Revealing the protein propionylation activity of the histone acetyltransferase MOF (males absent on the first)

Short-chain acylation of lysine residues has recently emerged as a group of reversible posttranslational modifications in mammalian cells. The diversity of acylation further broadens the landscape and complexity of the proteome. Identification of regulatory enzymes and effector proteins for lysine acylation is critical to understand functions of these novel modifications at the molecular level. Here, we report that the MYST family of lysine acetyltransferases (KATs) possesses strong propionyltransferase activity both in vitro and in cellulo. Particularly, the propionyltransferase activity of MOF, MOZ, and HBO1 is as strong as their acetyltransferase activity. Overexpression of MOF in human embryonic kidney 293T cells induced significantly increased propionylation in multiple histone and non-histone proteins, which shows that the function of MOF goes far beyond its canonical histone H4 lysine 16 acetylation. We also resolved the X-ray co-crystal structure of MOF bound with propionyl-coenzyme A, which provides a direct structural basis for the propionyltransferase activity of the MYST KATs. Our data together define a novel function for the MYST KATs as lysine propionyltransferases and suggest much broader physiological impacts for this family of enzymes.

Short-chain acylations of lysine residues in cellular proteins, such as acetylation, propionylation, and crotonylation, are reversible posttranslational modifications (PTMs) 5 that modulate functions and properties of protein targets (1). The variety of lysine acylation has been proposed to correlate with divergent biological outputs (2,3). Etiologically, lysine acylations rely on acyl-CoA molecules to serve as the acyl donors, which are key metabolic intermediates in the Krebs cycle, fatty acid oxidation, and amino acid degradation (4). The level of acyl-CoA intermediates fluctuates with nutritional status and altered activities of metabolic enzymes. Toxic accumulation of acyl-CoA molecules is associated with a broad variety of complications (5,6). A typical example is the propionic acidemia, an autosomal recessive metabolic disorder caused by the deficiency of propionyl-CoA carboxylase (7). Increases of propionyl-CoA and protein lysine propionylation level were reported in propionyl-CoA carboxylase-deficient patient fibroblast cells, but the disease-relevant targets and the corresponding regulatory enzymes have not been defined (5). It is also increasingly recognized that the gut microbiota-related short-chain fatty acids interact with and modulate the mammalian epigenetic machinery, especially histone acylation and methylation (8,9). Nevertheless, it remains poorly understood how the alterations of acyl-CoA abundance and protein acylation levels are causatively linked with the related disease phenotypes. Understanding the regulatory mechanisms of short-chain lysine acylation, such as lysine propionylation, will have profound significance in elucidating the pathophysiological mechanisms of metabolic diseases.
To understand the mechanistic links of acylation diversity with varied physiological outputs, a significant challenge is to resolve how different acylations are deposited, interpreted, and erased in the cell. In eukaryotes, protein acetylation marks are introduced by lysine acetyltransferases (KATs), which are classified into three major families based on primary amino acid The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S3 and Tables S1-S5. The atomic coordinates and structure factors (code 5WCI) have been deposited in the Protein Data Bank (http://wwpdb.org/). 1 Supported in part by a scholarship from the Chinese Scholarship Council. 2 Supported by National Institutes of Health Grant R01CA172495 and Department of Defense Grant W81XWH-15-1-0507. 3 Supported by National Institutes of Health Grants GM105933, DK107868, and GM115961 and the Nancy and Leonard Florsheim Family Foundation. 4 To whom correspondence should be addressed. E-mail: yzheng@uga.edu. 5 The abbreviations used are: PTM, post-translational modification; KAT, lysine acetyltransferase; KPT, lysine propionyltransferase; HEK293T, human embryonic kidney 293T; MOF, males absent on the first; CBP, CREBbinding protein; CREB, cAMP-response element-binding protein; CPM, 7-diethylamino-3-(4Ј-maleimidylphenyl)-4-methylcoumarin; PDB, Protein Data Bank.
cro ARTICLE sequence and domain organization, including the MYST, the PCAF/GCN5, and the p300/CBP family (10). These KATs, together with lysine deacetylases, orchestrate the dynamics of lysine acetylation in the cell. Numerous studies have shown that aberrant expression and dysfunction of KATs are associated with various kinds of disease phenotypes, such as inflammation, neurodegenerative disorders, and cancers (11)(12)(13). In 2007, Chen et al. (14) reported that p300 and CBP can propionylate the human proteins histone H4 and p53 with high-resolution mass spectrometry. Later on, GCN5 and PCAF, members of the GNAT superfamily, were also found to possess propionyltransferase activity (15)(16)(17). These findings largely increased mechanistic complexity and diversity of the KAT-related physiological activities. The MYST KATs represent a distinct family whose sequence and function are dramatically different from the members of the p300/CBP and PCAF/GCN5 families. It remains an open question whether or not the members of the MYST family possess lysine propionyltransferase (KPT) activity ( Fig. 1) (18). In this work, we conducted a combined suite of biochemical, cellular, and structural studies and demonstrated that the MYST KATs have genuine KPT activity both in vitro and in cellulo.

Propionyl-CoA is an abundant metabolite in human cells
Several novel acylations other than acetylation on protein lysine residues have been identified, such as propionylation, butyrylation, crotonylation, succinylation, and malonylation (14,19,20). Thus far, the cellular abundance of the acyl-CoA molecules for corresponding acylations remains poorly characterized. To understand the regulation of protein propionylation, we first quantified and compared the cellular abundance of endogenous propionyl-CoA and acetyl-CoA in the human embryonic kidney 293T (HEK293T) cells using an LC-MS/MS method we recently developed (42). Following cell culture, acyl-CoA molecules were extracted in 50% methanol and subjected to LC-MS/MS detection. Deuterated acetyl-CoA and propionyl-CoA were used as internal standards for calibration. Triplicate experiments were performed and showed that the cellular propionyl-CoA concentration is 92 ng/mg cellular proteins, and acetyl-CoA is 754 ng/mg cellular proteins. Thus, the abundance of propionyl-CoA is about 12% that of acetyl-CoA (Fig.  2). Considering that acetyl-CoA is a rich and principal metabolite for cell growth, propionyl-CoA, even 8 times less abundant than acetyl-CoA, has the potential to serve as an ample source for lysine propionylation.

The recombinant MYST KATs have strong lysine propionyltransferase activity
A few studies have demonstrated that GCN5/PCAF and p300/CBP members were able to carry out lysine propionylation, butyrylation, crotonylation, etc. in addition to their intrinsic acetylation activity (3,14,17). In contrast, although the MYST enzymes represent the largest KAT family in human cells, their cofactor promiscuity is much less studied. Very recently, MOF was shown to carry out butyrylation and crotonylation on histone substrates (21,22). This evidence hints at the cofactor promiscuity of the MYST enzymes. Herein, we thoroughly investigated the novel acyltransferase activity of MYST KATs comparatively and from different perspectives. First, we screened and compared the KPT activity of the three families of KATs on their histone H3 or H4 peptide substrates (H3-20 and H4-20, the N-terminal 20-amino acid sequence of histone H3 and H4). A fluorogenic assay was used to determine the kinetic constants of KATs, including K m and k cat , with respect to acetyl-CoA or propionyl-CoA under the initial velocity condition (23). The k cat /K m ratio was calculated to quantitatively compare the KPT and KAT activities of individual KATs. In agreement with the previous studies showing that PCAF/GCN5 and p300/CBP members possess KPT activity (15,17,24), our kinetic measurement showed that the KPT activity of GCN5 and HAT1 is almost equally strong compared with their KAT activity, whereas the KPT activity of PCAF and p300 is about 40 and 30% of their acetyltransferase activity (Table S1). Interestingly, the MYST KAT members, including MOF, HBO1, and MOZ, show strong KPT activity, with ratios of their KPT/KAT activity of 0.84, 0.99, and 1.04, respectively. The MYST KATs Tip60 and MORF also showed appreciable KPT activities, which are 12 and 17% of their KAT activity. After we demonstrated the propionyltransferase activity of the KAT enzymes on histone peptides, we then tested the KPT activity of the MYST KATs on the proteome of the 293T cell using an anti-propionyllysine antibody (Fig. 3). Compared with the untreated cell lysate, sole propionyl-CoA co-incubation did not induce increase of lysine propionylation on cellular proteins, whereas treatment of the cell lysate with propionyl-CoA and individual MYST KATs induced extensive propionylation of multiple histone and non-histone proteins. In particular, MOF showed the strongest KPT activity on cellular proteins. Together, the steady-state kinetic data and the Western blot analysis of whole-protein propionylation strongly demonstrate that the MYST KATs are bona fide KPTs. Next, we focus on MOF for further biological, structural, and biochemical investigation of the novel KPT activity of the MYST enzymes.

MOF and lysine propionylation MOF acetylates and propionylates largely shared substrates with subtle variation
After proving the KPT activity of the MYST enzymes on cellular histones and non-histones, we investigated the similarity of MOF-mediated sub-acetylome and sub-propionylome. Peptide proteomic analyses were applied to identify MOF acetylation and propionylation sites on the H4-20 peptide. Monoacylated H4-20 is the major product from the MOF-catalyzed acetylation reaction, and the diacylated H4-20 is also present with lower yield than the monoacylated product (Fig.  S1A). Notably, the acylation sites were slightly different: Lys-8, Lys-12, and Lys-16 were found acetylated, whereas Lys-8 and Lys-16 were propionylated. In further detail, monoacetylation of H4-20 occurs on Lys-8 and Lys-16, whereas diacetylated peptides have acetylated Lys-8 and Lys-12, Lys-16 and Lys-12, and Lys-8 and Lys-16, respectively. For H4-20 propionylation, Lys-8 -and Lys-16 -monopropionylated peptides were detected, and dipropionylated peptides have propionylated Lys-8 and Lys-16 simultaneously (Fig. S1 (B and C) and Table S2). Thus, both Lys-8 and Lys-16 are prone to undergo both acetylation and propionylation by MOF, whereas Lys-12 may be a weak acetylation site of MOF. Next, we tested MOF substrate specificity on the histone proteins and nucleosomes with Western blot analysis. Using free histones as substrates, MOF was able to modify all of the four core histones for both acetylation and propionylation (Fig. 4A), whereas MOF cannot acylate histone H3 on the nucleosomes (Fig. 4B). Both MOF catalytic domain and full-length MOF showed the same substrate specificity, suggesting that the chromodomain did not affect the interaction between MOF and nucleosomes. Accumulated studies indicated that the histone proteins are not the only substrate of MOF (2,(25)(26)(27); therefore, we next explored the MOF substrate profile on the cellular proteome. Because the extracted cellular proteome has gone through acylations by endogenous KAT enzymes, it is hard to accurately discriminate MOF sub-acylome from the bulk acylome with immunoblot assays. Thus, a radiometric gel assay was performed to compare the substrate acetylation and propionylation driven by MOF (Fig. 5). In accord with the Western blot results shown in Figs. 3 and 4, little non-enzymatic labeling on lysine residues driven by the chemical reactivity of acetyl-CoA and propionyl-CoA was observed. Treatment of the cell lysate with acyl-CoA and MOF together induced strong acylation  of multiple proteins from the cellular proteome. A highly identical modification pattern was observed between lysine acetylation and propionylation, which is suggestive that MOF acetylome and propionylome may largely overlap. MOF autoacetylation was previously reported by Yang et al. (28). We observed that MOF underwent autopropionylation with similar activity compared with autoacetylation. The fact that multiple non-histone proteins were being acylated by MOF is in agreement with the previous findings that MOF targets both histone and non-histone proteins (2, 26, 29 -31).Overall, MOF may have highly identical acetylome and propionylome, whereas the modification levels at individual lysine residues could differ to varying degrees.

MOF is a genuine lysine propionyltransferase in cellulo with a distinct substrate profile
The above biochemical experiments demonstrated that the recombinant MYST KATs catalyze propionylation of histone and non-histone proteins in vitro. We next focused on testing MOF propionyltransferase activity in the cell and determining the MOF propionylome. A model of MOF overexpression was created in HEK293T cells by lentivirus transient transfection (Fig. 6A). Following cell culture and cell lysate extraction, the acylation levels of cell proteome were examined with Western blot analyses. Both lysine acetylation and propionylation in extracted histones were moderately up-regulated in the presence of MOF overexpression (Fig. 6B). Particularly, MOF overexpression induced increased levels of acylation on both histone H4 and H2A/2B, whereas histone H3 acylation did not change, which is consistent with the in vitro data shown in Fig.  4B. Therefore, H3 is probably not a significant in cellulo substrate of MOF. Of note, H4 Lys-16 propionylation was upregulated with MOF overexpression (Fig. 6C), in accord with the previous knowledge that H4 Lys-16 is a primary substrate of MOF (32). We also observed that multiple nonhistone proteins were acetylated and propionylated at enhanced levels in response to MOF overexpression, and the Western blot profiles of substrate propionylation and acetylation are highly identical (Fig. 6D). This result is in agree-

MOF and lysine propionylation
ment with the radiometric gel assay (Fig. 5) and strengthens our hypothesis that MOF have shared substrates for KAT and KPT activities. In addition, we detected the change of lysine propionylation level on cellular proteome in response to varying propionyl-CoA concentrations (Fig. S2). The 293T cells were treated with sodium propionate, which can be converted to propionyl-CoA in the cells through the acyl-CoA synthase pathway (33), and cellular proteins were extracted for Western blot detection. Treatment of both the control and MOF-overexpressed cells with sodium propionate induced increase of lysine propionylation on both histone and non-histone proteins. Interestingly, the Western blot patterns show an appreciable difference between the control cells and MOF-overexpressed cells upon propionate treatment, suggesting that some proteins are only propionylated in the MOF-overexpressed cells. These results indicate that MOF may have a distinct substrate propionylome, and the unique MOF function cannot be compensated by other KPT enzymes. Moreover, we knocked down MOF expression in the 293T cells and found that the lysine propionylation level was slightly decreased upon down-regulation of MOF expression (Fig. S3). Together, MOF displayed in cellulo KPT activity on both histone and non-histone proteins. The propionylation substrate profile of MOF is highly similar to that of acetylation but is partially distinct from other KPTs.

Proteomic profiling of MOF-mediated propionylome
Tandem MS was carried out to further investigate the propionylation substrates of MOF in the 293T cellular proteome. The MOF-overexpressed and control cells were treated with deuterated sodium propionate (sodium propionate_D5), followed by cell lysate extraction, affinity enrichment with anti-propionyllysine antibody, and LC-MS/MS analyses. As shown in Fig. 7A, the number of d 5 -propionylated lysine sites (Kpr_D5) in cellular proteins greatly increased, from 36 to 60, upon MOF overexpression. Excitingly, 43 unique propionylation sites were identified in MOF-overexpressed cells that were not present in the control cells. These emerging sites most likely represent bona fide cellular substrates of MOF ( Fig. 7B and Table S3). These proteins include known MOF substrates, such as p53 and MOF (2,28). H4 Lys-16 was found in the propionylation list, which is consistent with the in vitro and in cellulo studies (Table S2 and Fig. 6) and also matches the previous literature, showing that MOF can acetylate H4 Lys-16 in mammalian cells (31). Additionally, we identified several new histone propionylation sites at H4 Lys-12 and Lys-16 and H2A Lys-5 and Lys-9 in the MOF-overexpressed cells, which is in agreement with the Western blot data indicating that the propionylation levels of histone H2A/H2B and H4 were up-regulated upon MOF overexpression (Fig. 6B). It is worth mentioning that 17 of the total 60 Kpr_D5 sites in the MOF-overexpressed cells were found in the control cells as well. These shared modification marks could be the common substrates of MOF and other KPTs. Also, 19 Kpr_D5 sites were only identified in the control cells and not in the MOF-overexpressed cells. These lysine sites might be the substrates of other KPTs, such as GCN5 or p300, and their propionylation levels are not high enough to be detected in the MOF-overexpressed cells due to the strong competition from the overexpressed MOF. Overall, these proteomic data demonstrate that MOF propionylates a wide range of cellular proteins. The finding that MOF propionylates multiple protein substrates indicates broader involvement of MOF in biological regulation and its versatile functions.

The crystal structure of MOF in complex with propionyl-CoA further supports the KPT activity of the MYST enzymes
To gain structural insight into the KPT activity of MOF, we solved the crystal structure of the MOF catalytic domain in complex with propionyl-CoA at 1.78 Å resolution (Table S4 and Fig. 8A). Binding of propionyl-CoA to the MOF catalytic domain did not cause any major structural change in MOF structure, as the MOF⅐propionyl-CoA binary structure is almost identical to that of the previously solved MOF⅐acetyl-CoA complex (PDB entry 2GIV). The C␣ of these two structures can be superimposed with root mean square deviation of just 0.17 Å. Similar to the MOF⅐acetyl-CoA structure, the overall fold of MOF⅐propionyl-CoA adopts an elongated shape. The structure is composed of a central core region, flanked on opposite sides by N-and C-terminal domains (Fig. 8A). The catalytic site of the MYST family of KATs involves a catalytic cysteine (Cys-316 in MOF) and a conserved glutamate (Glu-350 in MOF), which functions as general base for enzyme catalysis. An autoacetylated lysine at the active site (Lys-274 in MOF) is also critical for the KAT activity (34,35). Our structure shows that the active site of MOF can well accommodate propionyl-CoA, and the F o Ϫ F c omit map shows clear density for the terminal methyl group in propionyl-CoA (Fig. 8B). Like acetyl-CoA, propionyl-CoA binds to MOF between the core domain and C-terminal domain in a bent conformation, interacting mostly with the ␤10-loop-␣4 region of the core domain and loop-␣5 region of the C-terminal domain (Fig. 8A). When the two complex structures are overlaid together, the propionyl-CoA molecule in the MOF⅐propionyl-CoA structure can superimpose with the corresponding moiety of the acetyl-CoA molecule in the MOF⅐acetyl-CoA complex structure very well (Fig. 8C). Most of the interactions between MOF and acetyl-CoA in the MOF⅐ acetyl-CoA structure are still conserved in the MOF⅐propionyl-CoA structure. However, propionyl-CoA makes two extra hydrophobic interactions through its propionyl group with Val-314 and Pro-349 in MOF (Fig. 8C). Sequence alignment reveals that both Pro-349 and Val-314 in MOF are conserved amino acid residues through all of the five eukaryotic MYST KATs (Fig. 9). These two interactions could contribute to the proper orientation of the propionyl group. The crystallographic data provide direct structural evidence that the MYST KATs behave as a dual activity enzyme for catalyzing both propionylation and acetylation.
To test the necessity of the two residues Val-314 and Pro-349 of MOF for its KPT activity, we conducted site-directed mutagenesis to replace MOF Val-314 or Pro-349 with Ala and then tested the KAT and KPT activity of WT MOF, MOF-V314A, and MOF-P349A using the fluorescent 7-diethylamino-3-(4Ј-maleimidylphenyl)-4-methylcoumarin (CPM) assay to obtain the kinetic constants k cat and K m with respect to acetyl-CoA and propionyl-CoA (Table S5). Both MOF-V314A and -P349A showed decreased intrinsic enzymatic activity on acetylation, retaining 30 and 20% activity compared with the WT MOF. The KPT activity of MOF-V314A abolished 55% of WT MOF, whereas mutation of Pro-349 to Ala abolished Ͼ95% of the KPT activity of MOF. These data suggest that the interaction between the propionyl group and MOF involves Pro-349 more than Val-314; therefore, the conserved Pro residue contributes more to the KPT activity of MYST KATs.

Discussion
The discovery of chemically diverse acylations in proteins represents an exciting area of research in biology. It embodies the potential and important regulatory roles of cellular acyl-CoA metabolites in the modulation of epigenetics and signal transduction. It would be vitally important to determine the cofactor promiscuity of different KATs to elucidate the biochemical etiology of cellular protein acylations. In this study, we found that the MYST family of KATs showed strong bona fide KPT activity. Especially, MOF, MOZ, and HBO1 exhibited as strong KPT activity as their classic KAT activity on histone substrates. The analysis of cellular proteins with Western blotting showed that all members of the MYST KATs promote extensive lysine propionylation, not only on histones but also on non-histone proteins. These findings that the MYST enzymes exhibited a widespread propionylome suggest their much broader physiological roles in the regulation of biological processes. We found that MOF has highly identical acetylome and propionylome profiles, but the modification levels at the individual lysine residues were slightly distinct. That the propionylome of MOF is highly identical to its acetylome indicates the cofactor promiscuity of the MYST enzymes and predicts

MOF and lysine propionylation
that the dynamics of distinct acylation marks in proteins is greatly influenced by the metabolic fluctuation of cellular acyl-CoA variants. We, for the first time, solved the crystal structure of a MYST enzyme bound with propionyl-CoA, which provides direct evidence supporting the KPT activity of the MYST KATs. Moreover, our data suggest that the residue Pro-349 in MOF, conserved through the MYST family of KATs, is required for its KPT activity. Future efforts will be needed to clarify how the sub-propionylomes of individual KATs differ from each other and to address how lysine propionylation impacts the properties of the substrate proteins in various biological pathways.

Quantification of cellular acetyl-and propionyl-CoA
HEK293T cells were cultured to 90% confluence with Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% streptomycin-penicillin. Cells were washed with ice-cold PBS buffer followed by fixing in methanol at Ϫ80°C for 15 min. 75 ng of deuterated acetyl-CoA and propionyl-CoA were added as internal standards. Cells were collected in 50% methanol with gentle scrape. Cell suspension was centrifuged at 16,000 ϫ g, and supernatant was collected for LC-MS/MS analysis.

MOF and lysine propionylation
ter with an electrospray ionization source (Milford, MA) was applied for LC-MS/MS analysis. The column temperature was controlled at 32°C. The mobile phase A was 10 mM ammonium acetate, and mobile phase B was acetonitrile. Analytes were separated using a gradient method, with a 0.4-ml/min flow rate as follows (time, percentage mobile phase B): 0 min, 6%; 15 min, 30%; 15.01 min, 100%; 22.50 min, 100%; 22.51 min, 6%. The injection volume was 30 l, and the autosampler injection needle was rinsed with methanol after each injection. Samples were analyzed by the mass spectrometer in positive-ion electrospray ionization mode. Nitrogen was used as the desolvation gas at a flow rate of 500 liters/h. The desolvation temperature was 500°C, and the source temperature was 120°C. Argon was used as the collision gas, and the collision cell pressure was 3.5 ϫ 10 Ϫ3 millibars. The capillary voltage was 3.2 kV, the cone voltage was 42 V, and the collision energy was 22 eV. Multiplereaction monitoring functions were applied for the detection of each acyl-CoA and internal standards. The ion transitions monitored were 810 3 303 for acetyl-CoA, 813 3 306 for d 3 -acetyl-CoA, 824 3 317 for propionyl-CoA, and 829 3 322 for d 5 -propionyl-CoA.

Kinetic characterization of acetyl-CoA and propionyl-CoA in KAT-mediated histone modification
Synthetic histone peptides H3-20 and H4-20 (20 amino acids from the N terminus of histone H3 and H4; H3-20, Ac-ART-KQTARKSTGGKAPRKQL, H4-20, Ac-SGRGKGGKGLGKG-GAKRHRK) were used as acyl acceptor substrates. Acyl-CoA at varied concentrations was incubated with individual KAT enzymes and peptide at a fixed concentration. KAT enzymes deposit acyl groups on the ⑀-amine of peptide lysines and release CoASH simultaneously. The fluorogenic probe CPM was added to react with the by-product CoASH to form the fluorescent CoAS⅐CPM complex (23). The fluorescence intensities were measured with a microplate reader (FlexStation 3). The catalytic rate was determined from the fluorescence intensity. Kinetic constants, including binding affinity (K m ) and catalytic efficiency (k cat ), were determined by fitting the acyl-CoA concentration-catalytic rate to the Michaelis-Menten equation. All of the samples for the kinetic assays were duplicated, and the final results are presented as means Ϯ S.D.

Detection of acetylation and propionylation level on proteins
For in vitro studies, recombinant or extracted cellular proteome was incubated with acyl-CoA and KAT enzymes, followed by Western blot detection with anti-acetyllysine (PTM Biolabs, product no. PTM-101) and anti-propionyllysine antibodies (PTM Biolabs, product no. PTM-210). The radiometric gel analysis was used to compare the acetylome and propionylome of MOF. For the in cellulo studies, cellular proteome or core histone proteins were extracted from the control and MOF-overexpressed cells and detected with different antibodies.

Profiling of MYST KATs in vitro propionylome
The whole lysate of HEK293T cells was extracted using a gentle cell lysis buffer, M-PER (mammalian protein extraction reagent) (Thermo Scientific, product no. 78501) together with sonication at 30% amplitude using a sonicator (Fisher, model 120 Sonic Dismembrator). The protein propionylation level was tested with Western blotting using pan-anti-propionyllysine antibody (PTM Biolabs, product no. PTM-201).

Identification of MOF acylation sites on H4-20 peptide using LC-MS/MS analysis
100 M H4-20 peptide, 200 M acyl-CoA were incubated with 1 M MOF enzyme for 1 h at 30°C followed by desalting with C18 Ziptips. The desalted samples were analyzed by an ACQUITY UPLC system (Waters, Milford, MA) coupled to a Waters SYNAPT G2 mass spectrometer (Milford, MA). Peptides were separated on the HALO C18 peptide column (2.7 m, 4.6 ϫ 100 mm; Advanced Materials Technology, Wilmington, DE). Mobile phase A was water containing 0.01% formic acid, and B was acetonitrile. The injection volume was 10 l. Peptides were separated with a 0.3 ml/min isocratic flow of 5% B. MS tune parameters were as follows: capillary, 2.00 kV; sample cone, 35 V; extraction cone, 4.0 V; source temperature, 120°C; desolvation temperature, 500°C; desolvation gas, 500 liters/h. Data were first collected in the full scan mode in the mass range of 300 -1900. To identify acetylation/propionylation sites, data were acquired by the data-dependent acquisition mode. For data-dependent acquisition parameters, a 1-s MS survey scan in the m/z range of 300 -1900 was followed by MS/MS scans of up to three ions, when intensity rose above 1500 counts/s. MS/MS was acquired in the m/z range of 100 -1900, with a 2-s scan rate. MS/MS scan was switched to MS survey scan after three scans. Trap collision energy was set using charge state recognition, applying the default files for 1-4 charge states. Files containing MS/MS spectra were processed with Proteinlynx Global Server version 2.4 software (Waters, Milford, MA) to identify PTM sites.

Comparative analysis of MOF substrate specificity on both acetylation and propionylation
20 g of cell lysate was incubated with 2 M MOF and 50 M carbon-14 -labeled acetyl-CoA or propionyl-CoA for 3 h at 30°C. The reaction was quenched with the addition of 5ϫ SDS-PAGE loading dye, and the proteins were resolved on a 4 -20% gradient SDS-polyacrylamide gel (Bio-Rad). The gel was dried in vacuum and exposed to a phosphor screen for 72 h. The autoradiograph was scanned with the GE Storm 865 imager (GE Healthcare), and the in-gel proteins were imaged with Coomassie Brilliant Blue stain for protein loading control.

Study of in vivo MOF acyltransferase activity
Full-length MOF-encoding sequence was inserted into XbaI and EcoRI sites of lentivirus vector FuCRW to generate the MOF-overexpression plasmid (36). Plasmid transfections were performed using Lipofectamine 3000 (Thermo Fisher Scientific). 24 h before transfection, the HEK293T cells were seeded at a density of 1 ϫ 10 6 cells/well for a 6-well plate. The cells were transfected with 8 g of MOF-vector or vector plasmid as described in the manufacturer's protocol, and the transfected cells were maintained for 72 h before harvest. 20 mM sodium propionate was added into the medium 24 h before the cellular proteins were extracted. Whole-cell lysate and histones were MOF and lysine propionylation extracted from both MOF-overexpressed and normal cells for the study of cellular protein acetylation and propionylation changing in response to the overexpression of MOF. The extracted proteins were subjected to Western blot analyses.
For the MOF knockdown study, control siRNA and siRNAs targeting KAT8 (ON TARGET plus TM siRNA) were purchased from Dharmacon (Dharmacon, Lafayette, CO). siRNA transfections were performed using Lipofectamine 3000 (Invitrogen). HEK293T cells with ϳ80% confluence in a 6-cm dish were transfected with 300 pmol of siRNA (15 l of 20 M stock) and 30 l of Lipofectamine 3000 and maintained for 72 h. Then the cellular proteins were extracted for Western blot analyses.

Proteomic profiling of MOF sub-propionylome
The whole-cell lysates and core histone proteins were extracted from the HEK293T cells treated with 20 mM deuterated sodium propionate with or without MOF overexpression. The extracted whole-cell lysates or histones were precipitated using trichloroacetic acid. The resulting protein precipitate was washed twice with ice-cold acetone and digested with trypsin using a procedure described previously (37). To enrich the propionylated peptides, the tryptic digest in NETN buffer (100 mM NaCl, 1 mM EDTA, 50 mM Tris-HCl, 0.5% Nonidet P-40, pH 8.0) was incubated with pan-anti-Kpr antibody (PTM-201, PTM Bio, Chicago, IL) that was immobilized to protein A-agarose beads at 4°C for 6 h with gentle rotation. After incubation, the beads were washed three times with NETN buffer and twice with double-distilled H 2 O. The enriched peptides were eluted with 0.1% TFA. The eluates were vacuum-concentrated, and the peptides were suspended in 0.1% TFA followed by desalting using C18 Ziptips. The peptide samples were directly loaded onto a homemade capillary column (10-cm length, 75-m internal diameter) packed with Reprosil C18 resin (3-m particle size, 100-Å pore size, Reprosil) on an EASY-nLC TM 1000 system (Thermo Fisher Scientific). The binding peptides were eluted with a gradient of 5-90% HPLC buffer B (0.1% formic acid in 90% acetonitrile (v/v)) in buffer A (0.1% formic acid in water (v/v)) over 60 min at a flow rate of 200 nl/min. The eluted peptides were ionized and introduced into a Q-Exactive mass spectrometer (Thermo Fisher Scientific) using a nanospray source. Full MS scans were acquired over the range of m/z 300 -1400 with a resolution of 70,000, which was followed by data-dependent MS/MS fragmentation of the 15 most intense peaks with a resolution of 17,500 at 27% normalized collision energy. For all of the experiments, the dynamic exclusion time was set to 25 s. Peptide identification was performed with MaxQuant (version 1.3.0.5) against the UniProt human protein database. Oxidation on methionine, protein N-terminal acetylation, lysine acetylation, lysine propionylation, lysine d 5 -propionylation, lysine mono-/di-/trimethylation, and arginine mono-/dimethylation were set as variable modifications. Carbamiodomethylation on cysteine was set as a fixed modification. False discovery rate thresholds for protein, peptide, and modification site were specified at 0.01. Peptides with MaxQuant score below 40 or site localization probability below 0.75 were removed. In addition, all of the identified peptides were manually verified.

Crystallization of MOF with propionyl-CoA and X-ray crystal structure determination
Purified recombinant MOF protein (5 mg/ml) was mixed with propionyl-CoA (Sigma) at 1:5 molar ratio of protein/compound and crystallized using the hanging-drop vapor diffusion method at 20°C by mixing 1 l of the protein solution with 1 l of the reservoir solution. MOF⅐propionyl-CoA was crystallized in buffer containing 20% PEG 3350, 0.2 M sodium malonate, pH 5.0. Crystals were soaked in the corresponding mother liquor supplemented with 20% glycerol as cryoprotectant before freezing in liquid nitrogen.
X-ray diffraction data for MOF ϩ propionyl-CoA was collected at 100 K on a Rigaku FR-E superbright X-ray generator. The data set was processed using the HKL-3000 suite (38).
The structure of the MOF⅐propionyl-CoA complex is isomorphous with PDB entry 2GIV. REFMAC (39) was used for structure refinement. The graphics program COOT (40) was used for model building and visualization. MOLPROBITY (41) was used for structure validation.