Function of the MYND Domain and C-Terminal Region in Regulating the Subcellular Localization and Catalytic Activity of the SMYD Family Lysine Methyltransferase Set5

SMYD lysine methyltransferases target histones and nonhistone proteins for methylation and are critical regulators of muscle development and implicated in neoplastic transformation. They are characterized by a split catalytic SET domain and an intervening MYND zinc finger domain, as well as an extended C-terminal domain. Saccharomyces cerevisiae contains two SMYD proteins, Set5 and Set6, which share structural elements with the mammalian SMYD enzymes.

ABSTRACT

SMYD lysine methyltransferases target histones and nonhistone proteins for methylation and are critical regulators of muscle development and implicated in neoplastic transformation. They are characterized by a split catalytic SET domain and an intervening MYND zinc finger domain, as well as an extended C-terminal domain. Saccharomyces cerevisiae contains two SMYD proteins, Set5 and Set6, which share structural elements with the mammalian SMYD enzymes. Set5 is a histone H4 lysine 5, 8, and 12 methyltransferase, implicated in the regulation of stress responses and genome stability. While the SMYD proteins have diverse roles in cells, there are many gaps in our understanding of how these enzymes are regulated. Here, we performed mutational analysis of Set5, combined with phosphoproteomics, to identify regulatory mechanisms for its enzymatic activity and subcellular localization. Our results indicate that the MYND domain promotes Set5 chromatin association in cells and is required for its role in repressing subtelomeric genes. Phosphoproteomics revealed extensive phosphorylation of Set5, and phosphomimetic mutations enhance Set5 catalytic activity but diminish its ability to interact with chromatin in cells. These studies uncover multiple regions within Set5 that regulate its localization and activity and highlight potential avenues for understanding mechanisms controlling the diverse roles of SMYD enzymes.

INTRODUCTION

The majority of protein lysine methyltransferases characterized to date possess a SET [Su(var)3-9, Enhancer of zeste, Trithorax] domain, which catalyzes mono-, di-, or trimethylation of the primary amine of the lysine side chain (1, 2). The SMYD lysine methyltransferases comprise a subfamily of SET domain-containing enzymes which are defined by a split SET domain and an intervening zinc finger domain known as MYND (myeloid translocation protein, Nervy, Deaf) (3, 4). In metazoans, the SMYD proteins, which include SMYD1, SMYD2, SMYD3, SMYD4, and SMYD5, make important contributions to skeletal and cardiac muscle development (4–6). For example, Smyd1 has been implicated in muscle development, differentiation, and cardiac muscle function in zebrafish and mice (7–16) and linked to cardiac disease in humans (5, 17). Zebrafish Smyd2 has also been reported to play a role in proper cardiac and skeletal muscle development (18), and human SMYD3 regulates a transcriptional program linked to muscle atrophy (19). In addition to these developmental roles, altered regulation of SMYD proteins is associated with neoplastic growth. The roles of SMYD2 and SMYD3 have been most well defined in pancreatic ductal adenocarcinoma (20, 21) and leukemia (22, 23). Altogether, these findings indicate critical roles for this subfamily of SET domain enzymes during development and cancer.

The SMYD proteins are known to target both histone and nonhistone substrates for methylation (6). The substrates of mammalian SMYD2 and SMYD3 have been the most well characterized, and proteomic studies have identified a diverse set of candidate targets (20, 24–26). Biochemically, SMYD2 has been characterized to methylate Hsp90, Rb, p53, and MAPKAPK3 (21, 24, 27, 28), and histone H4 at lysine 5 (K5) and MAP3K2 are targets of SMYD3 (20, 29). In addition, histone H3K4 has been suggested to be a target of SMYD1 (10) and SMYD5 has been linked to H4K20 methylation (30). Consistently, SMYD proteins primarily appear to have both cytoplasmic and nuclear localization, although the extent to which an enzyme is found in one compartment or the other may be dependent on cell type, growth conditions, and other factors (20, 31, 32). This diversity of substrates and variability in subcellular localization indicate that there are likely to be multiple mechanisms regulating the catalytic activity and functions of these enzymes within the cell.

Structure-function analyses have probed different components of the SMYD proteins to determine their contribution to activity and function. The zinc finger MYND domain has been predicted to be both a protein-protein and protein-nucleic acid interaction domain (4). The MYND domain was originally defined to bind proline-rich sequences through investigation of ZYMND11/BS69 (33), and the MYND domains of some SMYD proteins have similarly been shown to interact with proline-rich targets (24, 34, 35). MYND domain sequences are often highly basic (4, 36, 37), and the MYND domain of SMYD3 has been reported to interact with DNA (37). In structural studies, the MYND domain appears to be in contact with the SET domain, but residues within the MYND domain do not contribute to cofactor or substrate binding (36, 38). Consistent with these findings, deletion of the MYND domain of SMYD2 does not appear to disrupt methyltransferase activity on histones (34); however, the consequences of deletion or mutation of the SMYD3 MYND domain is less clear. Loss of the domain in an N-terminally truncated SMYD3 did not impede catalytic activity on bulk histones (39), but the MYND domain-deleted SMYD3 was not able to be stimulated by DNA, as shown for the wild-type protein (37).

The majority of mammalian SMYD proteins also contain a conserved C terminus, named the C-terminal domain (CTD), which has tetratricopeptide repeat (TPR) domains and often mediates protein-protein interactions (24, 37). In crystal structures, the CTD is positioned near the active site, and there are a number of acidic residues within the domain which may be important to substrate binding by the enzymes (37, 40). Deletion of the CTD in SMYD1 enhances its methyltransferase activity, indicating an inhibitory role of the domain (36); however, removal of the SMYD3 CTD or mutation of acidic residues within the domain is reported to decrease catalytic activity (37, 41). While further biochemical and structural studies are required, these data at least suggest that there may be multiple mechanisms through which the CTD contributes to enzymatic activity for these proteins.

There are two SMYD enzymes in budding yeast, known as Set5 and Set6, which carry the same domain structure and high sequence similarity to the mammalian proteins and are considered to be ancestral members of this protein family (3). Yeast Set6 is not well characterized; however, we previously demonstrated that Set5 is an H4K5, K8, and K12 methyltransferase (42) that contributes to gene repression and genome stability near transposable elements and telomeres in conjunction with Set1, the yeast H3K4 methyltransferase (43, 44). Similar to the mammalian SMYD proteins, Set5 is localized to both the cytoplasm and nucleus, where it interacts with chromatin (42). In addition to its role in genomic stability and stress responses with Set1, overexpression of Set5 has been shown to confer resistance to acetic acid, oxidative stress, osmotic stress, and heat stress (45, 46), indicating that it may play general or stress-specific roles in protecting cells. Although nonhistone substrates of Set5 have not yet been identified, its cytoplasmic localization and stress-responsive phenotypes suggest it is also likely to target other proteins for methylation, similar to human SMYD2 and SMYD3.

In this study, we performed mutational analysis of domains within Set5 to determine their contribution to protein localization and methylation activity with the aim of uncovering potential regulatory mechanisms for Set5 activity, which may be applicable to the broader family of SMYD enzymes. The MYND domain of Set5 is highly similar to those of the mammalian SMYD proteins, although its role in Set5 function is not yet clear. In addition, while Set5 also has an extended C-terminal region after the SET domain (Fig. 1A), it does not have recognizable TPR domains as present in the mammalian SMYD enzymes. We therefore refer to this C-terminal region as the CTR in Set5, since it is not yet clear whether it plays an analogous role to mammalian SMYD CTD sequences. Our results identify a role for the MYND domain in promoting Set5 localization with chromatin in cells, possibly via a stabilizing interaction with DNA, and uncover extensive phosphorylation of Set5, particularly within the CTR, that we implicate in the regulation of Set5 catalytic activity and subcellular localization.

FIG 1.

The MYND domain promotes Set5 chromatin association. (A) Domain architecture of wild-type Set5 and deletion constructs used for expression in yeast. Set5 contains a split SET domain (blue), a zinc finger MYND domain (yellow), and a post-SET (PS) domain (purple). The extended C-terminal region is designated CTR. (B) Immunoblot of whole-cell extracts probed with anti-Set5 showing levels of WT, MYNDΔ, PSΔ, and CTRΔ Set5 expressed from a pRS316 vector with an endogenous SET5 promoter in set5Δ yeast cells. EV, empty vector (pRS316 alone). Anti-Hxk2 is shown as a loading control. *, nonspecific bands detected by anti-Set5. (C, left) Immunoblots of subcellular fractionation of yeast cells probed with anti-Set5, anti-Hxk2 (soluble loading control), or anti-H3 (chromatin loading control). Total, supernatant (sup.), and pellet are shown for cells carrying an EV or expressing Set5_WT or Set5_MYNDΔ. (C, right) Quantitation of the relative amount of Set5 in the supernatant and pellet fractions from three independent experiments. Error bars represent the standard errors of the mean (SEM). (D and E) Similar to panel C, except that cells carrying an EV or expressing Set5_WT or Set5_PSΔ (D) or Set5_CTRΔ-expressing cells (E) were analyzed and anti-H4 was used as the chromatin fraction loading control. Quantitation is from three independent experiments.

RESULTS AND DISCUSSION

The MYND domain promotes Set5 interaction with chromatin.

In yeast and mammalian cells, the SMYD methyltransferases show both nuclear and cytoplasmic distribution, although the extent to which these proteins can be found in either the nucleus or the cytoplasm appears to vary among cell types (20, 31, 32), suggesting that their subcellular distribution may be regulated. We have previously shown that Set5 localizes to both the nucleus and the cytoplasm, with a smaller fraction of the protein associated with chromatin (42). Based on the high sequence similarity and their shared methyltransferase activity on H4K5, Set5 appears to be most similar to human SMYD3 (29, 42, 47). Depending on the cell type and growth conditions, SMYD3 shows both cytoplasmic and nuclear distribution (20, 31, 32), and deletion of its CTD has been reported to decrease nuclear localization (41). The MYND domain of SMYD3 is required for DNA binding of SMYD3 in vitro (37), although its role in chromatin interaction or nuclear localization in cells has not been reported. We therefore tested the contribution of the MYND domain, the post-SET (PS) region, which often contributes to catalytic activity (37), and the extended C-terminal region (CTR) of Set5 to its localization by expressing Set5 with domain deletions (Fig. 1A) from a plasmid using the endogenous SET5 promoter in set5Δ cells. The mutant proteins were expressed to a similar level as wild-type Set5 (Fig. 1B), suggesting that the domain deletions were not inherently destabilizing in cells. To investigate the localization of Set5, we used a subcellular fractionation assay in which insoluble chromatin is separated from soluble nuclear and cytoplasmic material. Similar to our previous findings, wild-type Set5 is predominantly found in the soluble cytoplasmic and nuclear fraction, with a smaller pool at chromatin. Deletion of the MYND domain substantially reduced Set5 levels in the chromatin fraction (Fig. 1C), although deletion of either the post-SET (Fig. 1D) or the CTR (Fig. 1E) did not significantly alter the localization of Set5 under the conditions tested. These data suggest that the MYND domain promotes Set5 chromatin association in cells, but the removal of the CTR does not appear to change the Set5-chromatin interaction, nor does the post-SET region drive chromatin interaction in cells.

The MYND domains of Set5 and SMYD3 share approximately 33% sequence identity, with the most conserved residues being those required for zinc coordination (Fig. 2A). Human SMYD3 has previously been reported to interact with DNA via its MYND domain (31, 37). The MYND domain of SMYD3 is highly basic, with an isoelectric point (pI) of 9.59 (48) supporting a charge-based interaction between the domain and DNA. The MYND domain of yeast Set5 is also basic, although it has a slightly lower pI of 8.62 and fewer basic residues (Fig. 2A). Based on the MYND dependence of Set5’s interaction with chromatin in cells (Fig. 1C), we postulated that Set5 may also interact with DNA and that this may help sequester a pool of the protein at chromatin. We tested DNA binding in vitro using an electrophoretic mobility shift assay (EMSA) with the Widom 601 DNA sequence as a binding substrate and recombinant glutathione S-transferase (GST) fusions of either wild-type SMYD3 or Set5, or mutant versions lacking the MYND domain (Fig. 2B), which were expressed and purified from Escherichia coli (Fig. 2C). We tested the binding of GST-Set5_WT at three concentrations and observed weak binding to the 601 DNA sequence at the highest concentration shown (Fig. 2D; lane 7, 250 pmol of protein). Although this binding appeared weak, it was dependent on the MYND domain, since GST-Set5_MYNDΔ did not cause any shift in DNA mobility (Fig. 2D, lanes 8 to 10). Compared to Set5, GST-SMYD3_WT bound DNA at all concentrations tested (Fig. 2D, lanes 11 to 13), suggesting it has higher affinity for DNA than Set5. Consistent with previous reports, DNA binding by SMYD3 is dependent on the MYND domain, since GST-SMYD3_MYNDΔ showed substantially reduced ability to bind DNA (Fig. 2D, lanes 14 to 16). We also tested whether or not the CTR contributes to DNA binding in vitro. Here, we observed some decrease in DNA binding of GST-Set5_CTRΔ compared to the wild-type protein (Fig. 2E), although not to the same extent as the loss of the MYND domain.

FIG 2.

Yeast Set5 and human SMYD3 interact with DNA via their MYND domains. (A) Sequence alignment of the MYND domains of yeast Set5 and human SMYD3. Identical residues are highlighted in yellow, and similar residues are highlighted in gray. The percent identity between the two domains is approximately 33%. The conserved cysteines are required for zinc coordination. (B) Domain architecture of wild-type Set5 and deletion constructs used for expression of GST-tagged proteins in E. coli. GST-tagged proteins are used in experiments presented in this figure and Fig. 3. The domain architecture of human SMYD3 and SMYD3_MYNDΔ is also shown. The CTD of SMYD3 is shown in light gray. (C, top) Coomassie blue-stained SDS-PAGE gel of wild-type and deletion mutant GST-Set5 proteins expressed and purified from E. coli. (C, bottom) Coomassie blue-stained SDS-PAGE gel of GST-SMYD3 and GST-SMYD3_MYNDΔ expressed and purified from E. coli. (D) Agarose gel of electrophoretic mobility shift assay (EMSA) using 1.5 pmol of Widom 601 DNA and purified GST, GST-Set5_WT, GST-Set5_MYNDΔ, GST-SMYD3_WT, and GST-SMYD3_MYNDΔ. For all proteins tested, 50, 100, and 250 pmol of protein were incubated with the DNA. (E) EMSA performed as in panel D with 100 and 250 pmol of GST, GST-Set5_WT, and GST-Set5_CTRΔ.

Overall, these data suggest that Set5 interacts with DNA in vitro in a MYND domain-dependent manner, although to a much lesser extent than human SMYD3. There is some dependence of the Set5-DNA interaction on the CTR in vitro; however, we did not observe any substantial change in chromatin localization in the absence of the CTR in cells (Fig. 1E), suggesting that it is not likely to be an important mediator of the Set5-DNA interaction in vivo. In cells, the MYND domain interaction with DNA may help to stabilize Set5 association with chromatin, leading to decreased chromatin binding in the absence of the MYND domain (Fig. 1C). However, the relatively weak association in vitro suggests that protein-protein interactions likely also contribute to Set5 chromatin localization. In addition, it is plausible that other regions of the protein promote the Set5-chromatin interaction in cells, as we do not observe a complete loss of chromatin binding upon loss of the MYND domain. Future experiments on Set5 protein-protein interactions will help elucidate additional mechanisms through which Set5 subcellular localization is maintained and regulated.

Requirements for the catalytic activity of Set5 outside of the SET domain.

We next tested whether the MYND domain, the post-SET, or CTR of Set5 contributes to its methyltransferase activity. In vitro methylation assays were performed with recombinant GST-Set5_WT and deletion or truncation mutants as shown in Fig. 2B and C, including GST-Set5_MYNDΔ, GST-Set5_CTRΔ, GST-Set5_PSΔ, and GST-Set5_PSΔ+CTRΔ, which includes deletion of both the CTR and the post-SET domain. We monitored methyltransferase activity using radiolabeled S-adenosylmethionine ([³H]SAM) as the methyl donor or using unlabeled SAM and immunoblotting with antibodies against the H4K5me1 and H4K8me1 marks to detect methyltransferase activity. For these experiments, a new antibody against the H4K5me1 mark was generated (see Materials and Methods) that showed high specificity for the H4K5me1 mark using peptide dot blot assays (Fig. 3A). The specificity of the H4K8me1 antibody was demonstrated in previous work (42). Deletion of the MYND domain caused a very modest defect in methyltransferase activity at subsaturating concentrations of enzyme (Fig. 3B and C). This was most evident in the assay using radiolabeled SAM (Fig. 3B) to detect the methyl species rather than immunoblotting with antibodies against H4K5me1 and H4K8me1 (Fig. 3C). This may suggest that either the MYND domain contributes primarily to methylating H4K12me1 (which cannot be uniquely evaluated in this assay due to a lack of highly specific antibodies) or that small differences in methyltransferase activity against all three lysines show an additive difference when monitored together using radiolabeled SAM, but these differences are not apparent at the level of the individual mark. Nonetheless, our results largely agree with previous structural analysis and biochemical studies which indicate that the MYND domain has an independent role from the SET domain and does not significantly contribute to methyltransferase activity in SMYD family enzymes (34, 36, 38, 39).

FIG 3.

The MYND domain and the CTR of Set5 are not required for Set5 methyltransferase activity on H4, but the post-SET domain is required for catalytic activity. (A) Peptide dot blot indicating specificity of the anti-H4K5me1 antibody. The indicated biotinylated peptides were spotted in a dilution series from 1 μg to 15 ng and probed with either streptavidin-horseradish peroxidase (HRP) or anti-H4K5me1 and an HRP-conjugated secondary antibody. (B) Methylation assay showing the activity of recombinant GST-Set5_WT and GST-Set5_MYNDΔ on recombinant H4 as the substrate. [³H]SAM was used as the methyl donor, and the autoradiography (autorad.) is shown, as well as a silver stain, to demonstrate total protein loading of GST-Set5 and H4. Recombinant enzymes were used at 0.75, 1.25, and 5 pmol per reaction and 40 pmol of H4 per reaction. (C) Methylation assay was performed as in panel B except that unlabeled SAM was used as the methyl donor, and reaction products were subjected to immunoblotting with anti-H4K5me1, anti-H4K8me1 (42), anti-Set5 (42), and anti-H4 antibodies. (D and E) Experiments were identical to those for panels B and C except that recombinant GST-Set5_WT and GST-Set5_CTRΔ were used as enzymes. (F) Methylation assay as in panel B with 5 pmol of GST-Set5_WT, GST-Set5_PSΔ+CTRΔ, and GST-Set5_PSΔ.

We also monitored methyltransferase activity on H4 upon deletion of the Set5 CTR. In assays using radiolabeled SAM as a methyl donor or in immunoblots probing reactions performed with unlabeled SAM (Fig. 3D and E), we did not observe any substantial defects in methyltransferase activity on H4 with GST-Set5_CTRΔ. However, deletion of the post-SET domain either on its own or in addition to the CTR caused complete loss of methyltransferase activity (Fig. 3F), indicating that the post-SET domain is required for catalysis of H4 methylation by Set5. This finding is consistent with data from other SET-domain methyltransferases which require an intact post-SET domain for catalytic activity (37). In human SMYD proteins, the role of the CTD in catalytic activity may vary among enzymes. Whereas loss or mutation of the CTD appears to reduce SMYD3 activity (37, 41), truncation of the domain in SMYD1 enhances catalytic activity (36). In contrast, our biochemical analysis of Set5 shows that complete loss of this domain does not significantly influence the catalytic activity of the enzyme under the conditions tested. However, as further explored below, there may be a role for individual residues and posttranslational modification within the CTR in regulating Set5 activity.

The MYND, post-SET, and CTR domains are required for Set5-dependent repression of subtelomeric genes.

We previously demonstrated that Set5 works with the H3K4 methyltransferase Set1 to promote repression of telomere-proximal genes (43, 44). In the absence of Set1, there is a loss of silencing near telomeres, which is enhanced when Set5 is also deleted (43, 44). We have shown that this role for Set5 in gene repression is dependent on its catalytic activity, since mutation of a catalytic residue within the SET domain phenocopied deletion of Set5 (43). To investigate the role of the MYND, post-SET, and CTR domains in Set5-dependent subtelomeric gene repression, we expressed wild-type and mutant versions of Set5 (as shown in Fig. 1B) in set1Δ set5Δ cells and monitored gene expression at two subtelomeric genes, COS12 and YGL262W, using quantitative reverse transcriptase PCR. Expression of wild-type Set5 in set1Δ set5Δ cells showed decreased expression of COS12 and YGL262W compared to cells with the empty vector, indicative of restored repression when Set5 is added back to these mutants (Fig. 4A). However, expression of Set5_MYNDΔ is not able to restore repression of COS12 and YGL262W, indicating that loss of the MYND domain impairs Set5 function in gene expression control. Given that the methylation activity of Set5 is largely intact in the absence of the MYND domain (Fig. 3B and C), this phenotype may be due to its decreased association with chromatin (Fig. 1C). We also found that deletion of the post-SET region abrogated Set5 function in this assay (Fig. 4B), as expected based on the complete loss of methyltransferase activity observed for GST-Set5_PSΔ (Fig. 3F). Intriguingly, although we have not detected a defect in catalytic activity or chromatin localization when the CTR is missing, set1Δ set5Δ cells expressing SET5_CTRΔ showed a phenotype similar to that of the double mutant carrying an empty vector (Fig. 4C), indicating that the CTR does contribute to Set5’s repressive function. These data suggest that the CTR has an additional role in regulating Set5 function that has yet to be determined, possibly though mediating protein-protein interactions which may promote its role in repressing subtelomeric genes. Combined with our localization and biochemical experiments, these findings indicate that the MYND domain, the post-SET, and CTR all contribute to Set5’s function in subtelomeric gene repression together with Set1, although likely through different mechanisms.

FIG 4.

Set5-mediated subtelomeric gene repression requires the MYND domain, post-SET, and CTR. Quantitative reverse transcription-PCR (RT-PCR) of mRNA levels of subtelomeric genes COS12 and YGL262W in set1Δ set5Δ cells carrying either an empty pRS316 vector (EV) or a vector expressing SET5_WT or (A) SET5_MYNDΔ, (B) SET5_PSΔ, or (C) SET5_CTRΔ. Gene expression relative to set1Δ set5Δ cells expressing SET5_WT was calculated for the empty vector and SET5 mutant cells. The housekeeping gene TFC1 was used for normalization. The data are shown for five to seven biological replicates for each strain, and asterisks represent P values from unpaired t tests (*, P ≤ 0.05; ***, P ≤ 0.001; ns, not significant). We note that the expression of COS12 in set1Δ set5Δ cells with SET5_CTRΔ consistently showed increased expression compared to cells with SET5_WT, but the P value across the replicates shown did not reach the threshold of <0.05 due to variability in the data.

Phosphoproteomic analysis reveals extensive phosphorylation of Set5 and potential regulation of its catalytic activity.

We next sought to determine whether or not Set5 itself is subject to regulation via posttranslational modification. To do so, mass spectrometry-based phosphoproteomic analysis was performed on N-terminally FLAG-tagged Set5 immunoprecipitated from yeast under normal growth conditions. This analysis revealed numerous phosphorylation events on serine and threonine residues that were clustered in either the N-terminal region (NTR) or the CTR, along with two additional sites in the region between the MYND domain and the latter part of the SET domain (Table 1). We identified novel sites of phosphorylation, as well as sites that had previously been detected in large-scale proteomic studies (49–54) and in a recent phosphoproteomic analysis of methyltransferases from yeast (55). Table 1 summarizes all of the sites identified in our study and clusters them with sites previously found in other work to reveal the highly phosphorylated regions of Set5. The differences in sites identified between our work and other studies can most likely be attributed to different proteases used for peptide cleavage during sample preparation for mass spectrometry.

TABLE 1.

Phosphosite analysis of Set5^a

Protein region	Site	Reference(s) and/or source
N-terminal region (NTR)	S26	This study
	T90	This study
	T94	55
	S96	55; this study
Inter-MYND and SET	S301	This study
	T305	This study
C-terminal region (CTR)	S458	51, 54, 55; this study
	S461	53–55; this study
	S462	55; this study
	S466	51, 55; this study
	S469	55
	S470	55
	S475	52, 55
	S476	52, 55
	T504	This study
	T511	55
	S512	55; this study
	S517	49, 55; this study
	S520	55; this study

^aPhosphorylation sites on Set5 identified in this study (highlighted in boldface) and other works are listed, and the regions within the Set5 protein in which these sites were found are indicated. Mass spectrometry data have been deposited and are accessible through the Proteome Xchange Consortium, as described in Materials and Methods.

The CTR of Set5 shares just 22% identity with the CTD of SMYD3, and it does not have a clear TPR domain, as common in the majority of the mammalian SMYD proteins (3). SMYD protein structures solved to date indicate that this region of the protein is likely to contribute residues to the substrate binding pocket near the active site and that acidic residues, in particular, appear to be important for enhancing substrate-enzyme interactions (37, 40). Consistent with this, Xu et al. previously reported that deletion of the TPR domain or mutation of acidic residues within the SMYD3 CTD decreased its methyltransferase activity on histone substrates (37). In contrast, our results suggest that deletion of the full CTR in Set5 does not impact catalytic activity. However, we postulated that the extensive phosphorylation within the Set5 CTR may serve to regulate its catalytic activity, given the apparent role of this domain in other SMYD enzymes. To test this, we mutated ten serine and threonine residues within the CTR of Set5 (S458, S461, S462, S466, S475, S476, T511, S512, S517, and S520) to alanine or a phosphomimetic amino acid, aspartic acid, in our GST-Set5 expression construct. These mutants, named GST-Set5_{CTR_mutA} and GST-Set5_{CTR_mutD}, were expressed and purified from E. coli and used in in vitro methylation assays in the presence of radiolabeled SAM. We observed that mutation of the serine and threonine residues subject to phosphorylation in the CTR_mutA protein caused a very mild defect in methylation activity on H4 (Fig. 5A), indicating that the amino acid identity at these positions is not a critical contributor to catalytic activity. However, the CTR_mutD protein showed an enhanced methyltransferase activity relative to the wild-type protein (Fig. 5A, compare lanes 1 to 3 to lanes 9 to 11). This observation was further tested using immunoblotting of the in vitro methylation reactions with the H4K5me1 and H4K8me1 antibodies (Fig. 5B) and also with a more quantitative assay measuring incorporation of the radiolabeled methyl group on a peptide spanning the N terminus of H4 (Fig. 5C). In these assays, we also observed enhanced activity of the GST-Set5_{CTR_mutD} compared to the activity of GST-Set5_WT. These data suggest that the additional negative charge imparted by the aspartic acid residues within the CTR promote enhancement of Set5 catalytic activity toward H4. Although complete deletion of the CTR did not affect methyltransferase activity (Fig. 3D and E), these data suggest that the posttranslational modification of specific residues within the CTR does impart regulatory control on Set5 catalytic activity. Investigation of the contribution of some of the individual phosphosites to methyltransferase activity by generating individual mutations of serine/threonine to aspartic acid did not identify one specific site that contributed to enhanced activity (data not shown). However, further investigation using structural analysis and mutation of different combinations of sites may reveal whether individual phosphorylation events or simply a cumulative increase in phosphorylation within the CTR is the main contributor to this regulation of Set5 activity.

FIG 5.

Phosphorylation within the CTR modulates Set5 enzymatic activity and chromatin association. (A) Methylation assay performed as described for Fig. 3B using GST-Set5_WT, GST-Set5_{CTR_mutA}, and GST-Set5_{CTR_mutD} as enzymes. A total of 40 pmol of recombinant H4 was incubated with 0.37, 0.75, 1.5, and 3 pmol of enzyme. Short (4 h) and long (24 h) exposures of the autoradiograph are shown. (B) Methylation assay performed as described for Fig. 3C using GST-Set5_WT, GST-Set5_{CTR_mutA}, and GST-Set5_{CTR_mutD} as enzymes at 0.37, 0.75, 1.5, and 3 pmol of enzyme. (C) Counts per minute (cpm) of tritium-labeled H4 peptide following incubation with either GST-Set5_WT or GST-Set5_{CTR_mutD}. After the methylation assays, the biotinylated H4 peptides were subjected to pulldown with streptavidin magnetic beads, and the radiolabel was counted in a scintillation counter. Error bars indicate the SEM of three or four replicates. Asterisks represents P values of ≤0.05 from unpaired t tests. (D) Subcellular fractionation performed as described for Fig. 1C from set5Δ yeast cells with EV, SET5_WT, SET5_{CTR_mutA}, or SET5_{CTR_mutD}. (E) Quantitative RT-PCR of mRNA levels of subtelomeric genes COS12 and YGL262W in set1Δ set5Δ cells carrying either an empty pRS316 vector (EV) or vector expressing SET5_WT, SET5_{CTR_mutA}, or SET5_{CTR_mutD}. The relative expression is shown for five to seven biological replicates for each strain, and asterisks represent P values from unpaired t tests (*, P ≤ 0.05; ***, P ≤ 0.001; ns, not significant).

Our subcellular fractionation experiments indicate that complete loss of the CTR does not alter the broad localization of Set5 (Fig. 1D); however, the phosphosite-specific investigation into Set5 methyltransferase activity indicates that there may be additional roles for residues in the CTR that are masked upon full deletion of the region. Therefore, we tested whether or not alanine or aspartic acid mutations of phosphosites within the CTR affect Set5 protein localization in cells. The Set5 mutants, Set5_{CTR_mutA} and Set5_{CTR_mutD}, were expressed from a plasmid driven by the SET5 promoter in a set5Δ strain and showed similar expression level to wild-type Set5 (Fig. 5D, total lanes), and subcellular fractionation was performed on these strains. Set5_{CTR_mutA} showed a very mild decrease in chromatin localization compared to wild-type Set5; however, in cells expressing Set5_{CTR_mutD}, a larger decrease in chromatin localization was observed (Fig. 5D). This suggests that there may be a mild contribution of specific amino acids or regions within the CTR to the Set5-chromatin interaction; however, phosphorylation may also regulate this interaction in cells. Additional experiments will be required to determine whether this difference is due to phosphorylation at specific residues which may interfere with chromatin association or protein-protein interactions, or whether the increased negative charge within the CTR is disrupting interactions between Set5 and chromatin in cells. Overall, this indicates that there may be multiple functions for phosphorylation within the CTR, including regulating catalytic activity and chromatin localization, and suggests that the CTR may be particularly important for controlling activity of Set5 toward any nonchromatin substrates.

We also tested whether or not mutation of Set5 phosphosites impacted its role in subtelomeric gene repression. SET5_{CTR_mutA} and SET5_{CTR_mutD} were expressed in set1Δ set5Δ cells, and the expression of COS12 and YGL262W was analyzed as described for Fig. 4. Neither of the mutants was able to restore repression of these genes to the same extent as wild-type Set5 and cells expressing the mutants more closely resembled those with the empty vector (Fig. 5E). This suggests that mutation within the CTR inhibits Set5’s function in gene repression in cells, similar to complete deletion of the CTR (Fig. 4C). Although Set5_{CTR_mutD} shows enhanced methyltransferase activity, it also has decreased chromatin binding in cells. The increased catalytic activity may not be sufficient to overcome the loss of binding of Set5 to chromatin, thereby preventing it from promoting repression of subtelomeric genes. This further indicates the possibility that phosphorylation of the Set5 CTR is particularly relevant to regulation of non-chromatin-associated substrates of Set5.

In summary, our findings show that the MYND domain is important for Set5 localization to chromatin. Similar to human SMYD3, Set5 binds DNA in vitro, although its affinity for DNA is substantially weaker than that of SMYD3. We speculate that this weak DNA binding, as well as additional protein-protein interactions, work to stabilize the interaction of Set5 with chromatin. The role of the MYND domain in SMYD3 subcellular localization, or that of other SMYD family proteins, has not yet been reported. Given the potential for both nucleic acid and protein binding to MYND domains, it may be that differential interactions with the MYND domain are important for dictating the subcellular localization of the proteins under different conditions or cell types. Neither the MYND domain nor the CTR appears to contribute to Set5 catalytic activity on their own, since their complete deletion does not dramatically alter Set5-mediated methylation of H4 in vitro. However, along with the post-SET region, they are required for Set5’s role in repressing subtelomeric gene expression in coordination with Set1. This indicates that both the MYND domain and CTR have roles beyond regulating catalytic activity which are important for Set5 function in cells. In addition, phosphoproteomic analysis shows that the CTR of Set5 is highly phosphorylated, and our mutational analysis suggests that phosphorylation within the CTR may enhance methyltransferase activity of the enzyme. Intriguingly, in proteome-wide kinase activity screens, Set5 is a common substrate for a diverse set of kinases (56), and, in comparison to other lysine methyltransferases, it may be one of the most highly phosphorylated of the SET domain-containing proteins in yeast (55). This suggests the possibility of dynamic regulation of Set5 activity in the cell in response to signaling events, possibly directing its catalytic activity to either chromatin or nonchromatin substrates, depending on cellular conditions. Future investigation will seek to identify whether individual sites of phosphorylation are key to regulating Set5 under specific conditions, as well as which enzymes and signaling cascades are responsible for controlling these modifications. Moreover, while it has not been reported yet, these data also suggest that investigation of potential regulation of mammalian SMYD enzyme activity by phosphorylation may reveal new regulatory mechanisms. Overall, our findings highlight mechanisms for the regulation of yeast Set5 subcellular localization and catalytic activity and broaden our understanding of the biochemical regulation of SMYD family proteins.

MATERIALS AND METHODS

Yeast growth conditions, strains, and plasmids.

The genotypes for all S. cerevisiae strains used in this study are listed in Table 2. Standard medium was used to grow the yeast strains, including YPD (1% yeast extract, 2% peptone, 2% dextrose) and synthetic complete (SC) or dropout media (US Biological). The pRS316-SET5 yeast expression vector was generated by cloning a PCR fragment that amplified the SET5 locus, including its promoter and 3′ untranslated region, into pRS316. The pGEX-6P1 E. coli expression vector expressing GST-SET5 was previously described (42). Deletion constructs in GST-SET5 and pRS316-SET5 were generated using standard restriction enzyme cloning or Gibson assembly. The multisite phosphorylation mutants were generated through gene synthesis methods (Genewiz) in pRS316-SET5 and subcloned into pGEX-6P1-SET5. All sequences were confirmed by Sanger DNA sequencing. All of the plasmids used in this study are listed in Table 3.

TABLE 2.

Yeast strains used in this study

Strain	Background	Description	Source or reference
yEG001	BY4741	MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0	YKO
yEG231	BY4742	MATα set5Δ::HIS3MX	44
yEG775	BY4742	MATα set5Δ::HIS3MX + p045 (pRS316)	This study
yEG776	BY4742	MATα set5Δ::HIS3MX + p130 (pRS316-SET5)	This study
yEG777	BY4742	MATα set5Δ::HIS3MX + p353 (pRS316-SET5MYNDΔ)	This study
yEG778	BY4742	MATα set5Δ::HIS3MX + p354 (pRS316-SET5CTRΔ)	This study
yEG861	BY4742	MATα set5Δ::HIS3MX + p385 (pRS316-SET5CTR_mutA)	This study
yEG862	BY4742	MATα set5Δ::HIS3MX + p386 (pRS316-SET5CTR_mutD)	This study
yEG139	BY4741	MATa SET5::SET5-2xFLAG::KANMX	This study
yEG209	BY4741	MATa set5Δ::NATMX set1Δ::KANMX	42
yEG925	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p045 (pRS316)	This study
yEG926	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p130 (pRS316-SET5)	This study
yEG927	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p353 (pRS316-SET5MYNDΔ)	This study
yEG928	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p354 (pRS316-SET5CTRΔ)	This study
yEG929	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p355 (pRS316-SET5PSΔ)	This study
yEG930	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p385 (pRS316-SET5CTR_mutA)	This study
yEG931	BY4741	MATa set5Δ::NATMX set1Δ::KANMX + p386 (pRS316-SET5CTR_mutD)	This study

TABLE 3.

Plasmids used in this study

Plasmid	Description	Source or reference
p045	pRS316	70
p130	pRS316-SET5	This study
p353	pRS316-SET5MYNDΔ	This study
p354	pRS316-SET5CTRΔ	This study
p355	pRS316-SET5PSΔ	This study
p385	pRS316-SET5CTR_mutA	This study
p386	pRS316-SET5CTR_mutD	This study
p028	pGEX-6P1	GE Healthcare
p008	pGEX-6P1-GST-ySET5	42
p351	pGEX-6P1-GST-ySET5MYNDΔ	This study
p423	pGEX-6P1-GST-ySET5PSΔ	This study
p352a	pGEX-6P1-GST-ySET5CTRΔ	This study
p207	pGEX-6P1-GST-ySET5CTRΔ+PSΔ	This study
p403	pGEX-6P1-GST-ySET5CTR_mutA	This study
p404	pGEX-6P1-GST-ySET5CTR_mutD	This study
p398	pGEX-6P1-GST-SMYD3	This study
p399	pGEX-6P1-GST-SMYD3MYNDΔ	This study

Chromatin fractionation assay.

Yeast strains were grown to an optical density at 600 nm (OD₆₀₀) of ∼1.0 in 100 ml of synthetic complete medium lacking uracil (SC-URA) with shaking at 30°C. Cell pellets were processed for chromatin fractionation, as described previously (57–59). The fractions were incubated with sodium dodecyl sulfate (SDS) loading buffer at 95°C for 5 min, and equivalent volumes of each fraction were subjected to SDS-PAGE and immunoblotting.

Electrophoretic mobility shift assay.

The Widom 601 DNA sequence was PCR amplified from pGEM3z/601 (60; Addgene), purified, and concentrated using a PCR cleanup kit. Then, 1.5 pmol of 601 DNA was incubated with increasing amounts of recombinant proteins in EMSA buffer (100 mM Tris-HCl [pH 7.5], 750 mM NaCl). Reaction mixtures were incubated at 30°C for 30 min, loading dye was added, and DNA was resolved on a 1.5% agarose gel in 1× Tris-borate at 50 V for 1 h at 4°C. The gel was stained in Gel Green nucleic acid stain (Biotium) for 1 h and visualized using a Gel Doc EZ imager (Bio-Rad).

Recombinant protein purification.

Purification of GST-tagged proteins from E. coli was performed as described previously (20, 42). Protein expression was induced using 0.1 to 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside) at 37°C for 4 h for GST-Set5 variants, except GST-Set5_CTRΔ which was induced at 20°C overnight, and at 25°C for 4 h for GST-SMYD3. Cell lysates were prepared in 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5% NP-40, 10 mM phenylmethylsulfonyl fluoride, and a bacterial protease inhibitor mix (Thermo Scientific Pierce). The clarified lysate was bound to glutathione-Sepharose 4B (GE Healthcare) overnight at 4°C and eluted with 10 mg/ml glutathione (Sigma) in 50 mM Tris-HCl (pH 8.0).

In vitro methylation assays.

Methylation assays were performed as described previously (20, 42, 61). Briefly, different concentrations of GST-purified enzymes were added to 40 pmol of recombinant histone H4 (New England Biolabs) in 50 mM Tris-HCl (pH 8.0), 10% glycerol, 20 mM KCl, 5 mM MgCl₂, and either 2 μCi of [³H]SAM (Perkin-Elmer Life Sciences) or 0.3 mM unlabeled SAM (Sigma). Reaction mixtures were incubated at 30°C overnight, and proteins were resolved by SDS-PAGE, transferred to a polyvinylidene difluoride (PVDF) membrane, and imaged using autoradiography or subjected to immunoblotting. Total enzyme and substrate in each reaction were determined by SDS-PAGE and either silver staining (Silver Quest, Invitrogen) or immunoblotting. For methylation assays on H4 peptides, 0.5 μg of biotinylated H4 peptide (amino acids 1 to 23) was incubated with 2 μCi of [³H]SAM and the indicated amounts of recombinant GST-purified enzymes. Reaction mixtures were incubated at 30°C overnight, and peptides were pulled down using streptavidin magnetic beads (New England Biolabs) and washed with 50 mM Tris-HCl (pH 7.5), 300 mM NaCl, and 0.05% NP-40. Beads were resuspended in wash buffer and added to a scintillation vial containing scintillation fluid. Enzymatic activity was determined by reading the counts per minute (cpm) on a Beckman Coulter LS6500 scintillation counter.

Immunoblotting.

Immunoblots of subcellular fractionation samples, in vitro methylation assays, and other assays were performed after transfer from SDS-PAGE to a PVDF membrane. The following primary antibodies were used: polyclonal rabbit anti-Set5 (1:5,000) (42), mouse anti-H4 (1:1,000; Abcam, catalog no. ab31830), rabbit anti-H3 (1:1000; Abcam, catalog no. ab18521), rabbit anti-Hxk2-HRP (1:10,000; Novus Biologicals, catalog no. NBP2-44234), rabbit anti-H4K5me1 (1:1,000; 21st Century Biochemicals, Marlborough, MA), and rabbit anti-H3K8me1 (1:1,000) (42). The secondary antibodies used included IRDye 800CW-conjugated goat anti-rabbit (LI-COR, catalog no. 926-3211) and IRDye 800CW-conjugated goat anti-mouse (LI-COR, catalog no. 926-3210) antibodies. The rabbit anti-H4K5me1 antibody was generated in collaboration with 21st Century Biochemicals. The antibody was raised against an H4K5 monomethylated peptide, and specificity testing was performed using peptide dot blots with biotinylated histone peptides, as described previously (42, 62). Imaging was performed using a LI-COR Odyssey IR Western blotting scanner. For blots in which signal was quantified, Revert total protein stain (LI-COR) was used. Briefly, the blot was rinsed in water and incubated with total protein stain for 5 min. After staining, the blot was rinsed twice with wash solution and scanned on the 700 nm channel. The band intensity of protein was normalized to total protein stain intensity and quantified using the LI-COR Image Studio software.

Gene expression analysis by quantitative PCR.

Total RNA was extracted from 1.5 ml of an OD₆₀₀ of ca. 0.6 to 0.8 culture of yeast cells grown in SC-URA using the Masterpure yeast RNA purification kit (Epicentre). A Turbo DNA-free kit (Ambion) was used to eliminate genomic DNA, and an Accuris qMax cDNA synthesis kit cDNA was used to generate cDNA from 1 μg of total RNA. For quantitative PCR, 0.5 μl of the cDNA mixture was added to 1× qMax Green (Low Rox) PCR mix with gene-specific primers in a 10-μl reaction mixture and run in a Bio-Rad CFX384 real-time detection system. Gene expression values were determined relative to the control gene TFC1. Primer sequences for COS12, YGL262W, and TFC1 were reported by Jezek et al. (44). A minimum of five biological replicates are shown for each strain tested.

Phosphoproteomic analysis.

Wild-type (yEG001) and SET5::SET5-2×FLAG::KANMX (yEG139) yeast strains were grown in 2 liters of YPD at 30°C to late log phase, harvested, lysed, and subjected to anti-FLAG immunoprecipitations as described previously (61, 63). Phosphatase inhibitors (Roche phosSTOP tablet) were included in the lysis buffer to preserve protein phosphorylation. Set5-FLAG was eluted from the beads during two 30-min incubations with 0.2 mg/ml 3×FLAG peptide (Sigma) at 4°C. Samples were pooled, flash frozen in liquid nitrogen, and stored at –80°C until processing for mass spectrometry. Proteins were precipitated with 23% trichloroacetic acid and washed with acetone. Protein pellets were solubilized in 8 M urea–100 mM Tris (pH 8.5), reduced with 5 mM Tris(2-carboxyethyl)phosphine hydrochloride (Sigma-Aldrich), and alkylated with 55 mM 2-chloroacetamide (Fluka Analytical). Digested proteins were analyzed by four-step MudPIT using an Agilent 1200 G1311 quaternary pump and a Thermo LTQ Orbitrap Velos with an in-house-built electrospray stage (64). Protein and peptide identification and protein quantitation were performed using the Integrated Proteomics Pipeline (IP2; Integrated Proteomics Applications). Tandem mass spectra were extracted from raw files using RawConverter (65) with monoisotopic peak option. Peptide matching was done against the UniProt Saccharomyces cerevisiae protein database with reversed sequences and recombinant proteins using ProLuCID (66, 67) with a fixed modification of 57.02146 on cysteine and a differential modification of 79.9663 on serine, threonine, and tyrosine. Peptide candidates were filtered using DTASelect and the following parameters: –p 2 –y 1 –trypstat –sfp 0.01 –modstat –extra –pI –DM 10 –DB –dm –in –t 1 -a true –brief –quiet (65, 68). The phosphosite localization score was calculated (69).

Data availability.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD014756.