Structural, Kinetic and Proteomic Characterization of Acetyl Phosphate-Dependent Bacterial Protein Acetylation

1Post-translational modifications generate multiple protein isoforms, many of which have functional consequences. For example, Nε-lysine acetylation (lysine acetylation) can change protein size, conformation and/or charge, thus altering DNA binding affinity, enzymatic activity, protein stability, protein-protein interactions or protein localization,. It is well known that eukaryotes use lysine acetylation as a primary regulatory mechanism. Less well known is that lysine acetylation also occurs in bacteria. Proteomic studies of diverse bacteria have identified hundreds of acetylated proteins that function in various cellular processes,,,,,,,,. Such studies have led to the supposition that protein acetylation in bacteria might play an important physiological role.

2Lysine acetylation is typically thought of as the donation of an acetyl group from acetyl-coenzyme A (acCoA) to the ε-amino group of a lysine residue within a protein. It is a post-translational modification regulated by enzymes of opposing function, lysine acetyltransferases (KATs) and deacetylases (KDACs). Five protein families with KAT activity have been identified across all three domains of life,. Of these families, the GCN5-like acetyltransferases (GNATs) are distributed most widely. GNATs are ubiquitous within bacteria and can perform diverse functions. Some bacterial GNATs (e.g., Pat from Salmonella enterica) function as KATs, catalyzing Nε-acetylation of proteins, while others catalyze other reactions, including the acetylation of lipids or antibiotics,. Escherchia coli possesses more than 25 known or putative GNATs. Only one, YfiQ (a Pat paralog), has been reported to function as a KAT,. Two types of KDACs are known: the family of NAD+-dependent sirtuins and a family of metal-dependent enzymes. Putative bacterial homologs of each KDAC family exist; however, the sirtuin CobB is the only known KDAC in E. coli,,,.

3We recently reported that the acetylation patterns of RNA polymerase (RNAP) and of the response regulator and transcription factor RcsB differ substantially if purified from an acetate kinase (ackA) mutant rather than from its wild-type (WT) parent,,. Since ackA mutants can accumulate acetyl phosphate (acP), the high-energy intermediate of the phosphotransacetylase-acetate kinase (Pta-AckA) pathway ( Fig. 1A ),, we hypothesized that acP could regulate the acetylation of RNAP, RcsB, and other proteins.

4To test our hypothesis, we genetically altered the ability of cells to synthesize acP, to acetylate or deacetylate lysines, or to do both, and monitored global acetylation status by Western immunoblot analysis and quantitative label-free mass spectrometry. We used a high-throughput peptide array technology to determine if acP-dependent acetylation of susceptible lysines could occur in a non-enzymatic manner and x-ray crystallography to both confirm acP-dependent lysine acetylation of specific proteins and to examine the molecular environment of susceptible lysines. We mapped the locations of lysine residues modified in the different mutants onto previously determined protein structures in the Protein Data Bank (PDB) and performed kinetic experiments of in vitro acP-dependent acetylation.

5We conclude that lysine acetylation is common in E. coli, modifying thousands of lysines on hundreds of proteins involved in diverse and often essential cellular processes. We further propose the existence of at least two distinct mechanisms for lysine acetylation: a novel acetylation mechanism that uses acP as the acetyl donor, as reported recently by Weinert and co-workers, as well as the conventional mechanism that requires the acCoA-dependent KAT (YfiQ). We provide validated quantitative evidence that acP-dependent acetylation is common, especially modifying proteins that function in translation, central metabolism and transcription. We provide crystallographic evidence that acP can bind to substrate and co-factor binding sites and that acP-dependent acetylation of key lysine residues of the central metabolic enzymes triosephosphate isomerase (TpiA) and glyceraldehyde-3-phosphate dehydrogenase (GapA) could alter their function. Finally, we propose that acP-dependent acetylation can be both non-enzymatic and specific, with specificity determined by the local three-dimensional (3D) environment that surrounds the substrate lysine residue, as well as by its solvent accessibility and pKa.

6To test the hypothesis that acP plays a role in lysine acetylation, we grew E. coli at 37°C in tryptone broth buffered at pH7 (TB7), harvested them at several intervals, and subjected the lysates to Western immunoblot analysis using an anti-acetyllysine antibody. Relative to its WT parent, we observed more acetylation in the ackA mutant, especially during early exponential growth ( Fig. 1B, Fig. S1A) when acP concentrations are at their highest,,. In contrast, lysine acetylation in cells that cannot synthesize acP (pta ackA) more closely resembled the WT pattern ( Fig. 1C, Fig. S1A). To determine if the increased acetylation of the ackA mutant was due to acP, we grew a pta mutant in TB7 supplemented with acetate. Because this mutant retains AckA activity, it synthesizes acP from acetate independently of Pta. Since acetylation increased in an acetate-dependent manner ( Fig. 1D ), we conclude that lysine acetylation is mediated by acP itself, and not by the AckA and Pta enzymes that regulate acP levels.

7We next asked if this acP-dependent acetylation required the only known KAT in E. coli, comparing the acetylation profiles of an ackA yfiQ double mutant and the ackA single mutant (Fig. S1B). Due to the lack of substantial differences in the acP-dependent acetylation profile, we conclude that YfiQ is not involved and hypothesize that acP-dependent acetylation requires a novel mechanism.

8Growth on glucose has been reported to increase acP concentrations and to increase lysine acetylation,,. To test if the previously reported glucose-induced acetylation requires acP, we grew WT, ackA and pta ackA cells in TB7 supplemented with 0.4% glucose and determined the protein acetylation profile by Western immunoblot analysis. Consistent with the idea that acP is a key contributor to glucose-induced acetylation, most of the signal detected from WT cells increased in cells (ackA) that accumulated acP and decreased in cells (pta ackA) that could not synthesize acP ( Fig. 1E ). Since acP-dependent acetylation was observed along the entire range of protein molecular weights, we hypothesize that acP-dependent acetylation occurs on many proteins and that acP may have a physiologically relevant role in lysine acetylation of E. coli proteins. This role might be larger than that of the known KDAC (CobB) and KAT (YfiQ), as the acetylation patterns in mutants lacking these enzymes closely resembled the patterns in their WT parent ( Figs. 1F and G ). Since acetylation occurs in E. coli cells that lack acP (pta ackA) (Fig. S1C), we conclude that there exist at least two mechanisms for lysine acetylation: one that depends on acP and one that does not.

9To identify specific acetyllysine sites on individual acetylated proteins and to quantify sites that are clearly regulated by the acP-dependent and acP-independent mechanisms, we analyzed samples from these same strains using an in-depth mass spectrometry approach. We used a robust peptide-based affinity enrichment strategy using anti-acetyllysine antibodies (1) to identify acetyllysine-containing peptides from trypsinized total protein lystates and (2) to compare the relative levels of acetyllysine-containing peptides between the WT parent and its mutants, using a recently developed label-free quantitative proteomics method referred to as ‘Skyline MS1 Filtering',. The label-free approach provided us with flexibility in experimental design; in particular, it allowed us to directly compare acetylation in cells grown under more physiologically relevant conditions in complex growth media, in contrast to the SILAC method used by Weinert et al..

10Briefly, cells from different E. coli strains were grown in TB7 or TB7 plus 0.4% glucose and harvested just prior to entry into stationary phase (for WT, equivalent to Fig. 1C lane 2). From either 3 or 4 biological replicates of each strain/condition, total proteins were isolated, proteolytically digested, and immunoprecipitated using a polyclonal anti-acetyllysine antibody for affinity enrichment. Subsequently, each biological replicate and all acetyllysine-enriched samples were subjected to data-dependent mass spectrometric analysis on a TripleTOF 5600 in 3 technical injection replicates ( Fig. 2A ). These mass spectrometric acquisitions were later used for label-free quantification (MS1 Filtering), which allowed pairwise comparisons of peak areas and relative abundances of acetyl sites either between strains (i.e., ackA/WT) or between different growth conditions ( Fig. 2B ). Lists of candidate acetyllysine sites upregulated in the ackA or cobB mutants relative to the WT were then confirmed with an independent, quantitative verification methodology referred to as SWATH-MS2, as described in detail below ( Fig. 2C ).

11For five strains grown under two different conditions and harvested at a single time point, mass spectrometry confidently identified a total of 2730 unique acetyllysine sites on 806 acetylated proteins, as analyzed in 4 biological replicates and 3 technical replicates by LC-MS/MS (for unique acetyllysine sites, see Tables S1-5; for all identified acetyllysine peptides, see Table S6-7). The acetyllysine sites were not distributed equally among proteins. While many of the acetylated proteins were acetylated on one lysine, most were acetylated on at least two lysines and some contained as many as 24 acetyllysine sites ( Fig. 3A, Table S8). This distribution (i.e, the number of acetyllysine sites per protein) did not correlate strongly with molecular weight, but did show some correlation with the total number of lysines per protein (Fig. S2A–C, Table S8). This distribution also showed some correlation with the estimated protein abundance ( Fig. 3B, Table S8), as expected of any proteomic enrichment strategy. Although it is typical for acetyllysine sites of abundant proteins to be identified frequently, we also observed acetyllysine sites of proteins with very low estimated protein copy numbers per cell ( Fig. 3B ).

12Data-dependent mass spectrometry acquisitions verified the existence of both acP-dependent and acP-independent acetylation. In the absence of glucose, we detected 780 acetylated lysine sites on 355 proteins from WT, 1149 acetylated lysine sites on 448 proteins from the ackA mutant and only 320 acetyl sites from 166 proteins from the pta ackA mutant ( Table 1, Fig. 4A ). Since ackA mutants accumulate acP and pta ackA mutants cannot synthesize acP, lysine acetylation appeared to correlate with concentration of acP. Since some acetylation was observed in the pta ackA mutant, we verified that some acetylation must be acP-independent. We also verified the existence of glucose-induced acetylation. For cells grown in TB7 supplemented with glucose, we detected 1204 unique acetyllysine sites on 446 proteins from WT and 2473 acetyllysine sites on 751 proteins from ackA ( Table 2, Fig. 4B ), numbers that are considerably larger than those observed for cells grown in the absence of glucose. In comparison, relatively small effects of both YfiQ and CobB were observed. In the absence or presence of glucose, deletion of cobB resulted in a small change in acetyllysine sites relative to deletion of ackA ( Table 1 and 2, Fig. 4B and C ). In the presence of glucose, relative to the WT, the number of acetyllysine sites decreased slightly in the yfiQ mutant and increased slightly in the cobB mutants ( Table 2, Fig. 4D ). Taken together, these results confirm that YfiQ and CobB affect lysine acetylation in E. coli, but also support our proposition that acP is a major influence on lysine acetylation.

13To quantitatively compare the impacts of acP and CobB on lysine acetylation, we used MS1 Filtering, to compare the relative changes in acetyllysine abundances in both the ackA and cobB mutants relative to the WT, when grown in the presence of glucose. For each of the four independent biological replicates, relative peak areas were extracted for each unique acetyllysine peptide in three technical replicates (Fig. S3A). If the mutant/WT ratio was >2 with a significance of p<0.05 in at least 3 of the acquired 4 biological replicates, then an acetyl site was considered ‘regulated’, and a mean ratio across all biological replicates was determined, as demonstrated for select examples (Fig. S3B). For cells grown in the presence of glucose, 592 lysines on 292 proteins in the ackA mutant relative to its WT parent were identified as significantly increased ( Fig. 5A, Table S9 and S10). This represents 25% of the 2367 quantifiable sites for the ackA/WT experiment. In contrast, only 69 lysines in 51 proteins met these criteria in the cobB mutant ( Fig. 5B, Table S11 and S12). This represents 5% of the 1392 quantifiable sites for the cobB/WT experiment. Thus, the number of upregulated acetyllysine sites is substantially larger in the ackA mutant than in the cobB mutant (>8-fold for acetyl sites, >5-fold for proteins). Parallel analysis of total protein lysates (i.e., without affinity enrichment), and relative quantification of unmodified peptides revealed little evidence of any differential expression changes among proteins identified as acetylated. Specifically, acetylated proteins that contained regulated acetyllysine sites showed no changes in expression levels between the either the ackA or cobB mutants and WT ( Fig. 5C and 5D, Table S13 and S14), allowing us to conclude that acetylation of many lysines is dramatically upregulated in both mutants, but that the effect is much greater in the ackA mutant.

14The fold-change distribution of all quantifiable acetyllysine sites appears to be more specific for cobB (Fig. S4A, Table S15) than for ackA (Fig. S4B, Table S16). For the ackA mutant, the fold-change profile for upregulated acetyllysines was broader and more global than that of the cobB mutant, which exhibited profile changes for 10 times fewer distinct sites. However, both mutants displayed acetyllysine site specificity, as demonstrated for a subset of acetylation proteins (Fig. S4C and S4D). Furthermore, the acetylation sites upregulated in the ackA mutant can be distinctly different from sites de-acetylated by CobB (Fig. S4E).

15To independently validate the observed increase in acetylation of the acetyl sites regulated in the ackA mutant, we employed MS/MSALL with SWATH acquisition. SWATH is a comprehensive data-independent approach that is most similar to multiple reaction monitoring (MRM) assays in that it integrates sets of specific MS2 fragment ions for each targeted peptide ( Fig. 2A and 2C; for details, see methods). However, unlike MRM, a SWATH-MS2 acquisition samples all peptide ions within a large but defined m/z range (400–1000 m/z). Of the 592 acetyllysine sites that MS1 Filtering identified as regulated by ackA, SWATH-MS2 was able to quantify 526, with 507 of these sites (86% of all initial 592 candidates) meeting the same threshold for upregulation (≥2-fold increase, p-value ≤0.o5) (Fig. S5A, Table S17, including all SWATH assay details, Table S18). For example, Figure S5B shows a subset of these proteins whose candidate acetylation sites were verified by both MS1 Filtering and SWATH-MS2. A similar correlation between SWATH-MS2 and MS1 Filtering was observed for potential cobB acetyl target sites: out of 69 candidate sites quantified by MS1 Filtering, SWATH monitored 62 and verified 59 as upregulated in the cobB mutant (≥2-fold increase, p-value ≤0.o5) (Table S19 and Table S20).

16Now that our quantitative approaches had been independently validated, we used MS1 Filtering to re-examine the hypothesis that acP correlates with lysine acetylation, comparing the acetylation status of acP-sensitive acetyllysine sites in the ackA mutant grown in the absence of glucose to that of the same sites in the pta ackA mutant grown under the same conditions. The mass spectrometric results confirm the Western immunoblot analysis (compare Fig. S6 and Table S21 to Figs. 1B-C and S 1A), showing that acetyllysine-containing peptides decreased significantly (on average 6.9-fold) in the pta ackA mutant relative to the ackA mutant.

17To examine the potential biological impact of acP-dependent acetylation, we used PANTHER to perform functional analyses on those lysines whose acetylation increased substantially and significantly in the ackA mutant relative to the WT strain. The most common molecular function of these lysines was ‘catalytic activity,’ but the functions ‘binding’ and ‘structural molecule’ activity also were common (Fig. S7A, Table S22). By far, the most common biological processes (Fig. S7B, Table S23) were metabolic in nature, i.e., ‘metabolic process’ and ‘generation of precursor metabolites and energy’, but other processes were relatively common, e.g., ‘immune system process,’ ‘response to stimulus’ and ‘transport’. Enrichment analysis by DAVID using KEGG pathway ( Table 3 ) terms supports the hypothesis that acP-sensitive acetylation targets central metabolism: ‘glycolysis/gluconeogenesis,’ ‘pentose phosphate pathway,’ ‘pyruvate metabolism,’ the ‘TCA cycle,’ ‘fatty acid biosynthesis’ and ‘pantothenate metabolism.’ For example, almost every protein involved in glycolysis or the TCA cycle possessed at least one acetyl site that was upregulated in the ackA mutant (Fig. S8). Many of these proteins were amongst the most acetylated, including formate acetyltransferase (PflB), phosphoglycerate kinase (Pgk), transketolase 2 (TktB), enolase (Eno), and pyruvate dehydrogenase complex E1 component (AceE), and dihydrolipoyl dehydrogenase (LpdA), a subunit of the pyruvate dehydrogenase, α-ketoglutarate dehydrogenase and glycine cleavage complexes (Table S8). AcP-sensitive acetylation also tends to target translation (‘ribosome’ and ‘aminoacyl-tRNA biosynthesis’), nucleotide metabolism (‘purine and pyrimidine metabolism’), and ‘RNA degradation.’ These conclusions are supported by Gene Ontology analysis (Table S24) and extended to include DNA-dependent transcription (GO:0006351), including all the major components of the RNAP complex (α, β, β′, and σ70), as well as transcription termination (GO:0006353), including NusB, NusG and Rho.

18While the immunological and mass spectrometric evidence clearly supports a role for acP in lysine acetylation in vivo, the mechanism by which acP performs this role remained unknown. We knew that acP can donate its phosphoryl group to certain response regulators,,; therefore, acP could regulate protein acetylation indirectly, via a response regulator that regulates the expression or activity of a protein acetyltransferase or deacetylase. However, we currently possess no evidence for this mechanism. The acetylation profile of the ackA yfiQ double mutant was similar to that of an ackA mutant (Fig. S1B); thus, the only known KAT in E. coli cannot be intimately involved. Furthermore, deletion of yafP, the only putative KAT gene whose expression depends on acP, did not observably alter the global acetylome (data not shown).

19In contrast, substantial evidence exists to support the hypothesis that acP could function directly as an acetyl donor. AcP is a high-energy compound and its chemical properties favor such a reaction. In water, it undergoes rapid hydrolysis, which increases under basic or acidic conditions and elevated temperatures,. In the presence of Mg2+ or Ca2+ ions, acP hydrolysis increases,, because the metal ions coordinate with the phosphate to make it a better leaving group. The metal ions also neutralize phosphate's negative charge by decreasing the electrostatic repulsion. AcP also can react with imidazole to form acetylimidazole, which can hydrolyze or undergo an acyl transfer reaction to other functional groups,. By nucleophilic attack of its carbonyl carbon, acP can chemically acetylate amines, thiols, or hydroxyl groups. In the presence of such nucleophilic reagents, the acyl transfer reaction is preferred over hydrolysis.

20To determine if acP can react with peptides containing a lysine residue and, if so, to identify the acP chemical acetylation profile, we used a novel label-free method called SAMDI (Self-Assembled monolayers with Matrix-assisted laser Desorption-Ionization) mass spectrometry. This method monitors peptide modification of a peptide library immobilized onto a gold surface,,,, and revealed that acP can acetylate peptides non-enzymatically. At high concentrations of acP, the reaction was promiscuous and acetylated nearly every peptide. Upon the addition of salt and Mg2+, however, selectivity and specificity increased (data not shown). Under these conditions, the reaction was particularly favored when the acetylated lysine was adjacent to an additional positively charged residue (especially arginine (R) or lysine (K)) ( Fig. 6, Table S25).

21To see if the in vitro SAMDI peptide acetylation profile was recapitulated in vivo, we examined the protein primary sequences of the acetyl sites identified in our global mass spectrometric analysis. For this purpose, we only used the 592 acetyl sites identified as reproducibly, robustly, and significantly upregulated in the ackA mutant relative to its WT parent (≥2-fold increase in acetylation in at least 3 of 4 bioreplicates with a p-value<0.05). Of these 592 sites, we analyzed 541, compiling a table of neighboring amino acids from position -10 to position +10 relative to each acetyllysine. We then generated a sequence logo ( Fig. 7A, Table S26), using WebLogo. In contrast to the acetylation profile obtained from the in vitro SAMDI analysis, this analysis of acetyl sites identified by mass spectrometry of bacterial lysates revealed a high prevalence of negatively charged glutamate (E) and aspartate (D) residues in the -1 (∼30%) and +1 (∼18%) positions, adjacent to the acetyllysine residue. To determine if this prevalence was relevant to acetylation, we selected 4 proteins with multiple ackA-sensitive sites (LpdA, 13; PflB, 11; AceE, 10; and GroL, 11) and multiple ackA-insensitive sites (LpdA, 25; PflB, 39; AceE, 38; and GroL, 29) and generated sequence logos for both types of lysine ( Figs. 7B and 7C, Table S27). There was a strong tendency for negatively charged residues to be located adjacent to ackA-sensitive lysines, with a lesser tendency relative to insensitive lysines. In contrast, positively charged residues tended to be rarely adjacent to sensitive sites, but more commonly adjacent to insensitive sites. To more rigorously test these findings, we normalized the frequency of adjacent residues to the overall frequency of those residues in these 4 proteins ( Fig. 7D and 7E ). Negatively charged residues were more likely to be in the -1 position adjacent to ackA-sensitive sites than to insensitive sites (p<0.009), whereas positively charged residues were more likely to be in the -1 position adjacent to insensitive sites (p<0.03) (Table S28). On the basis of these results, we propose that adjacent negatively charged residues play a role in acP-dependent acetylation.

22Since the in vitro and in vivo linear (1D) sequences did not match, we examined the 3D acP-binding environment in native, folded proteins, asking whether positively charged lysine or arginine residues were located in the vicinity of acP binding sites and/or acP-sensitive lysine residues. To begin, we analyzed the structure of phosphotransacetylase (Pta) from Bacillus subtilis (PDB ID: 1XCO and ) in complex with acP. Pta, the enzyme that catalyzes the reversible reaction that interconverts acCoA and orthophosphate to acP and CoA, crystallized as a dimer (data not shown) with four molecules of acP bound in different locations within the structure ( Fig. 8 ). These acP molecules were coordinated to: (1) the ω and ω' nitrogen of R89, the ε amino group of K92, and the amide group of the side chain of Q138 ( Fig. 8A ); (2) the ω and ω' nitrogen of R88, the ε amino group of K141, and the amide group of the side chain of N296 ( Fig. 8B ); (3) a water molecule, the hydroxyl groups of the side chains of T129 and S303, and the δ amino group of R304 ( Fig. 8C ); and (4) a water molecule, the amide group of the side chain of N273, the ε amino group of K277, the hydroxyl group of the side chain of Y276 in one monomer and the hydroxyl group of the side chain of S303 from an adjacent monomer ( Fig. 8D ). Interestingly, in a study of Pta from Methanosarcina thermophilia (PDB ID: 2AF3), two molecules of CoA were bound to the enzyme and the acP sites observed in the Bacillus subtilis Pta correspond to CoA pyrophosphate and ribose-3-phosphate locations. Additionally, the arginine corresponding to R304 in the third site listed above was shown to be critical for catalysis. Thus, like phosphate binding sites, acP can be coordinated by positively charged residues, but also hydrogen bond donating groups such as a hydroxyl (i.e., from S, T or Y) or the side chain amide group (e.g., from Q or N).

23To further characterize acP's mode of binding, we crystallized the E. coli enzymes TpiA ( Fig. 9 ) and GapA ( Fig. 10 ). Both enzymes participate in glycolysis and glucoenogenesis (Fig. S8): TpiA interconverts dihydroxyacetone phosphate (DHAP) and glyceraldehyde 3-phosphate (G3P), while GapA converts the G3P to 1,3-bisphosphoglycerate using NAD+ as a co-factor. We chose these two enzymes because our mass spectrometric results revealed that most enzymes of the glycolytic and gluconeogenic pathways possess lysine residues that are significantly more acetylated in the ackA mutant than in the WT parent ( Fig. 11, Fig. S8, Table S9).

24We determined the 3D structure of TpiA in its apo-form (PDB ID: 4K6A) and the structure obtained from similar crystals that had been quickly soaked (<30 sec) with acP (PDB ID: 4MVA). The structures showed that acP bound near K11, where the substrate DHAP is seen to bind (e.g., in PDB ID: 1NEY) ( Fig. 9A,B ). Under these conditions, K11 was not acetylated and the phosphate moiety of the acP molecule interacted with the Oγ of S212 and and the main-chain nitrogen of G234. All other interactions were water-mediated. Although quantitative mass spectrometry revealed increased in vivo acetylation of K103 and K111, as well as K11 ( Fig. 9C ), in the short time of the crystal soak, no acP was found in the crystal structure near either K103 or K111 and neither residue was modified (data not shown).

25To obtain the structure of acP-soaked GapA (PDB ID: 4MVJ), we used a similar approach. The enzyme crystallized with four tetramers in the asymmetric unit and had either NAD+ or pyrophosphate bound in the NAD+-binding site. Not all molecules in the asymmetric unit had the same residues modified or ligands bound (Table S29, for more information). Quantitative mass spectrometry had shown increased acetylation of seven lysine residues, three above the active site (including K184) and four sites exterior to the active site (including K249) ( Fig. 10A ). The soaked crystals revealed that, within the short time of the crystal soak, acP-dependent acetylation had already occurred on four residues (i.e., K46, K249, K257, and K116) and three of these had well-defined electron density surrounding their acetyl groups ( Fig. 10B-D ). For these three lysines, no other positively charged residues were present in the surrounding environment, though they were located near carboxylates. Because the crystallographic experiments reveal acP binding and lysine acetylations that have reached significant fractional occupancy in the crystal environment in less than one minute, whereas the mass spectrometry data defines the fold increase in acetylation state, it should not be surprising that the results do not correspond precisely.

26Taken together, these crystallographic results confirm the hypothesis that acP can acetylate lysine residues in the absence of an acetyltransferase. They also demonstrate that acP binding can be coordinated by functional groups with positive charges, hydroxyls, amides, or combinations of each, while acetylated lysines frequently also have carboxyl groups in the immediate neighborhood.

27While crystallography showed that acP could be an acetyl group donor to proteins in vitro, it could not reveal the kinetics of that reaction. To obtain kinetic information, we used the pyruvate dehydrogenase subunit LpdA because mass spectrometry analysis of the E. coli lysates showed that it was highly acetylated in the ackA mutant relative to the WT parent (Table S9). Indeed, when incubated with acP, purified LpdA became acetylated in a dose- and time-dependent manner ( Fig. 12 ). The most obvious increases in overall acetylation occurred in the presence of acP concentrations (10 and 20 mM) within the endogenous range detected in an ackA mutant,. Using these acP concentrations, we monitored relative acetylation of specific LpdA lysine residues over time, performing MS1 Filtering quantitation on two technical mass spectrometric replicates of two independent incubation replicates. This analysis confirmed the overall increase in acetylation observed by Western immunoblot analysis (compare Fig. S9A to Fig. 12 ) and revealed specific relative increases in acetylation of 6 and 11 lysine residues when LpdA was incubated with 10 and 20 mM acP, respectively (Fig. S9A and B; Tables S30 and S31). These acetylated sites were similar to those detected in vivo, suggesting that acP could regulate some LpdA acetylation in vivo.

28Because we had previously published on the acetylation of RNAP, and because we observed substantially increased in vivo acetylation of several RNAP lysine residues in the ackA mutant relative to the WT parent, purified RNAP was incubated with a range of acP concentrations for 30 minutes. By Western immunoblot analysis, we observed acetylation of the subunits β (RpoB) and β′ (RpoA) ( Fig. 13 ). Using MS1 Filtering (Fig. S10, Table S32), significant relative increases in acetylation were observed for K297 of RpoA, for K1065 of RpoB, and for K493 and K593 of RpoD (the σ70 subunit). On the basis of these data and those obtained with LpdA, we conclude that acP-dependent acetylation depends on both acP concentration and duration of incubation.

29For particularly reactive lysine residues of LpdA and RNAP, we assessed the correlation of their acetylation with their pKa and surface accessibility (Tables S33–36). Typically, the surface accessibility of the acetylated lysine was more important than its pKa. One interesting exception was K1065 of RpoB, whose acetylation status increased dramatically (14.8 fold) upon acP incubation. Although K1065 is not exposed to the surface, it has a somewhat lower calculated pKa value (9.2) than other acetylated RpoB lysines. This lower pKa, combined with the identity and location of neighboring residues, could explain the elevated reactivity of K1065.

30We used a label-free quantitative mass spectrometric workflow to provide comprehensive, reproducible, and validated evidence that lysine acetylation is abundant in E. coli. This label-free approach, MS1 Filtering followed by SWATH-MS2 validation, allowed us the flexibility to analyze cells in tryptone broth, a complex mixture of amino acids, short peptides, nucleotides, and vitamins that forms the basis for many of the most common media, including Luria broth. This approach in general will permit global mass spectrometric analyses of organisms that must be grown in complex media or feedstocks, such as fastidious pathogens (e.g., Mycobacterium tuberculosis, Haemophilus influenzae or Neisseria meningitidis), industrial strains (e.g., those being developed for biofuel production), or bacteria grown in tissue culture to test associations with eukaryotic cells (e.g., urinary microbiota in association with human urothelium).

31Using our label-free approach, we detected 2730 unique acetyllysine sites on 806 acetylated proteins from an isogenic set of strains harvested at a single time point ( Tables 1 and 2, Fig. 4, Tables S1–S5). The detected acetyllysine sites were not distributed equally. Most of the acetylated proteins contained one or two acetylated lysines, but some contained as many as 24 ( Fig. 3A, Table S8). Whereas the distribution of acetyllysines did not correlate with molecular weight, it trended with the total number of lysines per protein (Fig. S2) and with estimated protein abundance ( Fig. 3B ). The dynamic range of detection of acetylated sites covered about 7 orders of magnitude, identifying acetylated lysines from proteins with very low, as well as very high, estimated protein copy numbers per cell ( Fig. 3B, Table S8). Since the TripleTOF 5600 mass spectrometer when combined with HPLC separation typically shows a dynamic range of 4-5 orders of magnitude, we conclude that expansion of the dynamic range was achieved by our acetyl-enrichment strategy.

32Our quantitative mass spectrometric assessment of lysine acetylation was performed in a label free environment simulating physiological conditions that could not have been achieved easily using metabolic SILAC labeling strategies, which require specific growth media and strain manipulations. To assess both technical and biological variance, we performed 3 technical replicates on either 3 or 4 independent biological replicates. Strict statistical requirements were applied to identify acetylation sites that were upregulated in mutants relative to their WT parent, arriving at final candidate lists of 592 and 69 lysines that were significantly more acetylated in the ackA or cobB mutants, respectively. Our use of an MRM-like data-independent method (SWATH-MS2) confirmed the majority of the acetylated candidates (86% and 86% for the ackA- and cobB-sensitive sites, respectively). An added advantage of SWATH-MS2, a quantitative ‘targeted proteomics’ method, centers on its ability to monitor many sites with a quantitative data-independent acquisition manner. As such, it is a high-throughput assay that requires little prior assay development. To our knowledge, these data are the most statistically rigorous and validated quantitative analysis of global acetylation in bacteria.

33By buffering the complex medium, we eliminated pH effects caused by metabolism of amino acids (alkalinization) and/or glucose (acidification). As growth under these conditions did not depend on the glucose but rather on the underlying tryptone, this allowed us to rigorously test the proposal that exposure to glucose increases lysine acetylation,, in a physiologically relevant condition (i.e., natural transition into stationary phase) and to identify lysines that become acetylated in the additional presence of glucose in the medium. Using both immunological and mass spectrometric analyses on our isogenic set of strains, we verified that glucose does indeed increase global protein acetylation, about two-fold under the tested growth conditions ( Tables 1 and 2, Fig. 4 ).

34The mechanism by which glucose increases lysine acetylation remains unknown; however, it may involve the acCoA-to-CoA ratio. We recently reported that glucose induces transcription from the envelope stress response promoter cpxP, that this effect can be mimicked by growth of these cells in the presence of the CobB inhibitor nicotinamide, that glucose-induced cpxP transcription depends on K298 of the C-terminal domain of the α subunit (α-CTD) of RNAP, and that K298 is acetylated under these conditions. Intriguingly, glucose-induced cpxP transcription is dampened by three distinct genetic manipulations that reduce the acCoA-to-CoA ratio. Additional evidence suggests that certain other glucose-induced promoters behave similarly (L. I. Hu and A. J. Wolfe, unpublished data).

35One possible consequence of a high acCoA-to-CoA ratio is the accumulation of acP. Indeed, we showed that the glucose effect upon lysine acetylation was enhanced when acP levels were high (ackA), but nearly abolished when acP was absent (pta ackA) ( Fig. 1E ). We conclude that glucose-dependent enhancement of lysine acetylation requires acP. We further propose that simultaneous exposure to both mixed amino acids and glucose (e.g., in the urine of a diabetic patient) may be a physiologically relevant condition under which the previously reported acP-dependent acetylation might play a key role.

36Our immunological analyses of isogenic sets of strains provided definitive genetic evidence that much, but not all, of lysine acetylation in E. coli is mediated by acP ( Fig. 1 ), that this effect behavior is not strain-specific (Fig. S1A), and that acP-dependent acetylation does not depend on YfiQ (Fig. S1B). These results strengthen and extend the proposals of Weinert and co-workers.

37The glucose-induced acP-dependent acetylation identified by label-free quantitation ( Fig. 5, Table S9) was enriched in essential processes such as translation, central metabolism and transcription ( Table 3, Fig. S8, and Table S24). This enrichment does not seem to be associated with the glucose, but rather with acP, as a similar pattern was observed in the 84 acP-dependent acetyllysines that were more acetylated in ackA cells grown in the absence of glucose compared to the pta ackA cells grown under the same conditions (Fig. S6 and Table S21).

38Previous reports highlighted the tendency of bacterial proteins involved in translation and central metabolism to be acetylated,,,,,,,. To our knowledge, however, ours is the first attempt to perform pathway enrichment analysis, which reveals that these processes are particularly sensitive to acP-dependent acetylation. An obvious question is why? For central metabolism, the answer might be as simple as providing a mechanism by which cells respond to the overflow metabolism that occurs when the carbon flux exceeds the capacity of central metabolic pathways, e.g. the TCA cycle. One major consequence of overflow metabolism is the conversion of acCoA to acP, accumulation of which would cause acetylation of the glycolytic enzymes that convert carbon to acCoA and thus to acP. A similar rationale might be applied to the proteins of translation. WT cells that accumulate acP are growing rapidly. Perhaps acP-dependent acetylation of ribosomal proteins and other translation-associated proteins alters their function to accommodate more rapid growth.

39Since it was clear that acP influenced lysine acetylation, we sought the mechanism by which it performs this function. We knew that acP can donate its phosphoryl group to certain response regulators,,, but we found no evidence that acP regulates lysine acetylation by controlling the expression of yafP, the only putative KAT gene whose expression depends on acP. In contrast, we obtained substantial evidence to support the hypothesis that acP could function directly as an acetyl donor, as proposed,. In vitro, acP acetylated GapA ( Fig. 10 ), LpdA ( Fig. 12 ), and RNAP ( Fig. 13 ). Thus, acP can clearly function as an acetyl donor for lysine acetylation. Furthermore, the process appears to be non-enzymatic in nature as (i) acP acetylated peptides in vitro ( Fig. 6 ), (ii) acP-dependent acetylation of GapA occurred within the crystal lattice, and (iii) a kinetic analysis of acP-dependent acetylation of purified LpdA showed the reaction to be slow and non-saturating ( Fig. 12, Figs. S9 and S10, Tables S30 and S31 and data not shown), which are characteristics of non-enzymatic reactions. Thus, we conclude that acP-dependent acetylation does not require an acetyltransferase. We further conclude that acP-dependent acetylation must be catalyzed by residues that surround the lysine substrate (see below). Whether acP-dependent acetylation occurs non-enzymatically in vivo remains to be determined. However, the lysines of GapA, LpdA and RNAP that became acetylated in vitro closely paralleled those identified as upregulated in the ackA mutant (compare Fig. 10, Fig. S9, and Table S30–S32 to Table S9). Moreover, there is precedence for non-enzymatic modifications; for example, acCoA-dependent acetylation and succinyl-CoA-dependent succinylation in mitochondria. Indeed, some non-enzymatic modifications have been implicated in human disease processes (e.g., glycation in diabetes).

40The effect of acP-dependent acetylation on cellular processes remains to be fully explored; however, some evidence exists. First, the most common molecular function of acP-sensitive lysines was ‘catalytic activity’ (Fig. S7A, Table S22). Indeed, four of the ten central metabolic enzymes whose structures were analyzed ( Fig. 11 ) have lysines within their active sites with significantly increased acetylation in the ackA mutant relative to the WT parent (Table S9); acetylation of these residues would be expected to reduce enzymatic activity. For example, the soaking of TpiA crystals with acP revealed an acP molecule bound near K11 in both active sites, in a position similar to that of bound DHAP. Thus, the binding of acP or the modification of this residue would be expected to inhibit binding of the substrate and thus TpiA activity. If acP were to acetylate K11, a site identified in our mass spectrometry experiments (Table S9), it would likely inactivate the enzyme. This expectation is based on the observation that the cationic side chain of K11 in yeast TpiA is critical for catalysis because it stabilizes the transition state of the reaction. Removing this charge in the yeast enzyme by substituting it with a glycine residue effectively rendered the enzyme inactive, and caused a decrease in activity of four-orders of magnitude for isomerization of G3P and a nearly 50-fold decrease in affinity for G3P.

41Although acP was bound in the active site of TpiA, it was not oriented properly for efficient nucleophilic attack on the carbonyl carbon ( Fig. 9B ). Unlike other structures of TpiA with the substrate DHAP (PDB ID: 1NEY), the transition-state analog 2-phosphoglycolic acid (PDB IDs: 2BTM, 1AW1), or orthophosphate (PDB ID: 1AG1) bound in the active site, our acP-bound structure did not efficiently trigger lid-closure of Ω loop 6 (comprised of residues 168-178) ( Fig. 9B ). Our crystal structure showed alternative conformations for this loop, likely due to crystal packing and contacts between different molecules of the asymmetric unit, especially the main chain nitrogen of T174 of one monomer and the main chain carbonyl oxygen of K152 of another monomer. Such effects likely contributed to the improper orientation of acP, prohibiting efficient acetyl transfer to K11.

42Acetylation also could inhibit enzyme activity by interfering with the binding of co-factors, allosteric effectors or protein-protein interactions. For example, in the case of GapA, acetylation of K183 of the yeast, sturgeon muscle, and rabbit muscle enzymes (which is homologous to E. coli K184) decreases the enzyme's affinity for NAD+,, resulting in a loss of enzymatic activity, possibly through altered inter-subunit interactions. This phenomenon could be more general, as many of the ackA upregulated acetyllysines of the structurally analyzed metabolic enzymes (Table S9, Fig. 11 ) were located exterior to the active site.

43Acetylation could impact protein-nucleic acid and protein-protein interactions. Glucose induces cpxP transcription in WT cells in a manner that depends on lysine acetylation, likely via K298 of the α-CTD of RNAP, which is acetylated under these conditions. In an ackA mutant, however, cpxP transcription is inhibited in a K291-dependent manner, a genetic mimic of a constitutively acetylated K291 (K291Q) causes inhibition in the WT parent, and K291, but not K298, is upregulated about three-fold in ackA mutants grown in the presence of glucose (Table S9). K298 is involved in DNA-binding, while K291 is known to be involved in at least one protein-protein interaction, and thus could interact with CpxR, the response regulator and transcription factor required for cpxP transcription. That acP-dependent acetylation could influence protein-protein interactions is supported by the work of Eisenbach and colleagues, who have shown that acetylation of CheY diminishes its affinity for its 3 interaction partners, CheA, FliM, and CheZ.

44An intriguing possibility is that acP, as a very small molecule, could gain access to interior surfaces of proteins that enzymes cannot; for example, the interior of a large complex (e.g., the ribosome or RNAP) where the active sites are often located. Indeed, K1065 of the β subunit sits inside the RNAP complex where it helps align the first ribonucleotide of every nascent transcript. Acetylation and thus neutralization of this key lysine would be expected to render the acetylated complex inactive. Because a deacetylase would be unable to gain access to the interior of RNAP, this inaction almost certainly would be permanent. Since RNAP is a scarce resource, acetylation-mediated inactivation of even a small fraction of RNAP could have a large global impact. In the ackA mutant, K1065 was upregulated ∼3.5-fold.

45Like K1065 of the RNAP β subunit, other lysines located in active sites of proteins that bind a phosphorylated substrate could be acetylated by acP. Indeed, acetylation of a strategically located lysine could influence the binding of not only substrates, but also of co-factors and allosteric effectors. The likelihood that such an acetylation would occur should depend upon competition with the concentration of those substrates, co-factors, or effectors, which could protect against modification. Thus, detection of these acetylations should vary with the method of detection (e.g., in vitro versus in vivo). This might explain why, of the glycolytic enzymes that we analyzed, only a handful of the detected acetylated lysines were near the active sites.

46Finally, since acP contains a phosphoryl group, its ability to compete with phosphorylated substrates, co-factors, or effectors would not actually require modification; it could simply bind to sites normally occupied by phosphorylated substrates, effectors, or cofactors and sterically hinder their binding (e.g., Fig. 9B, 10E, F ).

47Although acP-dependent acetylation seemed to have a broader set of targets and appeared to generally be less specific than CobB-sensitive acetylation (Fig. S4A and S4B), acP-dependent acetylation showed high specificity for selected proteins and for selected lysines (Fig. S4D) that could be differentiated from those that were sensitive to CobB (Fig. S4E). Biochemical, bioinformatic and crystallographic efforts to identify the characteristics of an acP-acetylatable lysine revealed a tendency toward a surface-exposure microenvironment enriched for residues with positive charges, hydroxyls or amides, but often with an adjacent negatively charged residue. In contrast to previous reports, we extended our analysis to the structures of proteins that mass spectrometry had detected as acetylated and found a uniqe three-dimensional microenvironment ( Fig. 8 – 10 ).

48It would appear that such a micoenvironment requires two properties: an affinity for acP and a way to enhance the reactivity of the lysine. The affinity requirement appears to be satisfied by the formation of salt bridges between the negatively charged phosphoryl group of acP and positively charged residues of the protein to be acetylated. An alternative or additional contribution is from hydrogen bonds between the phosphoryl group and hydrogen bond donors such as hydroxyl and/or amide functional groups that surround the lysine. The enhancement of the reactivity of the lysine might be satisfied by an adjacent negatively charged residue, an observation reported recently for acetyllysines in Thermus thermophilus. We base this concept upon studies of acetyl transfer by members of different KAT families,,. Acetylation is proposed to occur by direct transfer of an acetyl group onto the substrate following formation a ternary complex between the substrate, acCoA, and the KAT,,. However, acetyl transfer requires deprotonation of the lysine. At high pH, protonation can occur non-enzymatically. At neutral pH, however, the epsilon amino groups of most lysine side chains remain largely protonated due to their high pKa's and thus are not good nucleophiles. Several mechanisms for catalytic deprotonation have been proposed. For example, the KAT can use a negatively charged residue (i.e., D or E) as a catalytic general base to deprotonate lysine and thus activate it for acetylation. Structural and mutational analyses of the eukaryotic KAT Gcn5 protein suggest that deprotonation occurs via a conserved glutamate residue (E173),. Mutational analysis of E173 determined that this residue was not involved in binding to either acCoA or its histone substrate. Because E173 was not required for catalysis at high pH, a condition that non-enzymatically deprotonates lysine, the authors proposed that E173 functions as a general base to deprotonate lysine for nucleophilic attack of acCoA at neutral pH. This mechanism appears to be conserved as catalytic activity for the eukaryotic KAT Rv0998 depends on E235, which is positioned similarly to the conserved E173 of Gcn5.

49Several divergent eukaryotic KATs possess residues required for catalytic activity that could act as general bases to deprotonate the lysine substrate,,,. For example, Naa50p-dependent acetylation depends on tyrosine (Y73) and histidine (H112), residues proposed to act as general bases. Finally, instead of a catalytic general base, some KATs have a series of residues that may form a water channel or a ‘proton wire’ that could shuttle protons from the active site to internal proton acceptors or to the solvent,,. For acP-dependent acetylation, similar mechanisms could operate.

50We conclude that E. coli possesses two mechanisms for lysine acetylation: the conventional KAT- and acCoA-dependent mechanism and the acP-dependent mechanism described here. The full extent of the YfiQ-dependent acetylation remains to be determined. While some have suggested diverse roles,, the only well-established role is to acetylate acetyl-CoA synthetase and thus regulate its activity. In contrast, the effect of non-enzymatic acetylation of proteins by acP appears to be quite extensive. Concerning this novel mechanism of lysine acetylation, several questions remain unanswered. How does the cell cope with increases in acP-dependent protein acetylation? Does acP-dependent acetylation simply accumulate over time or is the process regulated? For example, CobB could recognize and deacetylate acetyllysines acetylated by acP. This mechanism is suggested by the sensitivity of some sites (e.g., K154 of RcsB) to both acP and CobB, but other acP-sensitive sites seem to be unrecognized by CobB. What is the effect of acP-acetylation on protein function? Our data suggest that it can inhibit enzymatic activity by acetylating key lysine residues in active sites, allosteric sites, or cofactor-binding sites. But, non-enzymatic acetylation by acP also could influence protein stability, oligomerization and protein-protein interactions, protein-nucleic acid interactions, protein translocation to various regions of the cell, cellular signaling mechanisms, or even host-pathogen interactions. What is the stoichiometry of acP-dependent acetylation? The fraction that any given lysine is acetylated at any given moment should determine the degree to which that modification alters the cellular process in which it functions. For rate-limiting enzymes or proteins in scarce supply, the stoichiometry should be crucial. Finally, how does the full extent of acP-dependent acetylation influence cellular physiology? Mutants of the Pta-AckA pathway exhibit pleiotropic phenotypes. How many of these phenoytpes result from dysregulated acP-dependent acetylation? Since acP plays many roles (e.g., generating ATP, phosphorylating certain RRs and acetylating proteins),, the answer to this question will require a series of precise genetic dissections.

51Chemical reagents and media components were purchased from Fisher Scientific (Pittsburgh, PA), Sigma Chemical Company (St Louis, MO), Difco (Sparks, MD), J.T. Baker (Center Valley, PA), Cardinal Health (Dublin, OH) and United States Biochemical (Cleveland, OH). Nitrocellulose was purchased from BioRad (Germany). HPLC solvents including acetonitrile and water were obtained from Burdick & Jackson (Muskegon, MI). Reagents for protein chemistry including iodoacetamide, dithiothreitol (DTT), ammonium bicarbonate and formic acid were purchased from Sigma Aldrich (St. Louis, MO). HLB Oasis SPE cartridges were purchased from Waters (Milford, MA), and proteomics grade trypsin was from Promega (Madison WI). Trypsin-predigested β-galactosidase (a mass spectrometric quality control standard) was purchased from AB SCIEX (Foster City, CA).

52All bacterial strains used in this study are listed in Table 4. Derivatives were constructed by generalized transduction with P1kc, as described previously. For strain construction, cells were grown in LB containing 1% (w/v) tryptone, 0.5% (w/v) yeast extract, and 0.5% (w/v) sodium chloride; LB plates also contained 1.5% agar. For Western immunoblot and mass spectrometric analyses, cells were grown at 37°C in TB7 [1% (w/v) tryptone buffered at pH 7.0 with potassium phosphate (100 mM)] or TB7 supplemented with 0.4% glucose or 25 mM acetate. Cell growth was monitored spectrophotometrically (DU640; Beckman Instruments, Fullerton, CA) by determining the absorbance at 600 nm (A600). Spectinomycin (100 µg/ml), ampicillin (100 µg/ml), and chloramphenicol (25 µg/ml) were added to growth media when needed.

53Overnight cultures were diluted into fresh TB7 at an OD600 of 0.1. Cells were harvested at OD600s of 0.5 and 1.0 and then at hours 8, 24 and 32. Cells were pelleted and lysed with 4X sample buffer (9% SDS, 0.36 M Tris pH 6.8, 45% glycerol, 2% bromophenol blue, 10% β-mercaptoethanol) and heated at 100°C for 10 minutes. Protein loading was normalized after protein amount was determined by BCA assay (Pierce 23225). Purified proteins were separated by SDS-PAGE with 4.6 M urea. The gel and the nitrocellulose membrane were rinsed in water and then equilibrated for 15 minutes in transfer buffer containing 25 mM Tris, 0.192 M glycine, and 20% (v/v) methanol. Proteins were transferred onto the membrane for 1 hour at 100 V. The blot was blocked with 5% (w/v) milk prepared in TBST for 2 hours at room temperature. The blot was washed with PBST two times for 5 minutes each. A rabbit polyclonal antibody raised against an acetylated lysine-containing peptide (Cell Signaling 9441) was used at a 1∶500 dilution in 5% BSA in PBST at 4°C overnight. The blot was washed 3 times for 5 minutes each and then one time for 10 minutes with PBST and then incubated with HRP-conjugated goat anti-rabbit secondary antibody (Cell Signaling 7074S) at a 1∶2000 dilution in 5% milk in TBST for 1 hour at room temperature. The blot was washed 4 times for 5 minutes each with PBST and exposed using 20X LumiGLOR Reagent and Peroxide (Cell Signaling 7003).

54Cell pellets from WT and isogenic mutant strains grown in TB7 or TB7 supplemented with 4% glucose were suspended in 6 mL of PBS and centrifuged at 4°C, 15,000 g for 20 min. The firm cell pellet was collected and re-suspended and denatured in a final solution of 6 M urea, 100 mM Tris, 75 mM NaCl containing the deacetylase inhibitors 1 mM tricostatin A and 3 mM nicotinamide. Samples were sonicated on ice (5X each), cellular debris removed, and the supernatant for each sample further processed for proteolytic digestion. Typically 1.5 mg of protein lysate was reduced with 20 mM DTT (37°C for 1 hour), and subsequently alkylated with 40 mM iodoacetamide (30 min at RT in the dark). Samples were diluted 10-fold with 100 mM Tris pH 8.0 and incubated overnight at 37°C with sequencing grade trypsin (Promega) added at a 1∶50 enzyme:substrate ratio (wt/wt). Samples were then acidified with formic acid and desalted using HLB Oasis SPE cartridges (Waters). Proteolytic peptides were eluted, concentrated to near dryness by vacuum centrifugation, and re-suspended in NET buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA). A small aliquot of each protein digestion (∼10 µg) was further desalted (C-18 zip-tips) for total protein lysate mass spectrometric analysis. The remaining tryptic peptide samples were used for affinity purification of lysine-acetylated peptides.

55The polyclonal anti-acetyllysine agarose antibody conjugate from ImmuneChem (ICP0380-100) was prewashed and incubated with 1 mg E. coli digested protein lysate (in NET buffer) overnight (4°C) at a 1∶25 antibody:peptide ratio (wt/wt). Beads were washed three times in NET buffer, and the enriched acetyl peptides eluted by washing three times in 1% trifluoroacetic acid/40% acetonitrile. Samples were concentrated to near dryness by vacuum centrifugation and resuspended in equal amounts of 0.1% formic acid/1% acetonitrile. Prior to mass spectrometric analysis the acetyllysine peptide enrichment samples were desalted using C-18 Ziptips (Millipore, Billerica, MA).

56Samples were analyzed by reverse-phase HPLC-ESI-MS/MS using the Eksigent Ultra Plus nano-LC 2D HPLC system (Dublin, CA) connected to a quadrupole time-of-flight TripleTOF 5600 mass spectrometer (AB SCIEX). Typically, mass resolution for MS1 scans and corresponding precursor ions was ∼35,000, while resolution for MS2 scans and resulting fragment ions was set at 15,000, a ‘high sensitivity’ product ion scan mode. Details for the mass spectrometric and chromatographic methods are provided in the Methods S1. Briefly, samples were acquired in data-dependent mode (DDA) on the TripleTOF 5600 to obtain MS/MS spectra for the 30 most abundant precursor ions following each survey MS1 scan. Single colonies were grown overnight for ∼15 hours in TB7 at 37°C with aeration, were diluted to an OD600 of 0.05 in 500 ml TB7 supplemented with or without 0.4% (w/v) glucose, and aerated for 7.5 hour in 1 L flasks at 37°C. The cells were pelleted at 4°C and washed with ice cold PBS, pelleted at 4°C again, and frozen in a dry ice and ethanol bath. The pellets were stored at −20°C before shipment on dry ice. For E. coli strains grown in TB7 containing 0.4% glucose, we analyzed three (yfiQ) or four (WT, ackA, cobB) independent biological replicates, each grown approximately 1 month apart. Each biological replicate was subjected to 3 technical MS replicates (i.e., three injection replicates on the mass spectrometer). Similarly, all strains grown without glucose supplement were analyzed in 3 independent biological replicates, each with 3 technical MS replicates.

57For selected samples, additional data sets were recorded in data-independent mode (DIA) using MS/MSALL with SWATH acquisitions for quantitative analysis. During SWATH-MS2 acquisition, a wide window of ∼25 m/z is passed in Q1 incremental steps over the full mass range from m/z 400–1000 (see Methods S1). SWATH-MS2 analysis was performed to confirm MS1 Filtering quantitative results on one biological replicate (with three technical MS replicates) for WT and its isogenic ackA and cobB mutants, all grown in TB7 supplemented with 0.4% glucose.

58Mass spectrometric data was searched using the database search engine ProteinPilot (AB SCIEX Beta 4.5, revision 1656) with the Paragon algorithm (4.5.0.0, 1654). Mass spectral data sets also were analyzed using a Mascot server version 2.3.02 after initially generating peak lists with the AB SCIEX mgf data converter version 1.3. All extensive details including search parameters, fixed and variable modifications, enzyme specificity, databases used, scoring, and false discovery rate analysis (FDR) are described in the Methods S1.

59MS1 chromatogram-based quantitation was performed in Skyline, an open source software project (http://proteome.gs.washington.edu/software/skyline) using a MS1 ion intensity chromatogram processing called MS1 Filtering. Briefly, prior to MS1 Filtering, comprehensive spectral libraries were first generated in Skyline from database searches of the raw data files using the BiblioSpec algorithm. Second, all raw files acquired in data dependent acquisition mode (DDA) were directly imported into Skyline 1.4, and MS1 precursor ions were extracted for all peptides present in the MS/MS spectral libraries (the MS1 resolution setting in Skyline for precursor ion extraction is set at 10,000). Quantitative analysis was based on extracted ion chromatograms (XICs) and the resulting precursor ion peak areas for M, M+1, and M+2; final quantitative comparisons were typically based on only the highest ranked (naturally most abundant) precursor ion, which was then compared between samples.

60Data sets from MS/MSALL with SWATH acquisitions were processed using the full scan MS/MS filtering module for data independent acquisitions within Skyline 1.4. Selected samples (1 biological replicate) containing acetylpeptides were subjected to SWATH-MS2 analysis (for confirmation of the quantitative MS1 Filtering results) and subsequently processed extracting the top 4-6 MS2 product ions using a product ion resolution setting of 10,000 (peak areas for the sum of the top 4-6 product ions are processed).

61All peak areas obtained from extracted precursor ion chromatograms using MS1 Filtering were exported from Skyline into excel sheets using the statistical program ‘MS1Probe’, (http://www.gibsonproteomics.org/resources/MS1Probe). For each biological replicate and E. coli strain, the peak areas for individual acetyllysine peptides were automatically assembled in a data array to subsequently calculate peak area ratios per peptide between different strains, i.e. ackA/WT (for each strain, the 3 technical MS replicates were grouped). Significance was assessed using two-tailed Student's t-test requiring p-values<0.05. To assess possible regulation of acetylation in E. coli mutants relative to their WT parent, for a given acetyllysine-containing peptide, we set a threshold of at least a two-fold change in the relative peak area between the mutant and the WT with significance of p<0.05 within each biological replicate (each with 3 technical MS replicates). To be considered a ‘candidate’, such significant up-regulation was then required in at least 3 independent biological replicates out of the four biological replicates. Subsequently, for a given acetyllysine-containing peptide the mutant/WT ratio from each biological replicate was log-transformed, an average ratio over the 4 biological replicates with standard deviation and coefficient of variation (CV) calculated, and the obtained average log ratio was transformed back to a natural ratio for final reports. In cases where a specific acetyllysine site was identified with different precursor ion charge states or that produced multiple acetyllysine containing peptides with additional missed proteolytic cleavages or modifications, the most reproducible and most robustly ionizing species per site was chosen for quantification. To determine if the concentration of acetylated proteins that showed candidate acetylation sites for acetyl regulation was similar or different in the mutant relative to its WT parent, levels were assessed by quantifying non-modified peptides from the whole protein lysates.

62The raw and processed data associated with this manuscript may be downloaded at ftp://sftp.buckinstitute.org:244/Ecoli_acetylome/. All details for peptide quantitation using MS1 Filtering, including peptide peak areas for all MS replicates, are provided as excel files.

63A Skyline spectral library file that contains all MS/MS spectra of PTM-containing peptides identified and subsequently used for quantitation in MS1 Filtering, was transferred to the public data-sharing Panorama site. Panorama provides an interactive web-based spectral viewer of all acetyllysine-containing peptides identified in this study (at 99% confidence and with expectation value <0.01, respectively). The spectral viewer can be accessed at http://proteome.gs.washington.edu/software/panorama/gibson_ecoli_acetylome.html. All MS/MS spectral details underlying this data also are displayed in Tables S6 and S7. Acetylation site assignment was initially suggested by search engines ProteinPilot and Mascot and confirmed by manual inspection using previously defined criteria.

64For functional analysis and protein ontology analysis ‘The Database for Annotation, Visualization and Integrated Discovery’ (DAVID v.6.7) was used. For pathway enrichment analysis, we used the Bonferroni correction as adjustment to p-values to determine statistical significance. To generate graphical ontology displays, Panther v8.1 classification system was used.

65Individual peptides were synthesized in 96 well filter plates, using methods previously described. A library was constructed in the format Ac-GXKZGC-CONH2, where X and Z positions represent all amino acids except cysteine. A total of 361 peptides were synthesized and stored in 0.1% TFA in water in a 384 well plate.

66The Ac-GXKZGC-CONH2 peptide library was diluted into 50 mM Tris-HCl pH 8.0 to a final concentration of 20 µM in a 384 well format. A Tecan EVO 384 head was used to transfer 3 µl of buffered peptide to 384 SAMDI biochips of self-assembled monolayers on gold, prepared as previously described. The cysteine-terminated peptides were captured onto monolayers presenting maleimide as a capture moiety. Peptides were immobilized onto surfaces for 1 h at room temperature. SAMDI Biochips were then washed with copious amounts of 100% ethanol, deionized ultra-filtered (DIUF) water, 100% ethanol, and then dried with nitrogen gas. Peptides were tested by SAMDI.

67For acP reactions, SAMDI biochip arrays with the immobilized Ac-GXKZGC-CONH2 peptide library were subjected to a mixture of 50 mM acP, 150 mM NaCl, 10 mM MgCl2 and 50 mM Tri-HCl pH 8.0. A total of 3 µl of the solution was transferred to the biochips and incubated for 1 h at 30°C. A negative control containing the same reaction buffer in the absence of AcP was used to determine the background level of acetylation. Reactions for AcP and controls were performed in duplicate and were quenched by washing biochips with 100% ethanol prior to SAMDI analysis.

68A 3 ml solution of 2′,4′,6′-trihydroxyacetophenone monohydrate (THAP) matrix (30 mg/ml in acetone) was applied to the biochips and allowed to dry prior to SAMDI. Mass analysis was performed using a 5800 MALDI-TOF/TOF (AB Sciex). A 355 nm Nd:YAG laser was used as a desorption/ionization source, and all spectra were acquired with 20 kV accelerating voltage using positive reflector mode. The extraction delay was 700 ns, 500 laser shots were applied, and the entire surface of the spot was randomly sampled. Each acquired spectrum was examined for complete peptide immobilization by monitoring the conversion of the maleimide monolayer. The extent of the acP reactions was calculated as previously reported.

69A consensus sequence for acP-regulated lysine acetylation was generated using the mass spectrometric data. Acetylated lysines that were significantly up-regulated (ratio>2, p-value<0.05) in at least 3 out of 4 biological replicates of the ackA mutant relative to WT cells were used in this analysis. Of the 592 unique sites that fit this description, 28 and 24 sites were less than 10 residues from the N-terminus or the C-terminus; these were removed from the following analysis, leaving 541 acetylated sites. Sequences of 21 amino acids, representing 10 amino acids N-terminal and 10 amino acids C-terminal relative to the acetylated lysine were analyzed and a consensus logo was generated using WebLogo.

70ASKA clones of gapA, tpiA and lpdA from E. coli K-12 were used to produce protein for crystallization experiments. The clones contain a six-histidine N-terminal affinity tag followed by TDPALRA spacer residues and then the ORF of the gene ligated into the pCA24N expression vector as described. Proteins were expressed in BL-21 (DE3) Magic cells and purified using previously described protocols,,.

71An additional purification step was introduced for the TpiA protein. First, 30 mg of concentrated purified protein (10 mM Tris-HCl pH 8.3, 500 mM NaCl, 5 mM beta-mercaptoethanol) was diluted into 50 mM Tris-HCl pH 8.3 buffer to obtain a final salt concentration of 150 mM NaCl. Half of this sample was used as a control with no modification and the other half was used for a reaction with AcP. Two ml of protein sample were first incubated with 50 mM acP for 30 min at 30°C. The sample was then incubated a second time with 50 mM acP in a final volume of 3 ml for 30 min at 30°C. The samples were diluted to 50 ml in 50 mM Tris pH 8.3 and 150 mM NaCl and loaded onto an 8 ml MonoQ 10/10 GL column (GE Healthcare). Using a linear gradient from 0-70%, the proteins were eluted and fractions collected. Each peak of native TpiA and acP-treated TpiA was concentrated and used for crystallization trials with the Classics II and PACT screens (Qiagen). The PACT screen (Qiagen) was used for crystallization trials of GapA. All crystallizations were performed in a 96-well plate with 100 µl of reservoir solution and 1∶1 µl reservoir to protein (7.3 mg/ml for GapA and 7.4 mg/ml for TpiA) at 22°C. Crystals grew within a month.

72Crystals were harvested from PACT D11 (0.2 M calcium chloride, 0.1 M Tris pH 8, 20% (w/v) PEG 6000) for native TpiA, Classics F11 (0.2 M NaCl, 0.1 M Bis-Tris pH 6.5, 25% (w/v) PEG 3350 for acP-treated TpiA, and PACT G2 0.2 M sodium bromide, 0.1 M Bis Tris propane 7.5, 20% (w/v) PEG 3350 for AcP-treated GapA. Crystals not treated with acP were transferred to 5 µl of reservoir solution and mounted and frozen in liquid nitrogen for data collection. The crystals from Classics F11 and PACT G2 were treated with acP by soaking using the following procedure: crystals were transferred to 5 µl of the reservoir solution, AcP powder was added directly to the solution (approximately 10 mM final concentration) using a pipette tip, and crystals were quickly drawn through the solution, mounted and flash frozen with liquid nitrogen. The time between adding the acP and plunging the crystal into liquid nitrogen was less than 30 sec. Data were collected at 100 K at the Life Sciences Collaborative Access Team (LS-CAT) 21ID-D, 21ID-F and 21ID-G beamlines at Argonne National Laboratory (Argonne, IL). HKL3000 was used for data processing, integration, and scaling. The GapA and TpiA structures were solved by Molecular Replacement with Phaser in the CCP4 suite, using the GapA structure from E. coli (PDB ID: 1S7C) and the TpiA structure from E. coli (PDB ID: 1TRE) as the search models. The models were refined using Refmac and acetylated lysines were modified using the CCP4 monomer library. Manual corrections were performed on workstations using COOT and the final models were validated using the validation server of the PDB. PDB Codes: 4K6A (native tpiA), 4MVA (acP-treated TpiA), and 4MVJ (acP-treated GapA). Figures were made using Pymol (Delano, 2002) and structure parameters are shown in Table S29.

73The following E. coli crystal structures were used for mapping lysines on glycolytic enzymes that were modified in the ackA mutant: PfkB (PDB ID: 3UQD) with ATP and fructose 6-phosphate (F6P); Fbp (PDB ID: 2OWZ) with F6P; GlpX (PDB ID: 3DIR) with fructose 1,6-bisphosphate (FBP); FbaA (PDB ID: 1B57) with phosphoglycolohydroxamic acid (PGH); TpiA (PDB ID: 4K6A, 4MVA) with acP; GapA (PDB ID: 1DC4 and 1DC6) with G3P from 1DC4 and NAD+ from 1DC6; Pgk (PDB ID: 1ZMR) with phosphoaminophosphonic acid-adenylate ester (ANP) and 3-phosphoglyceric acid (3PG) of 1VPE; GpmA (PDB ID: 1E59) with VO3; Eno (PDB ID: 1E9I) with SO4; and PykF (PDB ID: 1PKY) without a ligand.

741.25 µM of LpdA and various concentrations of acP were incubated at 37°C in 150 mM Tris HCl (pH 7.3 at room temperature), 10% glycerol, 10 mM MgCl2, and 150 mM NaCl at the indicated amount of time. For each acP concentration, time points were prepared at 0, 5, 10, 15, 30 min and at 1, 2, 3, 4, 5, 7 hours. To assess variability, each of the 4 time courses (at the 4 acP concentrations) was performed in ‘incubation duplicates.’ To stop the reaction, an equal volume of 2X SDS loading buffer was added and the reactions were heated at 95°C for 10 minutes.

75RNAP was purified as described. 1.25 µM of RNAP complex were incubated with acP at 10 mM in the acP in vitro acetylation buffer described for LpdA and acP. Time points at 0 min and at 30 min were collected. To assess variability, the experiments were performed in ‘incubation duplicates.’

76Following incubation with acP, purified proteins were separated by SDS-PAGE. The gel and the nitrocellulose membrane were equilibrated for 15 minutes in transfer buffer containing 25 mM Tris, 0.192 M glycine, 20% (v/v) methanol, and 0.1% (w/v) SDS. Proteins were transferred onto the membrane for 30 min using a Biorad TurboTransfer. The blot was blocked with 5% (w/v) milk prepared in TBST for 1 hour at room temperature. The blot was washed with TBST for 5 min. An anti-acetylated lysine-containing peptide (Cell Signaling 9441) was used at a 1∶1000 dilution in 5% BSA in TBST at 4°C overnight. The blot was washed 3 times for 5 min each with TBST and then incubated with HRP-conjugated goat anti-rabbit secondary antibody (Cell Signaling 7074S) at a 1∶1000 dilution in 5% milk in TBST for 1 hour at room temperature. The blot was washed 3 times for 5 min each with TBST and exposed using 20X LumiGLOR Reagent and Peroxide (Cell Signaling 7003). Quantification of acetylation in terms of average pixel density was performed using AlphaView.

77Samples (from different acP concentrations and different time points) were electrophoresed, each separated by an empty lane(s) to avoid cross contamination as previously described. To limit cross-contamination between samples, each gel lane was cut out and stained in its own container with SimplyBlue™ SafeStain (Invitrogen) and subsequently destained with water, or gels were stained with Coomassie Brilliant Blue and destained using methanol and acetic acid. The bands containing LpdA were excised. The RNAP complex was separated by electrophoresis and 4 distinct bands (RpoB/RpoC, RpoD, RpoA, and RpoZ) were excised. Subsequently, all gel bands were subjected to tryptic in-gel digestion as previously described. Mass spectrometric analysis was performed on the TripleTOF 5600 in data-dependent mode with a total acquisitions runtime of 90 min per sample. For LpdA incubations, we analyzed the following: Series A - 10 mM acP at time points 0, 10, 60, 240, 420 min. Series B - 20 mM acP at time points 0, 10, 60, 240, 420 min. For each series, two independent incubation replicates were analyzed as two technical injection replicates. For RNAP incubations (10 mM acP), we analyzed time points at 0 and 30 min in two independent incubation replicates and two technical injection replicates.

78pKa values for LpdA and RNAP subunits were calculated using PROPKA 3.1 at http://propka.ki.ku.dk/. PROPKA also calculates surface accessibility, defining a parameter referred to as buried ratio. As input for PROPKA, the following PDB structural files were used, PDB ID: 4JDR for LpdA, PDB ID: 4IGC for RNAP chain A (RpoA), chain C (RpoB), chain D (RpoC), and chain X (RpoD).