Malonylation of histone H2A at lysine 119 inhibits Bub1-dependent H2A phosphorylation and chromosomal localization of shugoshin proteins

Post-translational modifications of histones, such as acetylation and phosphorylation, are highly conserved in eukaryotes and their combination enables precise regulation of many cellular functions. Recent studies using mass spectrometry have revealed various non-acetyl acylations in histones, including malonylation and succinylation, which change the positive charge of lysine into a negative one. However, the molecular function of histone malonylation or succinylation is poorly understood. Here, we discovered the functions of malonylation in histone H2A at lysine 119 (H2A-K119) in chromosome segregation during mitosis and meiosis. Analyses of H2A-K119 mutants in Saccharomyces cerevisiae and Schizosaccharomyces pombe showed that anionic mutations, specifically to aspartate (K119D) and glutamate (K119E), showed mis-segregation of the chromosomes and sensitivity to microtubule-destabilizing reagents in mitosis and meiosis. We found that the chromosomal localization of shugoshin proteins, which depends on Bub1-catalyzed phosphorylation of H2A at serine 121 (H2A-S121), was significantly reduced in the H2A-K119D and the H2A-K119E mutants. Biochemical analyses using K119-unmodified or -malonylated H2A-C-tail peptides showed that H2A-K119 malonylation inhibited the interaction between Bub1 and H2A, leading to a decrease in Bub1-dependent H2A-S121 phosphorylation. Our results indicate a novel crosstalk between lysine malonylation and serine/threonine phosphorylation, which may be important for fine-tuning chromatin functions such as chromosome segregation.

neutralizes the positive charges of lysine residues. Second, acidic acylation, which encompasses malonylation 8 , succinylation 8 , and glutarylation 9 , turns the positive charge of lysine residues into a negative one. Third, hydroxylated acylation, which refers to 2-hydroxyisobutyrylation 10 and β-hydroxybutyrylation 11 , can allow the formation of hydrogen bonds with other molecules. Recent studies have suggested that hydrophobic acylation and hydroxylated acylation in histones are linked to active gene transcription, just like acetylation [10][11][12][13][14] . In contrast, the function of acidic acyl groups in histones is poorly understood. Moreover, crosstalk between non-acetyl acylations and other PTMs has not been reported yet.
Lysine succinylation and lysine malonylation (Fig. 1A) were originally discovered in non-histone proteins 15,16 , affording some insights about their effects. The succinylation of carbamoyl phosphatase synthase (CPS1) regulates its enzymatic activity 17 , while the malonylation in glyceraldehyde 3-phosphate dehydrogenase (GAPDH) is important for its enzymatic activity 18 . The following study described identification of these two PTMs in histones in several organisms, including Saccharomyces cerevisiae (budding yeast) 8 . Proteomic profiling studies have shown that proteins that undergo malonylation are typically different from the proteins undergoing acetylation or succinylation 18,19 , suggesting that histone succinylation and malonylation are regulated by distinct mechanisms. Removal of both succinylation and malonylation is catalyzed by Sirt5, an isoform of the sirtuin family of lysine deacylase 17 . One very recent study has revealed lysine acetyltransferase 2A (KAT2A, also known as Gcn5) to be a succinyltransferase 20 , whereas enzymes that catalyze lysine malonylation have not been identified to date.
In this study, we constructed and examined a series of malonyl-or succinyl-mimetic budding yeast mutants, and found the functions of malonylation in histone H2A at lysine 119 (H2A-K119) in chromosome segregation during mitosis and meiosis. H2A-K119 malonylation inhibits the interaction between Bub1 and H2A, leading to the loss of Bub1-dependent phosphorylation of H2A-S121, and consequently the failure of shugoshin proteins to localize to the centromere.

Results
Examining malonyl-or succinyl-mimetic mutants in budding yeast. We sought to address whether some instances of histone malonylation or succinylation were linked to chromosome segregation mechanisms during mitosis. Since the anionic character of malonylated or succinylated lysine can be partly mimicked by the mutation of lysine to glutamic acid or aspartic acid (Fig. 1A), we used plasmid shuffling to construct budding yeast histone mutants, in which a lysine residue was mutated to glutamic acid (Fig. S1A). In this study, we examined nine lysine residues that have been reported to be malonylated or succinylated in Saccharomyces cerevisiae (budding yeast) (Fig. 1A) 8 . H3K56E was excluded from our analysis because of its lethality 8 . Among the mutants we tested, only the H2A-K119E and H4K31E mutants showed sensitivity to benomyl, a microtubule depolymerizing agent which often enhances chromosome segregation defects, whereas no mutants were significantly sensitive to methylmethanesulfonate (MMS), a DNA-damaging agent (Fig. 1C). We decided to focus on the analysis of H2A-K119, since phosphorylation at proximal serine 121 (H2A-S121) is essential for the chromosomal localization of shugoshin (Sgo1) protein that ensures faithful chromosome segregation 21 .
H2A-K119 malonyl-mimetic mutants cause chromosome segregation defects in budding yeast. In order to examine whether the anionic property of H2A-K119 induced benomyl sensitivity, we further constructed several H2A-K119 mutants, including an aspartic acid mutant K119D as another malonyl-mimetic mutant, a glutamine mutant K119Q that neutralizes lysine's positive charge, and an arginine mutant K119R that keeps the positive charge. Immunoblotting experiments of yeast cell extracts showed that H2A mutants expressed at level comparable to the wild-type protein (Fig. S2A). Intriguingly, the K119D and K119E mutants showed benomyl sensitivity, whereas K119R and K119Q did not (Figs 2A, S2B). Consistently, we observed a high frequency of chromosome mis-segregation in benomyl-treated K119D and K119E mutants, but not in the benomyl-treated WT or K119R mutant (Figs 2B, S2B). As the H2A-S121A mutant, in which serine 121 of H2A was substituted with a non-phosphorylatable alanine, showed benomyl sensitivity and chromosome mis-segregation with a tendency similar to the K119D mutant, we hypothesized that H2A-K119 malonylation affected phosphorylation at H2A-S121. To address the hypothesis, we observed the centromeric localization of Sgo1-GFP in metaphase cells, which requires H2A-S121 phosphorylation 21 . Compared to the wild-type, Sgo1-GFP localization was dramatically reduced in the H2A-K119D and K119E mutants, as was observed in the H2A-S121A mutant (Fig. 2C). On the other hand, Sgo1-GFP localization in the K119R mutant was only slightly reduced (Fig. 2C). These data suggest that the negative charge at H2A-K119 may inhibit the H2A-S121 phosphorylation that is required for chromosomal Sgo1 localization, and thus induce chromosome mis-segregation during mitosis. The effects of H2A-K119 malonyl-mimetic mutation are conserved in fission yeast. Since H2A-K119 is widely conserved among several eukaryotes including fission yeast (Fig. 3A), we next constructed H2A-K119 mutants in fission yeast (Fig. S1B) and examined their phenotypes. ChIP (Chromatin immunoprecipitation) analysis showed that H2A mutants deposited at the centromeres as well as the wild-type H2A (Fig. S3A). We found that malonyl-mimetic K119D and K119E mutants showed sensitivity to a microtubule depolymerizing agent (TBZ) in the same way the H2A-S121A mutant had (Fig. S3B), suggesting that H2A-K119-related mechanisms are conserved between the two yeasts. In contrast, the H2A-K119R mutant showed weaker TBZ-sensitivity than these mutants. We next addressed whether H2A-K119 malonyl-mimetic mutations affected chromosomal localization of shugoshin proteins in fission yeast. In contrast to budding yeast that has only shugoshin Sgo1, fission yeast has two shugoshin paralogs, Sgo1 and Sgo2 (Fig. 3B) 22,23 . Fission yeast Sgo1 is transiently expressed during the first meiotic division (meiosis I) and maintains centromeric cohesion through the recruitment of the protein phosphatase 2A complex 22,24 , whereas Sgo2 is expressed during both mitotic and meiotic divisions and ensures proper attachment between the kinetochores and microtubules by recruiting the Aurora B kinase complex 23,25 . Intriguingly, we found that both Sgo2 centromeric localization during mitosis and Sgo1 centromeric localization during meiosis I were dramatically delocalized in the K119D and K119E mutants as well as the H2A-S121A mutant ( Fig. 3C-E). The lysine to arginine mutation (K119R) did not significantly affect shugoshin localization ( Fig. 3C-E). The mutation at H2A-K119 may not affect protein level of shugoshin proteins, since the level of Sgo2-GFP in any H2A mutants was comparable to the wild-type protein (Fig. S3C). Centromeric Sgo1 is essential for the protection of sister chromatid cohesion during meiosis I, so Sgo1 delocalization leads to the non-disjunction of sister chromatids at the second meiotic division (meiosis II) 21,22 . Consistently, sister chromatid non-disjunction at meiosis II was significantly increased in the K119D and K119E mutants as well as the H2A-S121A mutant (Fig. S3D). These results suggest that H2A-K119 malonylation-mediated inhibition of shugoshin localization is likely conserved between budding yeast and fission yeast.
H2A-K119 malonylation inhibits H2A-S121 phosphorylation by Bub1. As H2A-S121 phosphorylation is catalyzed by Bub1 kinase 21 , we addressed whether malonyl-mimetic mutations inhibited H2A phosphorylation by Bub1 in vitro. For this purpose, we purified fission yeast's recombinant Bub1C protein (SpBub1C), made up of the kinase domain and the N-terminus extension, which has been reported to be important for the kinase activity 26 of fission yeast Bub1 (Fig. 4A). The recombinant proteins of fission yeast H2A (SpHta1) were also purified. An in vitro kinase assay using these recombinant proteins showed that the H2A-K119D and the H2A-K119E mutants, as well as the H2A-S121A mutant, were not significantly phosphorylated by Bub1, whereas H2A-K119R mutant proteins were phosphorylated by Bub1 (Fig. 4B). An acetyl-mimetic H2A-K119Q mutation slightly inhibited Bub1-mediated H2A phosphorylation (Fig. 4B). We repeated the in vitro kinase assay using budding yeast's H2A (ScHta1) and Bub1 kinase (ScBub1C) proteins and obtained similar results (Fig. S4A,B). Therefore, we concluded that the K119 malonyl-mimetic mutation inhibits H2A-S121 phosphorylation by Bub1.
Although the effects of H2A-K119 malonyl-mimetic mutations are now clearer, it might be possible that malonyl-mimetic mutations do not completely mimic malonylated lysine. Therefore, we next sought to address the effects of true malonylation at H2A-K119 on H2A-S121 phosphorylation by Bub1. For this purpose, we synthesized H2A C-tail peptides (V109-K126 of SpHta1) with and without malonylation at K119. The mixture of both peptides was reacted with or without SpBub1C in the presence of ATP, and the phosphorylated peptides were analyzed by LC-MS/MS (Fig. 4C). S121 phosphorylation in the H2A C-tail peptides containing unmodified K119 (blue in Fig. 4D) was detected only when Bub1C was added, indicating that the peptide may be a good substrate for Bub1C (Fig. 4D). On the other hand, the amount of S121 phosphorylation was much lower in the H2A C-tail peptides that contain malonylated K119 (red in Fig. 4D). To exclude that the difference of ionization efficiency in mass spectrometry may affect our analysis, we compared that ionization efficiency between the two phosphorylated peptides (VPNINAHLLPKTSphosGRTGK and VPNINAHLLPKmalTSphosGRTGK) and found that it was similar (see Methods). Therefore, we concluded that Hta1-K119 malonylation inhibits the phosphorylation of H2A-S121 by Bub1. This result is consistent with the results from the H2A-K119 malonyl-mimetic mutant protein experiments, suggesting that the phenotypes of H2A-K119 malonyl-mimetic mutants likely reflect the effects of malonylated K119. H2A-K119 malonylation inhibits interaction between Bub1 and H2A. Finally, we addressed how H2A-K119 malonylation inhibits H2A-S121 phosphorylation by Bub1. We hypothesized that H2A-K119 malonylation might inhibit interaction between Bub1 and H2A. To investigate this hypothesis, we synthesized N-terminally biotinylated H2A-C-tail peptides (V109-K126 of SpHta1) with and without K119 malonylation (Fig. 5A), and examined the interactions between recombinant Bub1C proteins and these peptides. The K119-unmodified peptides efficiently pulled down Bub1C, whereas the K119-malonylated peptides did not (Fig. 5B). We also conducted a similar pull-down assay using budding yeast's H2A-C-tail peptides (L109-K126 of ScHta1, Fig. S5A) and Bub1C protein and obtained similar results (Fig. S5B). These data indicate that H2A-K119 malonylation inhibits the interaction between Bub1 and H2A, which is likely the reason why H2A-K119 malonylation inhibits Bub1-dependent phosphorylation of H2A.
It has been reported that P + 1 loop within the kinase domain of human Bub1 is important for interaction with its substrate Cdc20 26 . On the other hand, it remains unclear how Bub1 recognizes H2A. As our data suggest that the cationic property of K119 is important for the interaction between Bub1 and H2A, we hypothesized that anionic amino acids near the P + 1 loop of Bub1 may be critical for recognizing H2A as a substrate via electrostatic interaction. We focused on the glutamic acid (E929 of fission yeast Bub1) that is located at the end of the P + 1 loop, as this glutamic acid is widely conserved among budding yeast, fission yeast, and other eukaryotes including human (Fig. 5C). Intriguingly, the E929R mutation in SpBub1 dramatically reduced interaction between Bub1 and H2A (Fig. 5D), suggesting that interaction between the P + 1 loop of Bub1, which contains the conserved glutamic acid, and H2A C-terminus tail, which contains K119, may be crucial for H2A recognition by Bub1 (Fig. 6).

Discussion
For faithful chromosome segregation during mitosis and meiosis, Bub1 catalyzes phosphorylation of histone H2A at serine 121 in fission yeast and budding yeast (or threonine 120 in human) and recruits shugoshin proteins at the centromeres 21,[27][28][29] . In this study, we report that malonylation of histone H2A at lysine 119 has an inhibitory role in the regulation of H2A phosphorylation. Although the biological significance of the mechanisms in vivo remains to be elucidated, it would be a potentially important regulation of chromosome segregation. A model of the functions of H2A-K119 malonylation is shown in Fig. 6. Bub1 recognizes H2A via electrostatic interaction between Bub1's kinase domain, which contains the conserved glutamic acid, and H2A C-terminus tail, which contains K119, and catalyzes phosphorylation at serine 121 of H2A, which is read by shugoshin proteins. When H2A-K119 is malonylated, Bub1 fails to recognize H2A, leading to the loss of H2A phosphorylation and thus the delocalization of centromeric shugoshin and the subsequent mis-segregation of chromosomes during mitosis and Our study demonstrates the first example of crosstalk between non-acetyl acylation and phosphorylation. Inhibition of H2A-S121 phosphorylation by H2A-K119 malonylation was observed when H2A-C tail peptides were used as substrates, so this cross-regulation occurs in cis. So far, a few examples of cross-regulation in cis between acetylation and phosphorylation have been reported. H3-S10 phosphorylation promotes acetylation of H3-K14 by Gcn5 in vitro 30 . This is likely mediated by the direct interaction of Gcn5 with the phosphorylated H3-S10 residue. In contrast, H3-S10 phosphorylation completely blocks acetylation of H3-K9 31 , probably by blocking the access of HATs to H3-K9. Acetylation of H2B-K11 inhibits phosphorylation of H2B-S10 by Ste20 32 . Our data suggest that not only acetylation, but also non-acetyl acylations at lysine residues, may influence the phosphorylation of nearby serine/threonine residues. The fact that various non-acetyl acylations have been identified at a number of lysine residues in histones indicates that the occurrence of cross-regulation is much more widespread than we have previously supposed. For example, various types of non-acetyl acylations (crotonylation, butyrylation, succinylation, 2-hydroxyisobutyrylation, and β-hydroxybutyrylation) have been detected at H3K27 4,11 , possibly inhibiting H3S28 phosphorylation. It is therefore important to address to what extent different types of acylation at lysine residues can affect nearby phosphorylation.
It is unknown whether H2A-K119 malonylation is catalyzed by a specific malonyltransferase or introduced by malonyl-CoA via a non-enzymatic mechanism. In either case, the level of H2A-K119 malonylation may fluctuate according to metabolic conditions, as intracellular concentrations of malonyl-CoA depend on the metabolic status of the cell 33,34 . Establishing methods to quantitatively measure the stoichiometry of H2A-K119 malonylation will be important for further understanding of our findings. Our results indicate that H2A-K119 malonylation must be removed for proper centromeric localization of shugoshin proteins and faithful chromosome segregation during mitosis and meiosis. As Sirt5 can remove lysine malonylation in human cells, a yeast sirtuin homolog, such as Sir2, might remove H2A-K119 malonylation in budding yeast or fission yeast. Both H2A-K119 and H2A-S121 are widely conserved between yeasts and higher eukaryotes including humans (Fig. 3A), suggesting that this novel crosstalk might be conserved as well in humans. In the congenital metabolic disease malonyl-CoA decarboxylase deficiency (MCD), malonyl-CoA fails to be broken down into acetyl-CoA and carbon dioxide, and the resulting accumulation of cellular malonyl-CoA may lead to excess histone malonylation. Therefore, investigating whether H2A-S121 phosphorylation and shugoshin localization are defective in MCD patients may be important to understanding the molecular mechanisms of such metabolic diseases.

Construction and analyses of budding yeast strains.
All strains used in this study are listed in Table S1. To generate the series of point mutants of H2A and H2B, plasmids bearing a centromere sequence (CEN), a HIS3 or ADE2 marker, and gene cassettes of HTA1 and HTB1 carrying each mutation were transformed into a host strain in which the endogenous H2A and H2B genes had been deleted, and a plasmid containing wild type HTA1 and HTB1 genes with an URA3 marker had previously been introduced. First, transformants were selected by the introduced marker, and then counter-selected by 5-FOA to choose clones which had lost the originally introduced plasmid bearing wild type HTA1 and HTB1. To generate the series of point mutants of H3 or H4, we followed the same strategy, using plasmids with a centromere sequence, a TRP1 marker, and HHT2 and HHF2 gene cassettes carrying each mutation, and a host strain in which the endogenous H3 and H4 genes had been deleted and a plasmid containing wild type HHT2 and HHF2 genes with an URA3 marker already introduced. Tagging of endogenous SPC42 + by mCherry or SGO1 + by GFP was performed using the PCR-generated DNA constructs as previously reported 22 . To measure the frequencies of non-disjunction of CEN5-GFP, budding yeast cells were cultured in YPD medium for 7 hours at 29 °C in the presence of 10 µg/mL of benomyl, fixed by methanol, and stained by DAPI and Calcofluor. To observe mitotic ScSgo1-GFP localization, budding yeast cells expressing Sgo1-GFP and Spc42-mCherry were grown in YE4S medium at 25 °C until the OD 590nm reached 0.15-0.3, fixed by methanol, and stained by DAPI.
Construction and analyses of fission yeast strains. All strains used in this study are listed in Table S1.
Every h2a mutant strain was generated as previously described 21 . Specifically, the genomic hta1 and hta2 DNA fragments carrying each mutation were amplified by PCR, and then transformed into hta1::ura4 + and hta2::ura4 + strains respectively. Individual single integrants were selected by 5-FOA and confirmed by PCR. After crossing both integrants of hta1 and hta2, double mutants were selected. To visualize tubulin, P adh15 (a weak version of the adh1 + promoter)-mCherry-atb2 + was integrated at the Z locus using the nat r marker. To measure the centromeric Sgo2-GFP signal during mitosis, fission yeast cells expressing Sgo2-GFP and mCherry-Atb2 were grown in YE4S medium until the OD 590nm reached 0.5, fixed by methanol, and stained by DAPI. To observe SpSgo1-GFP localization during meiosis I, homothallic fission yeast strains were grown in YE4S medium at 25 °C until the culture became saturated, spotted on SPA plates, and incubated at 25 °C for 11 hours. Zygotes were arrested at metaphase I by shutting down both slp1 + and cut23 + genes by the substitution of their promoter to that of the rad21 + gene. To monitor meiotic cen2-GFP segregation of fission yeast, haploid strains of opposite mating types with and without cen2-GFP were independently grown in YEA medium at 26.5 °C until the OD 590nm reached 0.2-0.3, spotted on SPA plates after being combined, and incubated at 26.5 °C for 12 hours.

Construction of plasmids.
To express GST-ScBub1C and GST-SpBub1C, the C-terminus of budding yeast BUB1 + ORF (631 to 1021 in the amino acid sequence) or fission yeast bub1 + ORF (633 to 1041 in the amino acid sequence) was amplified by PCR and cloned into the pGEX6p-2 vector. To express GST-ScHta1 and GST-SpHta1, the full-length of either budding yeast HTA1 + ORF or fission yeast hta1 + ORF were cloned as described above. All of the point mutations of bub1 and histone genes were introduced by PCR mutagenesis with PrimeSTAR Max DNA polymerase (Takara).
Purification of recombinant proteins. The recombinant proteins of GST-SpBub1C, GST-ScBub1C, GST-SpHta1, and GST-ScHta1 were produced in E. coli (BL21C + ) using the pGEX6p-2 vector. All of the GST-fused proteins were affinity-purified by glutathione sepharose 4B resin (GE Healthcare) and eluted by adding glutathione. GST-SpBub1C, GST-Bub1C (E929R), and GST-ScBub1C were further purified by the gel-filtration system AKTA pure M1 with Superdex 200 10/300 GL column (GE Healthcare). In vitro phosphorylation assay with recombinant Hta1 proteins. The mixture of recombinant kinase GST-Bub1C and substrate GST-Hta1 (wild type or mutants) were incubated in kinase buffer (50 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 1% Triton X-100) for 30 min at 30 °C in the presence of [γ-32 P] ATP. Reacted proteins were separated by SDS-PAGE. Loading amount of GST-Hta1 proteins was checked by CBB staining and the phosphorylation was detected by autoradiography.
In vitro phosphorylation assay with fission yeast's Hta1-C-tail peptides. A 1 mM mixture of unmodified and K119-malonylated SpHta1-C-tail peptides was incubated in kinase buffer containing 100 µM ATP and 1 mg/ml BSA with or without recombinant GST-SpBub1C for 30 min at 30 °C. After the reaction, GST-SpBub1C was immediately denatured and precipitated by adding 4% trichloroacetic acid. After centrifugation, the supernatant was analyzed by LC-MS/MS in a 50-fold dilution with Milli Q.

LC-MS/MS analyses.
LC-MS/MS analyses were conducted using an AB Sciex Triple TOF 4600 equipped with an Eksigent ekspert microLC 200. LC was carried out as follows: 3C18-CL-120 column (0.5 mm l.D × 100 mm) using a linear gradient of 2-35% acetonitrile with 0.1% formic acid (v/v) versus water with 0.1% formic acid (v/v) over 6 min at 40 °C with a flow rate of 20 µL min −1 after 1 min equilibration. The amount of the injected samples was 5 µL. The eluent was monitored by on-line quadrupole time-of-flight mass spectrometer MS, operated in positive ion mode. Data-dependent analyses were conducted for characterization of synthetic peptides. Data-independent analyses were conducted for analysis of in-vitro phospholylation, with targeted precursor ions and collision energies of 496.27 (CE = 25) for the S121phos peptide and 517.27 (CE = 25) for the S121phos/K119mal peptide. Data analysis was carried out using PeakView software (AB Sciex, version 1.2.0.3). We compared ionization efficiency between the two phosphorylated peptides (VPNINAHLLPKTSphosGRTGK and VPNINAHLLPKmalTSphosGRTGK) as follows: The two phosphorylated peptides were mixed with BSA (0.125 mg/mL) and unphospholylated peptides (VPNINAHLLPKTSGRTGK and VPNINAHLLPKmalTSGRTGK, 2 mM each) in buffer (20 mM Tris pH 7.5, 4 mM MgCl 2 , 1% Triton). After incubation at 30 degree for 30 min, TCA was added to the final concentration of 4%, and incubated on ice for 30 min. After centrifugation (14000 rpm, 10 min), supernatant was analyzed by LC-MS/MS. Peak area was calculated on the software, and the obtained ratio of the peak area of unmalonylated one to malonylated one was 1.17 ± 0.05 (N = 5). Therefore, we considered that the difference of ionization efficiency did not affect our conclusion by LC-MS-based experiments. Data Availability. The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.