Recent research has implicated arginine methylation as a major regulator of cellular processes, including transcription, translation, nucleocytoplasmic transport, signalling, DNA repair, RNA processing and splicing. Arginine methylation is evolutionarily conserved, and it is now thought that it may rival other post-translational modifications such as phosphorylation in terms of its occurrence in the proteome. In addition, multiple recent examples demonstrate an exciting new theme: the interplay between methylation and other post-translational modifications such as phosphorylation. In this review, we summarize our current understanding of arginine methylation and the recent advances made, with a focus on the lower eukaryote Saccharomyces cerevisiae. We cover the types of methylated proteins, their responsible methyltransferases, where and how the effects of arginine methylation are seen in the cell, and, finally, discuss the conservation of the biological function of methylarginines between S. cerevisiae and mammals.
In the last decade, the methylation of proteins has received much interest from the scientific community, and is now known to be of common occurrence. A recent survey of the Swiss-Prot protein database [1, 2] ranked methylation as the fourth most common post-translational modification (PTM) based on the number of putative methylation sites . In Saccharomyces cerevisiae, protein methylation has been reported to occur at the N- and C-termini of polypeptides [4, 5], on glutamine , histidine , 2-(3-carboxy-3-aminopropyl)-l-histidine , farnesyl cysteine , arginine [10, 11], lysine , cysteine , aspartic and glutamic acid residues . Methylation reactions are catalysed by enzymes from the superfamily of S-adenosyl-l-methionine (SAM)-dependent methyltransferases. Using SAM as the methyl group donor, the methyltransferases catalyse transfer of the methyl moiety from SAM to their substrates, releasing the methylated protein and S-adenosyl-l-homocysteine as the products. Methyltransferases responsible for all of the above methylation types have been reported in S. cerevisiae, with the exception of those responsible for the methylation of cysteine, aspartic and glutamic acid. The metabolic cost of methylation is high; 12 ATP molecules per methylation event . This high metabolic cost and the fact that multiple types of amino acid residues may be modified suggest that methylation must play an important role in the cell.
In this review, we summarise our current understanding of arginine methylation and recent advances, with a focus on S. cerevisiae. We also discuss the conservation of biological function of methylarginine between S. cerevisiae and mammals.
Arginine methylation as a PTM
Arginine methylation is conserved in eukaryotes and is not found in prokaryotes . It may be detected by a variety of techniques but methyl site assignment has been revolutionised by the use of tandem mass spectrometry [17, 18]. Arginines may be methylated at any of its three guanidium nitrogen atoms. To date, four different forms of methylarginines have been identified. The most prevalent is ω-NG,NG-asymmetric dimethylarginine (aDMA) . Here, two methyl groups are added to one of the terminal nitrogen atoms of the guanidino group (Fig. 1). The remaining methylated derivatives include the ω-NG,N′G-symmetric dimethylarginine (sDMA), where one methyl group is added to each of the terminal nitrogen atoms, and ω-NG-monomethylarginine (MMA), in which only one methyl group is added to the terminal nitrogen atom. Finally, the rare δ-N-monomethylarginine (δ-MMA) has only been found in S. cerevisiae . In this modification, the methyl moiety is covalently attached to the δ-nitrogen atom of the guanidino group (Fig. 1). Protein arginine methyltransferases (PRMTs) catalysing all of the above modifications have been described in S. cerevisiae, and in vivo substrates with the various modifications have been identified, with the exception of sDMA [10, 20, 21]. Although most methylarginines are formed post-translationally , a minority of arginine methylation events have been reported to occur in a co-translational manner [21, 23].
Protein arginine methyltransferases in S. cerevisiae
PRMTs have been classified into four types depending on the modifications they catalyse: (a) ‘type I’ PRMTs catalyse the formation of both MMA and aDMA; (b) ‘type II’ PRMTs catalyse the formation of both MMA and sDMA; (c) ‘type III’ PRMTs only catalyse the formation of MMA; (d) ‘type IV’ PRMTs catalyse the formation of δ-MMA. To date, the majority of PRMTs have been reported to have type I and II activities [24, 25]. There are three type III PRMTs (from human, S. cerevisiae and Trypanosoma brucei) that have been identified [13, 26, 27]. The only type IV PRMT described to date is from S. cerevisiae  (Fig. 1).
In humans, a total of 11 PRMTs have been described to date [24, 25]. Three are evolutionarily well conserved, with ‘PRMT1-like’ methyltransferases being the most conserved followed by ‘PRMT5-like’ and ‘PRMT3-like’ methyltransferases [16, 25]. In contrast, only four PRMTs have been described in S. cerevisiae (Hmt1/Rmt1, Rmt2, Hsl7 and Sfm1) (Fig. 2). Of these, three (Hmt1/Rmt1, Rmt2 and Hsl7) possess the common seven-stranded β-sheet (Rossman-like) fold, while Sfm1 is a SPOUT family methyltransferase. Within the Rossman-like fold, a set of four sequence motifs (I, post-I, II and III) and a THW loop are evolutionarily conserved [28-30]. The I and post-I motifs and the THW loop have been previously reported to form part of the SAM-binding pocket . On the other hand, SPOUT family methyltransferases exhibit an unusual α/β fold, consisting of a series of five β-sheets sandwiched between α–helices, with a rare and deep trefoil knot on their C-termini [31-33].
The predominant S. cerevisiae arginine methyltransferase, Hmt1/Rmt1, is a type I PRMT and hence is capable of catalysing both MMA and aDMA modifications. Hmt1/Rmt1 was identified by two independent groups, and is the functional homologue of the human PRMT1 methyltransferase [10, 11]. The crystal structure of Hmt1/Rmt1 has been solved, and Hmt1/Rmt1 was found to homo-oligomerise to form a hexamer – specifically a trimer of dimers . The dimerisation of Hmt1/Rmt1 is mediated by two small helices to form a ‘donut-shaped’ molecule. Surface electrostatic potential modelling reveals an extensive acidic region around the central hole of the dimer. This negatively charged area has been suggested to be the region to which the positively charged arginine substrates bind. It has been reported that dimerisation of Hmt1/Rmt1 monomers is essential for catalytic activity .
A ‘PRMT1-like’ methyltransferase, Hmt1/Rmt1 is one of two PRMTs in S. cerevisiae that are well-conserved from unicellular eukaryotes to man (the other being Hsl7) [16, 25] (Fig. 2). Like its human homologue, it is responsible for most PRMT activity in S. cerevisiae, accounting for approximately 89% of all aDMA and approximately 66% of all MMA formed in vivo [10, 35]. Gary et al.  also demonstrated that, when cell extracts were methylated in vitro, Hmt1/Rmt1 was only responsible for the formation of approximately 83% of aDMA and was not required for MMA formation. This result suggests some redundancy, at least under in vitro conditions, between Hmt1/Rmt1 and other PRMTs in S. cerevisiae for formation of MMA.
Hmt1/Rmt1 has been shown to methylate both histone and non-histone proteins. In the case of histone proteins, Hmt1/Rmt1 asymmetrically dimethylates histone H4 on residue Arg3 (H4R3me2a) [36, 37]. Although its human homologue, PRMT1, also catalyses H4R3me2a methylation [38, 39], these two methylation events appear to have biologically distinct functions (described further below). In a separate report, Miranda et al.  observed that Hmt1/Rmt1 also methylates histones H2A, H2B, H3 and H4 in vivo. However, because the purified histone preparation was potentially contaminated with other methylatable proteins, this observation requires confirmation.
To date, 14 non-histone substrates have been described for Hmt1/Rmt1 (Table 1). Most of these are RNA-binding heterogeneous nuclear ribonucleoproteins (hnRNPs) that are involved in mRNA processing, rRNA processing, RNA splicing and nucleocytoplasmic transport of RNAs. A smaller group of substrates include proteins involved in DNA repair and the regulation of translation. Interestingly, the small ribosomal protein S2 (Rps2) is also methylated by Hmt1/Rmt1 [13, 41]. This Rps2 methylation event is normally catalysed by zinc-finger domain-containing ‘PRMT3-like’ methyltransferases [42, 43]. However, homologues of ‘PRMT3-like’ methyltransferases are not conserved in S. cerevisiae despite being present in Schizosaccharomyces pombe . Thus, it appears that Hmt1/Rmt1 functionally replaces PRMT3, at least in S. cerevisiae. This is particularly interesting as Hmt1/Rmt1 lacks the zinc-finger domain that is important for substrate (Rps2) recognition in vivo (Fig. 2) [42, 44, 45].
Table 1. Known substrates of arginine methyltransferases in S. cerevisiae
a Where available, biological process data were sourced from the Gene Ontology Consortium  through the Saccharomyces Genome Database  and from the Uniprot Consortium . The biological process GO data have been summarised for this table. b C, cytoplasm; N, nucleus; Nu, nucleolus; M, membrane; Mt, mitochondrion. c Where available, localisation data were sourced from the Gene Ontology Consortium  and from Huh et al.  through the Saccharomyces Genome Database . d Histone H4 was shown by Lacoste et al.  to be methylated by Hmt1/Rmt1 in vitro, but the same report suggested that Hmt1/Rmt1 is not the sole PRMT responsible for histone H4 methylation, and additional methyltransferases are likely to catalyse this H4R3me2a methylation. e No bona fide substrate of Hsl7 in S. cerevisiae has been described to date. However, Hsl7 has been shown to mono- and symmetrically dimethylate calf thymus histone H2A [40, 50].
Rmt2 was identified through motif sequence alignment of the conserved SAM-binding domain against the S. cerevisiae genome . Rmt2 is unusual in that it does not catalyse the addition of methyl groups to the ω-nitrogen atoms of the arginine guanidino group, but instead methylates the δ-nitrogen . Homologues of Rmt2 appear to be conserved in several fungal species, but this conservation is not seen in higher eukaryotes . However, Rmt2 does share 27% sequence identity with the human guanidinoacetate N-methyltransferase, an enzyme that is responsible for SAM-dependent methylation of the δ-nitrogen of guanidinoacetate into creatine . Interestingly, sequence analysis of Rmt2 has identified ankyrin repeats at its N-terminus (Fig. 2) ; such repeats have been shown to be a methyllysine recognition module . This suggests a possible interplay between methylarginine and methyllysine.
The ribosomal protein L12 (Rpl12a and Rpl12b) has been identified as a substrate of Rmt2 (Table 1) . The methylation of ribosomal protein L12 by Rmt2 has been shown to occur in a co-translational manner, a process for which there are only a few reported examples [20, 21, 23]. Recently, Rmt2 was also found to be associated with specific nucleoporins (Nup49, Nup57 and Nup100) . However, the exact role of Rmt2 and its association with ribosomes or nucleoporins remains to be elucidated.
As a ‘PRMT5-like’ methyltransferase, Hsl7 is the second of two PRMTs in S. cerevisiae that are evolutionarily well-conserved (Fig. 2) [16, 25]. Hsl7 was first described in a screen for synthetic lethal partners of histone H3 and H4 partial-deletion mutants . It was later identified as a homologue of human PRMT5 and shown to possess type II methyltransferase activity, capable of forming MMA and sDMA modifications on calf histone H2A [40, 49, 50]. However, this methylation of histone H2A was not observed in vivo. In addition, other reported in vitro substrates, histone H4 and myelin basic protein, were also reported to be poor substrates despite positive previous reports [40, 49, 50]. Unlike its mammalian homologue, which has well-established PRMT activity and known in vivo substrates, there is no reported bona fide substrate for Hsl7 nor has any cellular function associated with its methyltransferase activity been reported. Like Hmt1/Rmt1, Hsl7 has been shown to homo-oligomerise. However, it has not been determined whether oligomerisation of Hsl7 is required for its function .
Hsl7 is known to be associated with the cell cycle. It has been shown to associate with Hsl1 and Swe1 [52, 53]. The degradation of Swe1 is required for cells to exit the G2 phase, and Hsl7, together with Hsl1, has been demonstrated to be a negative regulator of Swe1 . Disruption of the HSL7 gene leads to an elongated bud phenotype and a delay in the G2 phase of the cell cycle [53, 54]. However, these established functions of Hsl7 and the elongated bud phenotype appear to be independent of its methyltransferase activity . It is possible that, Hsl7 methyltransferase activity may only be stimulated under specific conditions such as stress. In support of this, Hsl7 homologues in S. pombe and Arabidopsis thaliana, Skb1, have been shown to be important for cell survival during hyper-osmotic stress. In A. thaliana, cell survival during hyper-osmotic stress is methylation-dependent, but the exact mechanism is unknown in S. pombe [56, 57]. Interestingly, Hsl7-dependent Swe1 degradation and its localisation to the bud neck are disrupted during stress, including high osmolarity .
Sfm1 was very recently identified as a PRMT through gene deletion analyses . It is an unusual PRMT in that it is not part of the family of methyltransferases with the Rossman-like fold (like the other PRMTs) but instead is a member of the SPOUT methyltransferase family. SPOUT methyltransferases were previously only known to methylate rRNAs and tRNAs [59-62]. Thus, Sfm1 may be the founding member of SPOUT family protein methyltransferases. Although the crystal structure of Sfm1 has not been solved, there are several conserved features in this family.
The small ribosomal protein S3 (Rps3) has been identified to be monomethylated by Sfm1 at Arg146 (Table 1) . However, the functional implication of the methylarginine in Rps3 remains to be elucidated, as hypomethylated mutants did not exhibit growth defects. Because Sfm1 has been shown to only generate monomethylarginines on Rps3, we tentatively classify it as a type III PRMT.
There are several lines of evidence to suggest that there are PRMTs in S. cerevisiae that have yet to be described. First, as mentioned above, at least 11% of aDMA formation and 34% of MMA was unaccounted for . This suggests that at least another type I PRMT must exist. Second, as there is currently no evidence that Hsl7 methyltransferase activity is associated with any biological process, it is possible that a separate type II PRMT may be responsible for the small amount of sDMA observed in S. cerevisiae . Third, histone H3R2me2a methylation was reported to be evolutionarily conserved from yeast to man [63, 64]. In human cell lines, this modification is predominantly performed by PRMT6 and PRMT4/CARM1 to a lesser degree [63, 65]. A PRMT6 or PRMT4/CARM1 homologue does not exist in S. cerevisiae, but the H3R2me2a modification is still present in the absence of Hmt1/Rmt1, Rmt2 or Hsl7 . Fourth, because the biological function of monomethylated histone H3R2 (H3R2me) directly opposes that of H3R2me2a, it was suggested that two different methyltransferases are responsible . Finally, it was observed that Hmt1/Rmt1 was not solely responsible for H4R3 dimethylation, as hmt1/rmt1 mutants were still partially methylated [36, 37], and work by Samel et al.  found that Arg37 methylation on the histone H3-like protein Cse4 was not affected by Hmt1/Rmt1, Rmt2 or Hsl7 deletions (Table 1). Taken together, the presence of up to four additional PRMTs in S. cerevisiae is possible.
Recent bioinformatic studies have predicted an additional 33 putative methyltransferases in S. cerevisiae [30, 68, 69]. Of these, 15 were predicted to be protein methyltransferases , and six have since been confirmed to methylate proteins: See1 (lysine) , Efm1 (lysine) , Efm2 (lysine) , Tae1 (N-terminus) , Set5 (lysine)  and Hpm1 (histidine) . However, because the prediction of putative methyltransferase function is not perfect, it is possible that some protein methyltransferases may have been mis-classified. For example, Sfm1, which was originally predicted to methylate RNAs, is a protein methyltransferase . Hence, there is compelling evidence to conclude that additional uncharacterised PRMTs are present in the S. cerevisiae proteome.
Phenotypes of PRMT mutants
In S. cerevisiae, mutant with single gene deletions of the four known PRMTs and combinatorial gene deletions of Hmt1/Rmt1, Rmt2 and Hsl7 are viable without gross defects [19, 20, 40, 64, 73, 74]. Deletion of Hmt1/Rmt1, Rmt2 and Sfm1 produced no cellular phenotype. However, there is some evidence that, under specific conditions, hmt1/rmt1 mutants show a delayed transcriptional response to heat shock, dysfunctional recruitment of transcription-associated proteins to genes involved in the general stress response (e.g. in response to nitrogen starvation), and an increased tolerance to lithium and sodium [75, 76]. The hsl7 mutant has an elongated bud phenotype; however, this was found to be independent of its methyltransferase activity . In mammalian systems, some PRMTs are essential for viability: PRMT1 has been shown to be essential during mouse embryogenesis , PRMT5 deletion mutants are not viable under any circumstances , and PRMT4/CARM1−/− knockout mice are born smaller than wild-type mice and die shortly after birth [78, 79]. PRMT4/CARM1-deficient mouse embryos were also found to be defective in thymocyte development  and chondrocyte proliferation . Interestingly, stem cells deficient in PRMT1 or PRMT4/CARM1 are viable, without noticeable defects [77, 78]. Hence, it has been suggested that the effects of arginine methylation are subtle as opposed to acting as an on/off switch . This may explain why the current 4 S. cerevisiae and some mammalian PRMTs have no obvious deleterious effects.
PRMT substrate motifs
Protein arginine methylation most often occurs within glycine/arginine-rich (GAR) motifs , specifically the RGG motif that is commonly found in RNA- and DNA-binding proteins. Hmt1/Rmt1 and its human homologue PRMT1 are known to methylate within these RGG motifs [11, 83, 84]. However, there are exceptions to this. For example, PRMT1 has been shown to methylate motifs such as the RXR motif  and some RGX or RXG motifs . In another example, PRMT4/CARM1, a type I methyltransferase like PRMT1, does not methylate these RGG motifs. Instead, it methylates arginines within proline/glycine/methionine/arginine-rich (PGM) regions . These PGM motifs are usually found in spliceosome proteins and interact with WW domains and SH3 domains [86, 87]. PRMT5 has also been shown to methylate PGM motifs, and, interestingly, is known to methylate the same proteins as PRMT4/CARM1 (e.g. spliceosomal protein SmB) . As PRMT5 is a type II methyltransferase, this may suggest some interplay with PRMT4, as they catalyse different modifications on the same target protein.
Apart from primary sequences and motifs, other factors such as the 3D structure and site accessibility are also important for determination of methylarginine sites by PRMTs. For example, human PRMT3, a type I methyltransferase, only methylates the C-terminal GAR motif of the human Ewing sarcoma protein, while its paralogue, PRMT1, methylates all GAR motifs on the same protein . In another example, Wooderchak et al.  showed that methylation of similar sequences in the proteins eIF4A1 and fibrillarin by PRMT1 was selective. In support of these examples, bioinformatic work performed in our laboratory found that arginine methylation was strongly associated with disordered and surface accessible regions . These observations suggest that structural elements have an influence on site specificity.
In S. cerevisiae, as mentioned above, Hmt1/Rmt1 is known to methylate within the canonical GAR motifs. In contrast, consensus sequences for Rmt2, Hsl7 and Sfm1 have not been determined. Interestingly, recent bioinformatic work in S. cerevisiae identified two additional non-GAR methylarginine motifs: the GXXR and WXXXR motifs . Although more work is required to elucidate their corresponding PRMTs, these motifs provide additional evidence that arginine methylation may occur beyond the ‘RGG’ paradigm.
Arginine-mediated molecular interactions and the effects of methylation
In protein–protein interactions, arginine is over-represented at protein interfaces and often mediates intra- and inter-protein interactions [91, 92]. Surveys of protein structures revealed that arginines are the most common residue in all ionic salt bridges (at least 40%) [93, 94]. Similarly, approximately 74% of all energetically significant interactions between cations and aromatic rings (cation–π interactions) involve arginine residues . Cation–π interactions are important in protein folding and stability, with reports suggesting that they may be as strong as, if not stronger than, ionic salt bridges in aqueous environments [95, 96]. Through combined use of various interaction mechanisms (e.g. ionic, hydrogen bonds, cation–π and van der Waals), the arginine guanidino group may also display short-range arginine–arginine interactions despite carrying a positive charge. These short-range arginine–arginine interactions have been repeatedly found to mediate the oligomerisation and function of protein complexes [97, 98]. In protein–RNA interactions, approximately 45% of all hydrogen bonds are mediated through the arginine guanidino group [99, 100]. In addition, stacking of the arginine guanidium ion with RNA bases, possibly forming cation–π and van der Waals interactions, has also been reported to contribute to protein–RNA interactions . In protein–DNA interactions, arginine residues are frequently found to form hydrogen bonds with DNA, both through the phosphate backbone and the bases guanine, thymine and adenine [101, 102]. Similar to RNA, stacking of the arginine side chain may contribute to protein–DNA interactions, in which as much as 54% of all cation–π interactions involved arginine residues [102, 103].
The effects of arginine methylation are subtle; it slightly increases the bulkiness and hydrophobicity of the affected residue. In sharp contrast to phosphorylation or acetylation, methylation does not alter the charge of the residue, but may block important hydrogen bonds. In addition, methylation of arginine residues may enhance stacking and hydrophobic interactions with aromatic rings. The binding stability of dimethylated arginines to tryptophan residues was reported to be twice that of unmethylated arginines . A methylarginine-specific domain harnessing this increase in binding stability, the Tudor domain, has been described [85, 105]. In this methylarginine-dependent interaction, the increase in affinity of arginines for aromatic residues following methylation may facilitate binding to the aromatic ‘cage-like’ methyl-binding pocket of the Tudor domain (Fig. 3) . Such PTM-specific domains have also been described for methyllysine (chromodomain), acetyllysine (bromodomain) and phosphotyrosine (SH2 domain) .
Methylarginine-regulated cellular processes and interactions
Arginine methylation has been implicated in a plethora of cellular processes. These are mediated through protein–protein and protein–nucleic acid interactions. For instance, methylation of the ‘RGG box’ motif of nuclear-localised fragile X mental retardation protein affects its ability to bind mRNAs with the G-quartet structure; including its mRNA (FMR1) [23, 108-110]. Methylation of RNA-binding proteins and its function have been reviewed recently . Methylation may also affect protein–DNA interactions. The high-mobility group (HMG) A1 protein has been shown to be methylated at six sites within its three DNA-binding ‘AT hooks’ [112-115]. This indicates potential roles for arginine methylation in regulating HMGA1 DNA-binding activities. The various HMG protein modifications and their function have also been reviewed recently . Lastly, the majority of cellular functions affected by methylarginines are regulated through protein–protein interactions (reviewed in ). In S. cerevisiae, methylation of the hnRNP protein Npl3 leads to decreased self-association and interaction with the transcription elongation factor Tho2 [75, 117, 118]. Similarly, in mammalian systems, methylation of arginine residues adjacent to proline-rich motifs blocks binding to SH3 domains, but, interestingly, not WW domains . Arginine methylation may also positively facilitate protein–protein interactions and the formation of protein complexes. For example, placement of sDMA residues on SmB, SmD1 and SmD3 splicing factors and the Sm-like protein LSm4 allows interaction with the Tudor domains of the spinal muscular atrophy gene product (SMN) [105, 120-122]. Similarly, aDMA formation on the transcription elongation regulator 1 (CA150) by PRMT4/CARM1 enhances its interaction with SMN Tudor domains . The effects of methylarginine-mediated interactions are seen in a number of cellular processes, including chromatin remodelling, transcription, RNA processing and transport, translation, signal transduction and DNA repair. Some of these functions are conserved from yeast to man, and are discussed further below.
Transcriptional regulation through the histone code
Positive and negative regulation of transcription through arginine methylation of histones is well-documented in mammalian systems. To date, five mammalian PRMTs have been described as performing histone methylation: (a) PRMT1 and PRMT2 methylate histone H4R3, although PRMT1 has an approximately 600-fold higher activity than PRMT2 towards the same substrate [38, 39, 123]; (b) PRMT4/CARM1 methylates histone H3R2me2a, H3R17me2a and H3R26me2a ; (c) PRMT5 methylates histones H3R8 and H4R3 ; (d) although PRMT4/CARM1 methylates histone H3R2me2a, PRMT6 is the major methyltransferase for this activity . The formation of aDMA by PRMT1 and PRMT4/CARM1 is generally associated with gene activation, such as cooperation with p160-type co-activators (e.g. glucocorticoid receptor interacting protein 1 [GRIP1], nuclear receptor coactivator 1 [SRC-1] and activator of thyroid and retinoic acid receptor [ACTR]) and with histone acetyltransferases (e.g. p300 and CREB-binding protein [CBP]) [125-127]. In contrast, the formation of sDMA residues by PRMT5 is generally associated with gene repression [124, 128-131], but recent reports have demonstrated PRMT5 methylation-dependent activation of gene transcription [132, 133]. Importantly, as PRMT5 and PRMT1 methylates the same site on histone 4, the different dimethylation modifications produced by these enzymes (sDMA versus aDMA, respectively) may potentially produce antagonistic effects. [24, 134].
Transcriptional regulation by methylation of arginines in histones in S. cerevisiae is less well understood. Hmt1/Rmt1 has been shown to perform histone H4R3me2a methylation [36, 37]. However, in contrast with mammalian systems, in which PRMT1 dimethylation of histone H4R3 leads to transcriptional activation, the same methylation by Hmt1/Rmt1 in S. cerevisiae was instead observed to elicit transcriptional repression [36, 38, 39]. In line with this finding, Hmt1/Rmt1 had been previously found to be involved in the formation and maintenance of silent chromatin . Yu et al. reported that recruitment of the histone deacetylase Sir2 to chromatin, previously shown to be involved in chromatin silencing , is dependent on Hmt1/Rmt1 methyltransferase activity . Enzymatically inactive Hmt1/Rmt1 led to the decrease of Sir2 recruitment to silent chromatin regions, while over-expression of active Hmt1/Rmt1 led to a concomitant increase of Sir2 recruitment to these regions. Consequently, the inactive Hmt1/Rmt1 led to an increase in acetylation in these regions, while over-expression of active Hmt1/Rmt1 led to a decrease in acetylation and a concomitant increase in H4R3 dimethylation . Most interestingly, Hmt1/Rmt1 has been observed to preferentially bind and methylate acetylated histones compared to unmodified histones . It is therefore tempting to hypothesize that Hmt1/Rmt1 preferentially interacts with acetylated histones to recruit the histone deacetylase Sir2 to acetylated sites. This recruitment of Sir2 results in removal of acetylation marks and subsequent Hmt1/Rmt1-mediated dimethylation of H4R3, leading to transcriptional repression .
Histone H3R2me2a is evolutionarily well-conserved from yeast to man [63, 64]. This aDMA mark on histone 3 directly blocks formation of trimethyllysine on its neighbouring H3K4 (H3K4me3) residue by the COMPASS Set1 methyltransferase complex . This leads to a repression of gene transcription, as H3K4 tri-methylation is associated with transcriptional activation. By contrast, an MMA mark on the same residue (histone H3R2) does not block H3K4me3 formation and is associated with transcriptional activation . However, the PRMT(s) responsible for histone H3R2 methylation in both instances is currently unidentified.
Finally, the lysine methyltransferase Set5 has been recently identified as monomethylating histone H4K5, K8 and K12. As these sites are in close proximity to the histone H4R3 site of Hmt1/Rmt1 methylation, an interplay similar to that for the histone H3R2 and H3K4 modifications described above is possible .
Transcriptional regulation through non-histone proteins
Instead of regulating transcription through the histone code, arginine methylation may also affect transcription through methylation of non-histone proteins involved in transcription. This may lead to either activation or repression of transcription. This may be effected in two ways. One means involves methylation of co-activators, which may lead to modification of histones. This is an ‘indirect’ means by which PRMTs may affect epigenetics. Alternatively, PRMTs may directly methylate transcription factors to affect transcriptional activation. Mammalian examples of these two mechanisms have been described previously [24, 134].
In S. cerevisiae, transcription elongation and termination are regulated by accumulation of ‘anti-termination’ and ‘termination’ factors on the elongating transcript. The hnRNPs Npl3 and Hrp1 are examples of an anti-terminator and terminator, respectively. Simply put, Npl3 prevents transcription termination by factors such as Hrp1 at weak terminator sites [137, 138]. In addition to its anti-terminator functions, Npl3 has been shown to enhance RNA polymerase II activity [139, 140]. Importantly, the functions of Npl3 and Hrp1 and recruitment to the nascent transcripts appear to be dependent on Hmt1/Rmt1-methyltransferase activity [75, 140]. As a result, defective recruitment of Npl3 to the transcript, together with the concomitant increase in recruitment of Hrp1, may determine the site of transcription termination . Finally, methylation of Npl3 has also been implicated in proper dissociation of Tho2, another Hmt1/Rmt1 substrate and a member of the TREX transcription/export complex, from completed mRNA-protein complexes (mRNPs) that are ready for export. This dissociation allows recycling of Tho2 (and TREX) back to sites of transcription to stimulate transcription by RNA polymerase II [75, 140-142]. Failure to do so leads to decreased production of the target gene mRNA and delayed export out of the nucleus [75, 140]. Arginine methylation may also negatively regulate transcription. While examples are available for mammalian systems [143, 144], this form of transcriptional regulation has yet to be described in S. cerevisiae.
Arginine methylation has been implicated in many aspects of RNA metabolism, such as RNA splicing and nucleocytoplasmic transport. In higher eukaryotes, methylation of RG repeats of Sm- and Sm-like proteins leads to increased binding by the Tudor domain-containing SMN protein [105, 120-122]. The SMN protein is required for the cytoplasmic assembly of mature spliceosomal small nuclear ribonucleoproteins and their subsequent nuclear import . This preferential binding of SMN to symmetrically dimethylated Sm- and Sm-like proteins demonstrated a clear link between RNA splicing and methylarginines. Arginine methylation is also required for localisation of spliceosomal small nuclear ribonucleoproteins to Cajal bodies in the nucleus, where they are further processed before being incorporated into mature spliceosomes . The methylation of spliceosomal proteins is performed by PRMT5 and PRMT7 [147, 148]. Recently, PRMT4/CARM1 has also been identified as a regulator of alternative splicing by asymmetrically dimethylating the PGM motif of spliceosome components such as CA150 .
In S. cerevisiae, the RG motifs of Sm proteins, normally methylated in mammalian systems, are not conserved. That being said, proteins Npl3 and Snp1 involved in RNA splicing have been shown to be methylated in S. cerevisiae [149, 150]. However, the function of arginine methylation in RNA splicing in S. cerevisiae is only beginning to be elucidated. A recent report showed that arginine methylation by Hmt1/Rmt1 is required for the proper co-transcriptional recruitment of pre-mRNA splicing factors to intron-containing genes . In hmt1/rmt1 mutants, increased binding of splicing factors to the meiosis-specific gene (HOP2) was observed. This led to a concomitant increase in splicing of up to 50% for the HOP2 transcript ; Hop2 is a meiosis-specific protein whose transcript is encoded during vegetative growth but is only spliced during meiosis .
Cytonucleoplasmic transport of RNA and localisation of proteins
In S. cerevisiae, hnRNPs such as Npl3, Hrp1, Hrb1 and Nab2 are known to be involved in the nucleocytoplasmic transport of mRNA. The shuttling of Npl3 and Nab2 between the nucleus and cytoplasm has been found to be methylation-dependent [152-155]. Hypomethylated versions of these hnRNPs were found to accumulate in the nucleus, while over-expression of their methylating enzyme, Hmt1/Rmt1, relieved this defect and encouraged their export [153, 155]. However, not all hnRNP-mediated mRNA export is methylation-dependent. For example, although Hrb1 is methylated by Hmt1/Rmt1, its nucleocytoplasmic shuttling is not affected by this modification . Note, however, that there are questions regarding whether Hrb1 is a bona fide substrate of Hmt1/Rmt1 . In contrast, for Hrp1, an in vivo substrate of Hmt1/Rmt1 , nucleocytoplasmic shuttling was found to be heavily dependent on the nucleocytoplasmic shuttling of Npl3. Interestingly, it was also found that export of all four hnRNPs (Npl3, Hrp1, Hrb1 and Nab2) required ongoing RNA synthesis by RNA polymerase II [152, 153, 156]. This is in agreement with work by Hurt et al.  and Wong et al. , who demonstrated that Npl3, Hrp1 and Hrb1 are recruited to actively transcribed genes. This demonstrates a clear link between RNA synthesis and RNA export.
The arginine methylation-dependent localisation of proteins is evolutionarily conserved. In mammalian systems, the subcellular localisation of hnRNPA2, SAM68 RNA binding protein and helicase A is dependent on their methylation status. These proteins are substrates of PRMT1. Localisation of helicase A and hnRNPA2 to the nucleus requires methylation [158, 159]. In contrast, methylated SAM68 localises to the cytoplasm .
Ribosomal proteins are subjected to a variety of PTMs. In two separate reports, arginine methylation, particularly aDMA, was found to be the most abundant form of PTM in both the 40S and 60S subunits in HeLa cells [161, 162]. Arginine methylation of ribosomes is evolutionarily conserved and is found in S. cerevisiae [41, 90, 163, 164]. Recent reports have demonstrated that arginine methylation affects the biogenesis and stability of ribosomes [165-167]. In the fission yeast S. pombe, hypomethylation of ribosomal protein S2 (Rps2) reduces the efficiency of 40S subunit synthesis, leading to an imbalance of the 40S:60S free subunit ratio in the cytoplasm [43, 165]. The methylation on Rps2 is evolutionarily conserved from yeasts to man, and is catalysed by PRMT3 homologues, with the exception of S. cerevisiae, in which Hmt1/Rmt1 is responsible . Similarly, the human ribosomal protein S3 (RPS3) has been reported to be methylated by PRMT1 . It was found that RPS3 that is arginine methylation-defective due to arginine to alanine point mutations was not imported into the nucleolus. As a result, it was not incorporated into mature ribosomes and the 40S:60S subunit ratios were unbalanced, similar to hypomethylated Rps2 in S. pombe . Although Rps3 methylation is conserved in S. cerevisiae, methylation of conserved arginine sites at Arg63 and Arg64 (corresponding to Arg64 and Arg65 by PRMT1 in humans ) was not detected . Instead, methylation was found at Arg146 and was catalysed by Sfm1. The functional implication of this methylarginine remains to be elucidated.
Interestingly, ribosomal methylation in HeLa cells is cell cycle-dependent , and the Rps2 methylation in S. cerevisiae appears to be growth condition-dependent . These observations suggest that methylation may play a regulatory role in the growth of S. cerevisiae.
Regulation of arginine methylation
Localization of PRMTs
The localization of a PRMT may determine and limit the processes it is involved in. Hmt1/Rmt1 is found in the nucleus as well as the cytoplasm . Rmt2 is also found in the nucleus and cytoplasm, but in granulated and punctate structures [21, 47, 169]. Hsl7 is found in the cytoplasm and strongly localizes to the bud neck, and, to a lesser extent, the spindle pole bodies during the G1 phase [51, 169]. In addition to the subcellular localizations mentioned above, tissue localization may also play a role. For instance, human PRMT8 is only expressed in brain tissues despite an approximate 80% sequence similarity to PRMT1, which is expressed in all cell types .
Modulation of arginine methyltransferase activity via oligomerization
The formation of homodimers or homo-oligomers has been linked to activity for the enzymes Hmt1/Rmt1, Sfm1, PRMT1, PRMT4/CARM1 and PRMT5 [13, 34, 171, 172]. Homodimerization of Hmt1/Rmt1 has been reported to be essential for its enzymatic activity . Likewise, disruption of PRMT4/CARM1 homodimerization has also been reported to reduce its enzymatic activity . However, PRMT1 and PRMT5 have been observed to only be active in the form of multimers [171, 173, 174]. Although structural and enzymatic studies of Sfm1 have not been performed, studies of homologous SPOUT family methyltransferases suggest that Sfm1 methyltransferase activity may be regulated by its homodimerization [33, 175].
Modulation of arginine methyltransferase activity via protein regulators and PTMs
PRMT function may be modulated by interaction with various proteins. This may lead to inhibition, enhancement or activation of methyltransferase function or alteration of specificity/substrate preference. As an example of a blocking interaction, the human tumour suppressor gene DAL-1/4.1B inhibits PRMT3 methyltransferase activity both in vitro and in vivo . Similarly, in S. cerevisiae, the ring-finger proteins Air1 and Air2 are able to bind to Hmt1/Rmt1 and inhibit its ability to methylate Npl3 in a dose-dependent manner . In contrast, PRMT1 activity has been shown to be synergistically enhanced both in vitro (up to eightfold more) and in vivo through direct interactions with PRMT2 . Likewise, PRMT5 is only active in the presence of methylosome protein 50 (MEP50) . Proteins that interact with PRMTs may also modulate their substrate specificity. For example, the ‘co-operator of PRMT5’ protein may direct PRMT5 to preferentially methylate histone H4R3 over H3R8 [24, 131].
PRMT enzymatic activity may be altered through PTMs on the methyltransferases themselves. PRMT4/CARM1 is phosphorylated during mitosis at two possible sites, affecting its enzymatic activity differently. Higashimoto et al.  demonstrated that phosphorylation at Ser229 in the PRMT4/CARM1 dimerization interface prevented its homodimerization. This consequently disrupted its enzymatic activity . In a separate report, phosphorylation at Ser217 was observed to disrupt an essential hydrogen bond, between Ser217 and Tyr154, that is crucial for formation of the SAM-binding pocket (instead of PRMT4/CARM1 dimerization, as above) . Importantly, it was also observed that similar serine and tyrosine residues are completely conserved in type I but not type II PRMTs, suggesting that phosphorylation of Ser217 may be a general mechanism for regulation of type I PRMT function. Apart from PTMs deposited by other modifying enzymes, auto-methylation may also occur. PRMT6 and PRMT8 are strongly auto-methylated [180, 181], but the effect of auto-methylation has not been described. In S. cerevisiae, Hsl7 is phosphorylated by Hsl1, but this phosphorylation event has not been reported to affect its methyltransferase activity . There have been no reports to date of PTMs on Hmt1/Rmt1, Rmt2 or Sfm1.
Demethylation of arginine residues
Demethylases are enzymes that remove methyl groups and return the amino acid residue to its unmethylated form. The first identified demethylase, lysine demethylase LSD1, was observed to demethylate H3K4 . This discovery was significant as it was the first evidence that protein methylation is reversible. Since then, demethylases capable of demethylating mono-, di- and trimethyllysines have been reported [183-185]. Two lysine demethylases, Jhd1 and Jhd2, are conserved in S. cerevisiae [183, 186].
Recently, the Jumonji domain-containing protein JMJD6 was reported to be the first (human) methylarginine demethylase . However, two papers recently described JMJD6 as a lysine hydroxylase, with no detectable methylarginine demethylase activity [188, 189]. In addition, JMJD6 was found to preferentially bind ssRNA instead of proteins . Further work is required to truly understand JMJD6 function in vivo. There have been no reports to date of arginine demethylases in S. cerevisiae.
It is noteworthy that arginine and methylarginine residues may be converted to citrulline via deimination and demethylimination reactions [191, 192]. This directly antagonizes the effects of arginine methylation and prevents further methylation. However, the biological relevance of demethylimination has been questioned . Currently, peptidyl deimination and demethylimination have not been described in S. cerevisiae and are not discussed further here.
Interplay of arginine methylation with other PTMs
There is increasing evidence that PTMs may cross-talk with one another to achieve dynamic regulation and control of protein function. Arginine methylation may interact with other PTMs in both a positive and negative fashion. Mammalian examples have been reviewed recently . Here, we focus on the examples from S. cerevisiae.
As described above, Hmt1/Rmt1 has been observed to preferentially bind and methylate acetylated histones over unmodified histones. Acetylation of any of the four acetylatable lysine residues in histone H4 leads to enhanced methylation of H4R3 by Hmt1/Rmt1. Interestingly, acetylation of histone H4K8, the most proximal lysine residue, is the most influential. This shows that there may be cooperation between acetylation and formation of nearby methylarginine residues. Oddly, the acetylation and methylation of histone H4 in S. cerevisiae are functionally antagonistic . This is in direct contrast with mammalian systems, where most histone H4 acetylation leads to reduced PRMT1-mediated methylation (while enhancing PRMT5-mediated methylation). Functionally, this is similar to what is observed in S. cerevisiae; PRMT5 (like Hmt1/Rmt1) functions to repress transcription [38, 194].
In addition to positive interplay, methylation may also block the formation of other PTMs. As previously mentioned, Hmt1/Rmt1 methylation in the nucleus encourages the nuclear export of Npl3 [153, 155]. In contrast, phosphorylation of Npl3 by Sky1 kinase in the cytoplasm enhances its interaction with Mtr10 and results in its import into the nucleus [195, 196]. However, hypermethylation of Npl3 by Hmt1/Rmt1 within the Sky1 docking motif significantly attenuates Sky1-mediated phosphorylation of Npl3 [195, 197]. Hence, through dynamic interplay and a balance between the methylation export signals and the phosphorylation import signals, nucleocytoplasmic shuttling of Npl3 in S. cerevisiae may be achieved. Conversely, other PTMs may negatively affect arginine methylation. For instance, the C-terminal domain of mammalian RNA polymerase II is subject to extensive modifications. The residue Arg1810, methylated by PRMT4/CARM1, is important for normal expression of a variety of small nuclear and nucleolar RNAs. However, methylation at this site is blocked by Ser2 and Ser5 phosphorylation in vitro .
The study of protein arginine methylation has gained much momentum during the last decade. The role of arginine methylation is now known to not just be part of the histone code, but to also involve a wide variety of cellular processes. Many of these processes are evolutionarily conserved. Although it has not been reviewed here, it is important to note that PRMTs and their substrates have been increasingly implicated in multiple disease states in humans [24, 134]. As a result, PRMTs themselves have been considered as drug targets [25, 199]. However, before progress can be made in this area, our knowledge of the regulation of methylation events and methyltransferases needs to be dramatically expanded. This requires detailed study of the interplay between methylation and other PTMs such as phosphorylation, acetylation and lysine methylation. We believe that the methylproteome is far more complex than previously imagined, and understanding the methylproteome and its various regulatory mechanisms will be the focus of future research.
We apologise to those researchers whose original work could not be cited due to space constraints. M.R.W. and J.K.K.L. thank the Australian Research Council for financial support. J.K.K.L. acknowledges financial support from an Australian Postgraduate Award and the University of New South Wales Research Excellence Award.