Identification and characterization of two types of amino acid-regulated acetyltransferases in actinobacteria

Abstract One hundred and fifty GCN5-like acetyltransferases with amino acid-binding (ACT)-GCN5-related N-acetyltransferase (GNAT) domain organization have been identified in actinobacteria. The ACT domain is fused to the GNAT domain, conferring amino acid-induced allosteric regulation to these protein acetyltransferases (Pat) (amino acid sensing acetyltransferase, (AAPatA)). Members of the AAPatA family share similar secondary structure and are divided into two groups based on the allosteric ligands of the ACT domain: the asparagine (Asn)-activated PatA and the cysteine (Cys)-activated PatA. The former are mainly found in Streptomyces; the latter are distributed in other actinobacteria. We investigated the effect of Asn and Cys on the acetylation activity of Sven_0867 (SvePatA, from Streptomyces venezuelae DSM 40230) and Amir_5672 (AmiPatA, from Actinosynnema mirum strain DSM 43827), respectively, as well as the relationship between the structure and function of these enzymes. These findings indicate that the activity of PatA and acetylation level of proteins may be closely correlated with intracellular concentrations of Asn and Cys in actinobacteria. Amino acid-sensing signal transduction in acetyltransferases may be a mechanism that regulates protein acetylation in response to nutrient availability. Future work examining the relationship between protein acetylation and amino acid metabolism will broaden our understanding of post-translational modifications (PTMs) in feedback regulation.


Introduction
Protein lysine acetylation has emerged as an important metabolic regulatory mechanism in bacteria, since the discovery of reversible acetylation of acetyl-CoA (Ac-CoA) synthetase via enzymatic acetylation, such as acetyltransferase [1] and deacetylase [2]. Protein acetylation also can occur via nonenzymatic reaction, such as chemical acetylation, wherein the intracellular acetyl-phosphate (AcP) plays a critical role [3,4]. The intracellular level of protein acetylation is ultimately defined by the concentration of AcP (or Ac-CoA) and relative activities of acetyltransferases and deacetylases. These two classes of enzymes respond to intracellular nutritional status to control the acetylation of specific proteins that mold the metabolic network [5,6]. Multiple nutrient-mediated signaling pathways regulate the transcription of genes encoding acetyltransferases and deacetylases. In Bacillus subtilis, regulating the expression of the acuA gene, which encodes an acetyltransferase, is under the control of catabolite control protein A (CcpA), a global regulatory protein affected by the quality of the carbon source available to the cell [5]. In Escherichia coli, two acetylation-signaling pathways mediated by cAMP have been identified. The carbon regulating CRP-cAMP complex induces the expression of yfiQ, which encodes the Gcn5-like acetyltransferase YfiQ

In vitro assays of Acs
The specific activity of Acs was determined at 30 • C in a transparent 384-well microplate reader (Bio Tek Instruments, Winooski, U.S.A.) at 340 nm [16]. The standard reaction mixture contained 100 mM Tris/HCl (pH 7.8), 10 mM malate (pH 7.7), 0.2 mM CoA, 8 mM ATP (pH 7.5), 1 mM NAD + , 10 mM MgCl 2 , 3 units of malate dehydrogenase, 0.4 units of citrate synthase, and 0.3 μM SveAcs or AmiAcs. The reaction was initiated with 100 mM potassium acetate and the rate was determined continuously for 5 min by measuring NADH production at 340 nm. One unit was defined as the conversion of 1.0 μmol of NAD + to NADH per min in the presence of saturating levels of acetate.

CD spectroscopy
CD spectra were prepared using a CD spectrophotometer (Applied Photophysics, U.K.) and 1 mm cuvette equilibrated at 25 • C. All samples were desalted and dialyzed against buffer P (150 mM KF, 94 mM K 2 HPO 4 , and 6 mM KH 2 PO 4 , pH 8.0). Samples were diluted to 0.25 μg/μl and incubated with and without 2 mM Asn or 2 mM Cys. The CD spectra were collected between 190 and 260 nm, and three repeat scans were averaged. Parameters of the secondary structure were analyzed using CDNN 2.1 (http://bioinformatik.biochemtech.uni-halle.de/cdnn).

Site-directed mutagenesis and purification of PatA and Acs
Mutations at key sites in PatA and Acs proteins were introduced into pET28a using QuikChange Mutagenesis Kit (TransGen, Beijing, China) according to previously described procedure [12,17]. The mutations were confirmed by DNA sequencing. The overexpression and purification of mutant proteins were performed as described above.

Structural model of the ACT domain of AmiPatA
The structure of the ACT domain of AmiPatA was built by homology modeling from Swiss Model server (https: //swissmodel.expasy.org/) using four ACT domains with the greatest similarity to that of AmiPatA (PDB ID: 3W7B, 3O1L, 1PSD, 2J0X) according to the previously described procedure [12]. Discovery Studio 3.5 Client was used as the docking tool to predict amino acid-binding to AAPatA proteins. CHARMM forcefield was used to optimize the model of the ACT domain, and CFF forcefield was used to optimize the structure of Cys. We added Kollman charge to the residues of ACT domain of AmiPatA. LibDock was used to perform molecular docking. The conserved binding site was selected as docking center, and a sphere with a radius of 10Å were considered as binding region for docking studies.

Sequences and phylogenetic tree analysis of ACT and GNAT domains
Multiple alignments of the ACT and GNAT domains, as well as a phylogenetic analysis of full-length sequences of all ACT-GNAT proteins, were performed using ClustalW 2.1 and MEGA 5.10 software.

Data and statistical analysis
Microsoft Excel was used to quantitate protein concentrations and the activities of PatA, SveAcs, and AmiAcs. Graph-Pad Prism 5.0 was used for statistical analyses, preparation of graphs, and data analysis. Data are presented as means + − S.E.M. for three independent experiments (n=3); statistical significance was accepted at P 0.05.

PatA acetyltransferases with ACT-GNAT domain organization are found in actinobacteria
Previous work has shown that MaKat (Micau 1670, from M. aurantiaca) containing an ACT-GNAT domain organization is an amino acid-regulated Pat that can acetylate Ac-CoA synthetase [12]. One hundred and fifty-six proteins containing the ACT-GNAT domain organization were displayed in the InterPro database (v.54.0) and the sequences of these proteins were examined. Six repeated sequences were deleted. A total of 150 unique ACT-GNAT acetyltransferases, which we designated as AAPatAs, are listed in Supplementary Table S1. The AAPatA enzymes are unique to and are distributed in 22 species of actinobacteria. Remarkably, 107 AAPatAs (71.3%) were found in Streptomyces. Some AAPatAs were found in Amycolatopsis (10), Actinoplanes (5), and Micromonospora (5). We performed a phylogenetic tree analysis using the protein sequences of all AAPatAs ( Figure 1). Figure 1 shows that the sequence of AAPatA reflects the phylogeny of actinobacteria. In AAPatA sequence phylogeny, organisms are clustered according to orders, such as Streptomycetales (Streptomyces, Kitasatospora), Micromonosporales (Actinoplanes, Micromonospora, Verrucosispora), Pseudonocardiales (Saccharopolyspora, Pseudonocardia, Saccharothrix, Actinosynnema, Kibdelosporangium, Alloactinosynnema, Actinokineospora, Prauserella, Amycolatopsis), and Streptosporangiales (Streptomonospora, Microbispora, Streptosporangium). Exceptions included A4FK80 from Saccharopolyspora erythraea DSM 40517 (Pseudonocardiaceae), which was between Streptomycetales and Micromonosporales, and A0A0N0T6Y4 from Nocardia sp. NRRL S-836 (Corynebacteriales), which was dispersed amongst the Pseudonocardiales. The results revealed the sequence evolution of AAPatA acetyltransferases in actinobacteria, which can be used as a molecular marker for speciation of actinobacteria.

AAPatA can be divided into two groups: Cys-and Asn-dependent Pat
We overexpressed and purified 24 AAPatA acetyltransferases (indicated by the red dot in Figure 1) from all the ACT-GNAT proteins ( Table 1). Like MaKat from M. aurantiaca and other Pat (SlPat, Af Pat, 1YGH A: yeast, 1Z4R A: human, 1QSN A: TeT), the most highly conserved motif QXXGX (G/A) for Ac-CoA recognition and binding, and a glutamic acid residue provides as an active site to deprotonate the lysine residue of the acetyl acceptor, were observed in GNAT domains of all 24 AAPatAs (Supplementary Figure S1). The experimental data showed that all AAPatAs can directly acetylate the putative AMP-forming Ac-CoA synthetase (AmiAcs from Actinosynnema mirum strain DSM 43827) in vitro (Supplementary Figure S2). These results demonstrate that AAPatA is a novel type of Pat.
The ACT domain binds to more than one amino acid, by inducing conformational changes, it can generate variation in the activity of the catalytic domain. To identify the amino acid ligands of ACT domains in AAPatA acetyltransferases, we investigated the activities of all AAPatAs acetylating putative Ac-CoA synthetases with and without the addition of the 20 l-amino acids. Acetylation was measured continuously using a coupled enzymatic assay to monitor increase in absorbance at 340 nm ( A 340 ). These increases resulted from the formation of NADH by pyruvate dehydrogenase because of AAPatA-dependent CoA production [12][13][14]. The ACT domain allosterically regulates acetylation activity of AAPatA in response to addition of amino acids, as shown in Supplementary Figure S2. Adding Cys or Asn increased the values of A 340 , while no change in the values of A 340 were observed after the addition Cys of other amino acids. These increases in the values of A 340 suggest that the acetylation activity of AAPatAs is enhanced by the conformational changes resulting from Cys or Asn binding the ACT domain. The amino acid ligands of AAPatAs are listed in Table 1. We found that members of the AAPatA family are divided into two groups based on the allosteric ligands of the ACT domain: the Asn-regulated PatAs and Cys-regulated PatAs. The former are mainly found in Streptomyces; the latter are distributed in other actinobacteria. Furthermore, we performed a phylogenetic analysis of the sequences that make up the GNAT and ACT domains in the AAPatA proteins. As shown in Figure 2, the two groups were clustered according to the binding ligands, Asn (indicated by the red lines) or Cys (indicated by the blue lines) in both ACT and GNAT domains.
Multiple alignments of ACT domains in the 24 PatA proteins revealed that all ACT domains contain the conserved βαββαβ-fold and the highly conserved glycine residue between the first β1 strand and the first α1 helix; Asn-binding ACT domains have higher sequence conservation than the Cys-binding ACT domains (Figure 3). The β1 and α1 are the conserved motifs for amino acid recognition and binding [18]. As indicated by the red frame in Figure 3, conserved residues, such as DXPGXL, at the interface of β1 and α1, may play an important role in the binding of specific amino acids. Bioinformatics analysis and coupled enzymatic assay showed that AAPatAs, the protein lysine acetyltransferases (KATs) that acetylate the Ac-CoA synthetase, are divided into two groups: Cys AAPatA and

Cys enhances protein lysine acetylation by AmiPatA derived from A. mirum
To examine whether Cys regulates the enzymatic activity of A. mirum-derived AmiPatA, we used a spectrophotometric assay to assess the acetylation of AmiAcs by AmiPatA in the presence and absence of Cys. As shown in Figure  4A, the acetylation of Acs was enhanced in the presence of Cys, indicating an increase in the activity of the acetyltransferase AmiAcs. The initial rate of acetylation of AmiAcs by AmiPatA (indicated by the formation of NADH) was estimated to be 3.8 μM/min, which increased to 6.8 μM/min (an increase by 179%) in the presence of Cys. The acetylation level of AmiAcs was assessed by Western blotting using an anti-acetyl-lysine antibody (Anti-AcK) in the presence or absence of Cys. Cys increased the acetylation level of AmiAcs ( Figure 4B); furthermore, increases in the acetylation level of AmiAcs were dependent on the concentration of Cys ( Figure 4C). When the concentration of Cys was increased from 0.004 to 0.5 mM, the acetylation levels also increased. The concentrations of Cys ligand that saturated the acetylation of the protein were in the submillimolar range. The concentration is within the endogenous ranges reported for Cys, which are estimated to be 0.1-0.2 mM in E. coli cells [19], indicating that intracellular Cys is able to exert an effect on activity of AmiPatA enzyme in vivo. To investigate the effect of acetylation on enzyme activity, AmiAcs was incubated with AmiPatA in the presence or absence of Ac-CoA for 3 h. The activity of AmiAcs was reduced by half in the presence of Ac-CoA and AmiPatA, indicating that lysine acetylation effectively decreased the activity of AmiAcs ( Figure 4D). The acetylated AmiAcs protein was subjected to trypsin digestion, and the generated peptides were analyzed by MS. The acetylated lysine site (SGK (620) IMR) was identified in the conserved PXXXXGK motif of the AMP-forming acyl-CoA synthetases; this motif is recognized by the previously reported    bacterial acetyltransferases [17] (Supplementary Figure S3). We further determined the kinetic parameters of the AmiPatA catalyzed acetylation reaction of AmiAcs in the absence and presence of Cys using a coupled enzymatic assay previously described [12]. The data are shown in Table 2. The K m and k cat values of AmiPatA for AmiAcs are 45.5 μM and 0.03 s −1 , respectively. Cys was able to decrease the K m value of AmiPatA for AmiAcs by three-fold, and increase k cat value by 1.5-fold. The ACT domain is an amino acid-binding domain involved in amino acid metabolism; 3-phosphoglycerate dehydrogenase (3PGDH) in E. coli was the first enzyme found to contain the ACT domain [20] and folds with a ferredoxin-like topology. In the ACT domain, the most conserved part is the region at the interface between the first strand (β1) and the first helix (α1); the residues binding specific amino acids via hydrogen bonding are found at this conserved region [21]. To investigate key residues used to bind Cys in AmiPatA, the structure of the ACT domain in AmiPatA was modeled using four ACT domains with the greatest similarity (query cover: 85%, identity: 25%) to that of AmiPatA (PDB ID: 3W7B, 3O1L, 1PSD, 2J0X). The docking result showed that the residues involved in hydrogen bonding with Cys were Asp 16 , Ser 17 , Gly 19 , Ala 20 , and Leu 21 ; in particular, Asp 16 and Ser 17 were located in the loop between the first strand (β1) and the first helix (α1) of the AmiPatA ACT domain ( Figure 5A). The characteristic Gly 19 , followed by a hydrophobic residue, is the most conserved residue in the helix, necessary for maintaining the conformation of the strand-helix interface. Far-UV CD spectroscopy was conducted to confirm the conformational change in AmiPatA after the addition of Cys ( Figure 5B). CD spectra, recorded from 190 to 260 nm, showed an increase in ellipticity at 208 and 222 nm, indicating a decrease in the α-helical content of AmiPatA after the addition of Cys and suggesting that the binding of Cys altered the secondary structure of the protein. The characteristics of the secondary structure, calculated using CDNN version 2.1, indicated the presence of approximately 45.1% α-helices and 11.1% β-sheets in AmiPatA, and 35.8% α-helices and 15.8% β-sheets in AmiPatA after the addition of Cys. This phenomenon was observed in some proteins regulated allosterically by effectors [22,23].
To further identify the key residues involved in the binding of Cys, we mutated the acidic amino acid Asp 16 and the hydrophilic amino acid Ser 17 to a neutral and hydrophobic amino acid alanine. The CD spectra of D16A and S17A mutant proteins showed that Cys does not induce changes in the secondary structures of AmiPatA mutants ( Figure 5C,D). We also used an enzyme-coupled assay and Western blotting analysis to investigate the acetylation activity of the two mutants with and without the addition of Cys. As shown in Figure 5E,F, no increases in acetylation were observed in D16A and S17A mutants after the addition of Cys; this agrees with the CD assays that indicated no conformational changes induced by the addition of Cys. These results suggest that Asp 16 and Ser 17 at the interface between β1 and α1 may somehow affect the Cys-binding of AmiPatA. Taken together, these results indicate that AmiPatA has protein-acetylating acetyltransferase activity and that this activity is regulated allosterically by Cys binding.

Asn enhances protein lysine acetylation by SvePatA from S. venezuelae
Finally, we characterized SvePatA from S. venezuelae as a model representative of Asn-binding AAPatA acetyltransferases. To investigate the effect of Asn on the activity of SvePatA, we determined the initial rate of SveAcs acetylation by SvePatA with and without the addition of Asn. For this, we used a coupled enzymatic assay, where the amount of CoA released after SveAcs acetylation was measured by the formation of NADH from NAD + via pyruvate dehydrogenase. As shown in Figure 6A, SvePatA acetylated SveAcs in the absence of Asn (2.23 nmol of NADH formed per min/ml), and the rate of acetylation was increased by 176% in the presence of Asn (3.92 nmol of NADH formed per min/ml). Western blotting was conducted to detect the level of SveAcs acetylation with an anti-Ack antibody in the presence and absence of Asn. Acetylation of SveAcs was clearly observed after incubation with SvePatA and Ac-CoA; acetylation was significantly increased in the presence of Asn ( Figure 6B). Moreover, the increase in the acetyltransferase activity of SvePatA was dependent on the concentration of Asn ( Figure 6C). Previous work showed that Asn can reach cytoplasmic concentration of 0.51 mM in E. coli cells [24], which is comparable with the concentration of Asn needed to regulate allosterically activity of SvePatA enzyme. To examine the effect of acetylation on Western blotting analysis are shown in the AmiPatA mutants D16A and S17A. The gray values were quantitated by densitometry using ImageJ software. enzyme activity, SveAcs was incubated with SvePatA in the presence of Ac-CoA for 2 h. The activity of SveAcs was reduced by approximately 60%, indicating that lysine acetylation effectively decreased the activity of SveAcs ( Figure  6D). These results demonstrated that AAPatA proteins possessing the ACT-GNAT domain in non-Streptomycetaceae actinobacteria can function as acetyltransferases regulated allosterically by Asn.
As shown in Figure 7A, CD spectra of SvePatA displayed an increase in ellipticity at 208 and 222 nm after the addition of Asn, indicating that the binding of Asn altered the secondary structure of the protein. The docking result of Asn to SvePatA showed that the residues involved in hydrogen bonding with Asn were Asp 109 , Lys 112 , Pro 113 , Glu 114 , and Cys 139 ( Figure 7B). To further identify the key residues involved in the binding of Asn, we mutated the acidic amino acid Lys 112 and amino acid Pro 113 to alanine. Allosteric effect of Asn on SvePatA activity was still observed in two mutants K112A and P113A ( Figure 7C,D), indicating that Lys 112 and Pro 113 were not involved in the Asn binding of SvePatA.

Discussion
N-lysine acetylation is a dynamic, reversible, and regulatory post-translational modification (PTM) used to modulate enzyme activity in prokaryotes. Accumulating evidence suggests that metabolic networks are co-ordinated via reversible acetylation of enzymes that regulate cellular metabolism to maintain homeostasis in the rapidly changing microenvironment. The present study examined the archetypical acetyltransferases AAPatAs possessing GNAT and ACT domains, and shows a novel signaling pathway for regulating the acetylation of cellular proteins. These findings indicate that in actinobacteria, acetyltransferase activity and protein acetylation may be tightly correlated with intracellular amino acid metabolism. The conventional mechanism for reversible N ε -acetylation of proteins is enzyme-catalyzed acetylation and deacetylation, which relies on KATs and deacetylases (HDACs). Reversible enzyme-catalyzed acetylation is regulated at different levels, such as transcriptionally and allosterically, to adjust the acetylation of specific proteins responding to intracellular nutritional signals. In B. subtilis, the acuA gene encoding an acetyltransferase is controlled by the carbon catabolite control protein, CcpA [5]. In E. coli, the cAMP receptor protein CRP induces the expression of yfiQ, which encodes the acetyltransferase YfiQ, thus increasing the acetylation level of proteins in response to the intracellular cAMP signal [6]. In S. enterica, lolR, the repressor of myo-inositol catabolism, activates the transcription of pat and cobB [10]. There are three classes of KATs that catalyze the donation of an acetyl group from acetyl-coenzyme A (Ac-CoA) to an ε-amino group of a deprotonated lysine. These KATs include the GNAT family, the MYST family, and the p300/CBP family [25]. The genomes of actinobacteria encode approximately 40-80 GNAT acetyltransferases, 1-2 NAD + -dependent sirtuin deacetylases, or 1 NAD + -independent protein deacetylase. The genome of S. coelicolor encodes 77 putative GNAT acetyltransferases (Pfam00583), 2 sirtuin-type deacetylases (SCO0452, SCO6464), and an AcuC-like NAD + -independent deacetylase (SCO3330). The genome of S. avermitilis encodes 43 putative GNAT acetyltransferases, 1 sirtuin-type deacetylase (SAV 537), and 1 AcuC-like deacetylase (SAV 4729). The genome of S. lividans encodes 72 putative GNAT acetyltransferases, 2 sirtuin-type protein deacetylases (EFD65580, EFD71509), and 1 NAD + -independent deacetylase (EFD68590). The genome of Saccharopolyspora erythraea encodes 42 putative GNAT acetyltransferases, 1 NAD + -dependent deacetylase (SACE 3798), and 1 NAD + -independent deacetylase (SACE 1779). Thus, the number of GNAT acetyltransferases in a bacterial species may reflect the metabolic complexity of the species. Thus far, all the known bacterial Pat identified contain the GNAT domain having low sequence homology, and sharing a conserved motif A, sequence (R/Q)XXGX(G/A) for Ac-CoA recognition and binding, and a glutamate residue providing as an active site. It is worth noting that there has been considerable debate about the function of the conserved glutamate as a catalytic base [26,27]. In prokaryotes, some Pat have two different domain organizations: the multidomain (Class I) and single-domain (Class II). Several single-domain GNAT proteins have been shown to have acetyltransferase activity. These Class II enzymes, which are approximately 200 residues long, include BsAcuA from B. subtilis [28], RpKatA from Rhodopseudomonas palustris [29], and SacAcuA from Saccharopolyspora erythraea [17].
Class Iα Pat, such as the first discovered SePat in S. enterica [1], the homologs (EcPatZ from E. coli and RpPat from R. palustris) and variants (SlPatA from Streptomyces lividans) of SePat [6,29,30], are 800-1000 residues long and contain a GNAT catalytic and NAD + -forming CoA ligase domains; the function of the latter is undetermined, but it has homology with ADP-forming acyl-CoA synthetases. The catalytic mechanism and domains function of SePat and EcPatZ have been investigated [31,32]. Biochemical and thermodynamic analyses of SePat revealed that SePat has two distinct sites for Ac-CoA binding, and shows the positive co-operativity (with a Hill coefficient of