53K370me2 in cells . Surprisingly, K370me2 has a different role in regulating p53 function than K370me1: K370me1 represses p53 function, whereas K370me2 promotes p53 activity . The tandem Tudor domain of 53BP1 was identified by a protein domain array as a reader for p53K370me2. The interaction between p53 and 53BP1 is K370me2 dependent, as mutation of p53K370 to either alanine or arginine totally abolishes their interaction in cells, whereas knockdown of KDM1A leads to increases of p53K382me2 levels and p53–53BP1 interaction .
In more comprehensive studies, it was showed that, in vitro, the tandem Tudor domain of 53BP1 binds to several dimethylated lysine residues at the p53 C-terminus, with a preference for p53K382me2 [172,173]. The KD value of 53BP1 Tudor–p53K382me2 is about 0.9 µM, whereas that of 53BP1 Tudor–p53K372me2 is 20 µM, indicating that the binding of 53BP1 Tudor to p53K382me2 is much stronger . Further, generation of K382me2 in the p53K370R mutant retains the ability of p53 protein to interact with 53BP1 in vivo, suggesting that p53 K382 dimethylation is more likely the primary target of 53BP1 Tudor domains . Since other regions of p53 and 53BP1 also interact, further study is needed to define the specificity and functions of p53 C-terminal lysine methylation for 53BP1 association and other activities.
A recent study showed that KMT7 inhibits the association of p53 with SIRT1, a histone deacetylase that regulates p53 acetylation . Interestingly, KMT7 can also directly methylate SIRT1 at several K residues. It is currently not known whether the inhibition of p53–SIRT1 interaction by KMT7 is dependent on p53 methylation or SIRT methylation. Nevertheless, this study suggests that KMT7 might modulate p53 function indirectly as well as through a direct p53 methylation-dependent mechanism.
Methylation and function of p53 in vivo
Although studies using cultured cells have demonstrated that KMT7-mediated methylation of p53 at K372 is important for stabilization and transcriptional activation of p53 [18,19,167,168], two recent studies challenged the notion that KMT7-mediated p53 methylation plays a significant role in p53 function in vivo [174,175]. Testa and colleagues made an independent KMT7 knockout mouse strain, and defined its function in vivo. They found that deletion of KMT7 had no effect on p53-dependent cell-cycle arrest or apoptosis in response to DNA damage induced by radiation, genotoxic agents or oncogene c-myc. Both acetylation of p53 and transactivation of its target genes were preserved in the absence of KMT7, suggesting that in vivo KMT7 is not required for the p53-dependent gene expression program . In an independent study, Rossi and colleagues drew a similar conclusion that KMT7 is dispensable for p53-dependent senescence, transcription, cell-cycle arrest, and apoptosis in vivo . The conclusions are in sharp contrast to findings from a previous in vivo study by Gaudet and colleagues demonstrating that KMT7-mediated p53 methylation is required for p53 acetylation and is important for p53 activation . Although this discrepancy may arise from different genetic backgrounds of the mice used in these studies, independent studies from both the Rossi group and the Testa suggested that the antiserum raised against human p53K372me which was used by the Gaudet and colleagues  cannot detect the methylation of murine p53K369 in vitro , or cannot distinguish the unmethylated and methylated forms of murine p53 in vivo . Of note, a previous study showed that in MEFs the substitution of the six C-terminal lysines of p53 (including K369) with arginine had no effect on the stabilization and activation of p53 [176,177], suggesting that the C-terminal lysine residues play functional redundant roles in fine-tuning p53 functions in stress responses.
Methylation of other non-histone proteins
Methylation of a few non-histone proteins other than p53 has also been reported. Most of these proteins are either transcriptional factors (such as ERα, NFκB, E2F1, RB, and STAT3) or histone- or DNA-modifying enzymes (such as DNMT1 and KMT1C) (Table 1) . Interestingly, most of these proteins are methylated by KMT7. KMT7 is the only H3K4 mono-methyltransferase so far identified . It exhibits strong activity on free histones, but very weak activity on nucleosomal histones, indicating that its main function may be regulating proteins other than histones.
One mechanism by which KMT7 modulates proteins is regulation, either enhancement or inhibition, of protein–protein interactions, similar to histone methylation. For example, KMT7 methylates TBP-associated factor TAF10 and increases the affinity of TAF10 for RNA polymerase II, thereby helping in formation of pre-initiation complex [20,94]. Another example is that KMT7-mediated methylation of RB protein (at K873) promotes its association with HP1 protein for target gene repression and differentiation . Interestingly, KMT7 can also methylate RB protein at a distinct K residue (K810), which impedes binding of CDK and thereby prevents subsequent phosphorylation of the associated serine residue .
Another function of KMT7-mediated methylation is to influence protein stability. Methylation of p53 and ERα by KMT7 stabilizes these proteins, and is necessary for their efficient recruitment to target genes and for their transactivation activities [18,22]. Methylation of MYPT1, a RB protein regulator, by SET7 also stabilizes MYPT1 proteins, and subsequently decreases the steady-state level of phosphorylated RB proteins . In contrast, KMT7-mediated methylation of DNMT1, E2F1, STAT3, and the RelA subunit of NFκB negatively regulates their stability and transactivation activity [17,23–26,180].
It is worth noting that KMT7-mediated methylation in many proteins (such as p53, DNMT1, E2F1, STAT3, and MYPT1) undergoes demethylation by KDM1A, indicating a dynamic regulation of this modification on non-histone proteins . A conserved motif for KMT7 methylation was determined by a stepwise iterative approach using peptide arrays, and >90 protein peptides were identified as KMT7 novel substrates, suggesting that lysine methylation is a common modification for non-histone proteins .
Although most known methylation events on non-histone proteins are mediated by KMT7, a few other KMTs have also been shown to methylate non-histone proteins. For instance, KMT3C is reported to methylate p53 as well as RB, making it an attractive drug target [13,14]. Both KMT3B and SETD6, a novel KMT that has not yet been assigned a name in the new nomenclature, methylate the RelA subunit of NFκB [12,181]. Interestingly, SETD6-mediated methylation is recognized by the KMT1D ANK domain to anchor KMT1D to RelA target genes to repress gene expression under basal conditions . KMT1C is another KMT that methylates non-histone proteins. It methylates transcriptional factor C/EBPβ and chromatin-remodeling factor Reptin to negatively regulate their downstream target genes [5,6]. KMT1C also automethylates itself on a conserved methylation motif, similar to that of histone H3K9. This methylation is necessary and sufficient to mediate in vivo interaction with the epigenetic regulator HP1 to maintain a repressive chromatin . Lysine methylation of non-histone proteins is conserved from yeast to human. Methylation of proteins from budding yeasts (such as Ipl Aurora kinase and ribosomal proteins) and plants (Rubisco subunits) has also been reported [11,183–185]. It is conceivable that many proteins can be modified by lysine methylation, in addition to other types of PTMs.
In the past decade, a large number of KMTs and KDMs have been identified, significantly extending our knowledge of the dynamics of histone methylation and epigenetic regulation. However, the study of this modification of non-histone proteins is still in its infancy. Most studies of lysine methylation have focused on only a few target proteins via a candidate approach. High-throughput methods are needed to identify proteome-wide lysine methylation. One approach in identifying additional non-histone substrates of KMTs is to employ a peptide array to determine the substrate specificity of a given methyltransferase, and then to use the derived specificity profile of the enzyme to predict potential substrates [186,187]. Jeltsch and colleagues have taken steps using this approach to identify additional non-histone substrates of KMT7 and KMT1C [7,27]. As several KMTs show promiscuous activities, however, especially in in vitro systems, it is essential to determine whether the methylation events thus identified do occur in vivo.
Direct identification of protein methylation in vivo can be achieved through mass spectrometric analysis. A recently developed method called stable-isotope labeling by amino acids in cell culture enables proteomic analysis of proteins from whole-cell extracts or enriched fractions . This method provides increased confidence in identification of PTMs and their relative quantification. Because most protein PTMs occur at low abundance, however, some degree of enrichment is necessary prior to mass spectrometric analysis, and the most widely used and efficient method is affinity purification [189–191]. Therefore, a key step in generating a proteome-wide protein lysine methylation profile is to develop state-specific, but not sequence-dependent, antibodies recognizing distinct levels of methylation on lysine residues. With advances in new technologies, we foresee that, in the next few years, a proteome-wide profile of lysine methylation and its biological functions will soon be elucidated. Understanding of the essential role of this tiny modification will further extend our knowledge of ‘histone code’ to ‘PTM code’ in regulating proteins involved in diverse processes.
This work was supported by the grants from the Welch Foundation (G1718), CPRIT (RP110471), and the Center for Cancer Epigenetics at University of Texas MD Anderson Cancer Center. X.S. is a recipient of a Sydney Kimmel Scholar award.
We thank Kathryn Hale for critical reading of this paper. We apologize to researchers whose papers are not cited here because of space constraints.