Insights Into the Specificity of Lysine Acetyltransferases

Results : The structure of a GNAT was determined in complex with a protein substrate. Conclusion : Specificity of the interaction

diversity of protein substrates. Thus, a detailed characterization of the GNAT-substrate interface is critical for understanding GNAT specificity for protein targets given the prevalence of lysine in biological proteomes.
In bacteria many members of the multidomain AMP-forming acyl CoA synthetase family are regulated by Nε-lysine acetylation as was first identified in Salmonella enterica for acetyl-CoA synthetase (SeAcs WT ) (8). Acetylation of residue K609 of SeAcs WT by the S. enterica protein acetyltransferase Pat (SePat WT ) leads to enzyme inactivation that can be reversed by the CobB sirtuin-type NAD +dependent deacetylase (Fig. 1A) (9). SePat WT consists of a GNAT catalytic domain and a larger domain of unknown function (Fig. 1B) (10). Studies of SePat WT homologues in other bacteria show that these acetyltransferases can regulate a wide range of acyl-CoA synthetases which all share a conserved PX 4 GK motif near the C-terminus of these proteins. The lysine in this motif is catalytic and is the target of acetylation. The roles of the proline and glycine residues are unknown and are not critical for acyl CoA synthetase activity (11). Interestingly, the presence of the PX 4 GK motif is necessary but not always sufficient for acetylation of these enzymes suggesting that additional determinants lie outside the signature sequence (12).
To gain insight into how specificity governs recognition and modification in a reversible lysine acetyl-lysine signaling pathway, we characterized the interaction interface between a GNAT and a protein substrate. We determined the structure of the GNAT domain of the Streptomyces lividans PatA (SlPatA WT ) enzyme (Figs. 1C, 1D), and used the ClusPro 2.0 server to identify potential interacting surfaces between the GNAT domain of SlPatA (hereafter SlPatA GNAT ) and the C-terminal domain of SeAcs (hereafter SeAcs CTD ), a substrate of SlPatA WT for which a three-dimensional crystal structure was available (RCSB PDB # 1PG3, 1PG4) (13). We demonstrated that SlPatA GNAT acetylated SeAcs CTD in vivo and in vitro, and that reversing charges near a predicted interface prevented interaction and SeAcs CTD acetylation. To test the biological relevance of the interaction model, we tethered the SlPatA GNAT domain to SeAcs CTD in multiple orientations with crosslinkers that resulted in spacing between the proteins ranging between 2 and 20 Å. We identified a single orientation that resulted in acetylation of SeAcs CTD when SlPatA GNAT was linked to SeAcs CTD by a direct disulfide bond, demonstrating catalysis in an enzyme-substrate complex with limited movement. We report the structure of the catalytic complex comprised of the SlPatA GNAT domain and SeAcs CTD at 1.9Å resolution. The structure revealed a constellation of determinants needed for recognition of a protein substrate by the SlPat GNAT domain.
Molecular techniques. DNA manipulations were performed using standard techniques (17). Restriction endonucleases were purchased from Fermentas. DNA was amplified using Pfu Ultra II Fusion DNA polymerase (Agilent) or Herculase II Fusion DNA polymerase (Agilent). Site-directed mutagenesis was performed using the Quikchange TM Site Directed Mutagenesis kit (Agilent). Plasmids were isolated using the Wizard Plus SV Miniprep kit (Promega) and PCR products were purified using the Wizard SV Gel and PCR Clean-Up System (Promega). DNA sequencing was performed using BigDye® (ABI PRISM) protocols, and dyeterminator sequencing reactions were resolved and analyzed by capillary electrophoresis at the University of Georgia Genomics Facility. Oligonucleotide primer sequences are listed in Table 3.
Construction of complementation plasmids. The allele encoding the GNAT domain of wild-type SlPatA (residues M1-L194) was amplified from S. lividans TK24 genomic DNA using primers listed in Table 3. Primers used in the amplification changed the first codon of SlPatA from TTG to the more common ATG start codon. The amplified fragment was cut with EcoRI and KpnI and ligated into pBAD30 cut with the same enzymes. The resulting plasmid, pSlPatA11, directed the synthesis of SlPat GNAT in response to L-(+)-arabinose. The plasmid directing the synthesis of variant SlPatA GNAT E123Q was generated by site directed mutagenesis using primers listed in Table 3.

Plasmids for in vivo two-hybrid system assays.
Alleles encoding the wild-type SlPatA GNAT domain, variant SlPatA GNAT E121R, and variant SlPatA GNAT D185R were amplified from plasmids pSlPatA27, pSlPatA33, and pSlPatA35, respectively, using primers listed in the Table 3. Alleles encoding wild-type SeAcs CTD (residues D518-S652), variants SeAcs CTD R606E and SeAcs CTD R613D were amplified from plasmids pACS38, pACS42, and pACS44, respectively, using primers listed in the Table 3. DNA fragments were cut with NotI and BamHI, then ligated into plasmids pACTR-V-Zif-AP and pBRωGP, which had been cut with NotI and BamHI. The resulting plasmids are listed in Table 2.
SlPatA GNAT overproduction plasmids. The allele encoding the GNAT domain of SlPatA (residues M1-L194) was amplified from S. lividans TK24 genomic DNA with primers listed in Table 3. The primers used were designed to change the first codon of SlPatA from TTG to the more common ATG start codon. The amplified fragment was cut with KpnI and SalI and ligated into pKLD66 (18) cut with the same enzymes. The resulting plasmid pSlPatA14 directed the synthesis of wild-type SlPatA GNAT fused at its Nterminus to a H 6 -Maltose binding protein (MBP) tag cleavable by recombinant Tobacco Etch Virus (rTEV) protease using described protocols (19). Plasmids directing the synthesis of variants SlPatA GNAT S73C, SlPatA GNAT A110C, and SlPatA GNAT A164C were generated from plasmid pSlPatA14 using site-directed mutagenesis using primers listed in Table 3.

SeAcs
C-terminal domain (SeAcs CTD ) overexpression plasmids. The allele encoding SeAcs CTD (residues D518 to S652) was amplified from S. enterica LT2 DNA using primers listed in Table 3. The amplified fragment was cut with NheI and EcoRI and ligated into pTEV5 cut with the same enzymes. The resulting plasmid pACS38 directed the synthesis of wild-type SeAcs CTD fused at its N-terminus to a hexahistidine tag cleavable by rTEV protease. Plasmids directing synthesis of variants SeAcs CTD R606E, SeAcs CTD R613D, SeAcs CTD A238C, SeAcs CTD H567C, and SeAcs CTD D600C were generated from plasmid pACS38 by sitedirected mutagenesis using primers listed in Table 3.

Full-length SeAcs overproduction plasmids.
Plasmids directing the synthesis of variants SeAcs CTD R606E and SeAcs CTD R613D were generated from plasmid pACS33 ( Table 2) using site-directed mutagenesis. The resulting plasmids pACS60 and pACS61 directed the synthesis of variants SeAcs CTD R606E and SeAcs CTD R613D, respectively, each fused to a N-terminal H 6 tag cleavable by rTEV protease.
SlPatA GNAT overproduction and purification. Plasmids pSlPatA14, pSlPatA53, pSlPatA54, pSlPatA56 were transformed into the Δpka derivative of E. coli C41 (λDE3) (strain JE9314). The resulting strains were grown overnight and sub-cultured 1:100 (v/v) into 2 liters of LB containing ampicillin (100 µg/ml). The cultures were grown shaking at 25°C to A 600 ∼ 0.7 and H 6 -MBP-SlPatA GNAT synthesis was induced with IPTG (0.5 mM). Upon induction, the cultures were grown overnight at 25°C. Cells were harvested at 6000 x g for 10 min at 4°C in a Avanti J-2 XPI centrifuge fitted with rotor JLA-8.1000 (Beckman Coulter). Cell pellets were re-suspended in 30 ml cold His-Bind buffer A [tris(hydroxymethyl)aminomethane-HCl (Tris-HCl) buffer (50 mM, pH 8), NaCl (500 mM)], and imidazole (5 mM)) containing phenylmethanesulfonylfluoride (PMSF, 1 mM). Cells were placed on ice and lysed by sonication for 1 min (2-s pulse followed by 4 s of cooling) at level 7 in a model 550 sonic dismembrator (Fisher). The extract was cleared by centrifugation at 4°C for 30 min at 43,367 × g. His 6 -tagged SlPatA GNAT were purified from the soluble fraction by Ni-affinity purification using a 1-ml bed volume of His-Pur Nickel-NTA Resin (Thermo). Unbound proteins were eluted off the column by extensive washing with buffer A. The column was then washed with buffer B1 [Tris-HCl buffer (50 mM, pH 8), NaCl (500 mM), and imidazole (15 mM)] and SlPatA GNAT was eluted from the column with buffer C [Tris-HCl buffer (50 mM, pH 8), NaCl (500 mM), and imidazole (250 mM)]. All fractions containing H 6 -MBP-SlPatA GNAT were combined. rTEV protease was added to H 6 -MBP-SlPatA GNAT and the SlPatA GNAT /rTEV mixture was incubated at room temperature for 3 h. The SlPatA GNAT /rTEV mixture was dialyzed at 4°C against buffer C (Tris-HCl (50 mM, pH 8), NaCl (500 mM)) twice for 3 h and again against buffer C containing imidazole (5 mM) for 12 h. After cleavage and dialysis, protein mixtures were passed over the 1-ml HisTrap column using the buffers described above and the untagged proteins eluted in the flow-through and buffer A wash. Purified SlPatA GNAT was analyzed by SDS-PAGE. Fractions containing SlPatA GNAT were pooled together. SlPatA GNAT was stored in Tris-HCl buffer (50 mM, pH 8.0) containing NaCl (100 mM) and glycerol (10%, v/v). Tris(2carboxyethyl)phosphine (TCEP, 0.3 mM) and ethylenediaminetetraacetic acid (EDTA, 0.5 mM) were included in the storage buffer for SlPatA GNAT cysteine variants. SlPatA GNAT concentration was determined by measuring absorbance at 280 nm. The molar extinction coefficient used to calculate SlPatA GNAT concentration was 17,420 M -1 cm -1 . The purification protocol for SlPatA GNAT for crystallography was similar to the one described above except that the purified, untagged protein was dialyzed into Tris-HCl buffer (10 mM, pH 8.0) and concentrated to 11 mg/ml before flash freezing into liquid nitrogen.
Selenomethionine-labeled SlPatA GNAT was overproduced as follows. Plasmid pSlPatA14 was transformed into strain JE9314 and 1-L culture of the resulting strain was grown overnight in M9 glucose medium. The culture was used to re-inoculate 2 x 2L of fresh M9 glucose medium at a 1:100 inoculum and the culture was grown to an OD 600 of ~1. The culture was cooled on ice to 16°C for 10 min, and a defined amino acid mixture containing lysine (100 mg/L), threonine (100 mg/l), phenylalanine (100 mg/L), leucine (50 mg/L), isoleucine (50 mg/L), valine (50 mg/L), and selenomethionine (50 mg/L) was added to suppress methionine biosynthesis. The culture was grown at 37°C for 30 min before the addition of IPTG (1 mM) to induce SlPatA GNAT expression. The culture was grown overnight at 37°C and the protein was purified and stored for crystallography as described above.
Purification of SeAcs WT and SeAcs CTD . Plasmids pACS38, pACS44, pACS56, pACS57, pACS58, pACS60, and pACS61 were transformed into a Δpka derivative of E. coli C41 (λDE3) (JE9314) to prevent acetylation prior to overproduction. The resulting strains were grown overnight and sub-cultured 1:100 (v/v) into 2 liters of lysogeny broth (LB) containing ampicillin (100 µg/ml). The cultures were grown shaking at 25 °C to A 600 ∼ 0.7 and protein synthesis was induced with IPTG (0.25 mM). Upon induction, the cultures were grown overnight at 25°C. SeAcs WT and SeAcs CTD proteins were purified and stored as described above for SlPatA GNAT with modifications. During the first purification step, the His 6 -SeAcs-bound resins were washed with buffer B [Tris-HCl buffer (50 mM, pH 8), NaCl (500 mM), and imidazole (20 mM)] before His 6 -SeAcs proteins were eluted with buffer C. His 6 -SeAcs CTD -bound resins were washed with buffer B3 [Tris-HCl buffer (50 mM, pH 8), NaCl (500 mM), and imidazole (40 mM)] before elution with buffer C. In the second purification step SeAcs WT proteins and SeAcs CTD proteins did not adsorb to the column and were present in the flow-through fractions. Proteins were stored in Tris-HCl buffer (50 mM, pH 8.0) containing NaCl (100 mM) and glycerol (10%, v/v). TCEP (0.3 mM) and EDTA (0.5 mM) were included in the storage buffer for SlPatA GNAT cysteine variants. The molar extinction coefficients used to calculate protein concentrations were 72,152 M -1 cm -1 for SeAcs WT and 15,470 M -1 cm -1 for SeAcs CTD . Two-hybrid system assay and β-galactosidase activity measurement. E. coli strain KDZif1ΔZ harboring compatible plasmids were grown overnight in 1 ml of nutrient broth (NB) supplemented with kanamycin and carbenicillin. The following day, strains were sub-cultured 1:100 into 200 ml of NB supplemented with kanamycin, carbenicillin, and isopropyl β-D-1thiogalactopyranoside (IPTG, 200 µM) to induce expression from two-hybrid screen plasmids. Strains were grown for 3.5 h at 37°C with shaking at medium intensity in a BioTek plate reader (BioTek Instruments, Inc.). Absorbance values of the cultures were measured in the microtiter plate at 650 nm wavelength using a Spectramax Plus 384 spectrophotometer (Molecular Devices).
SlPatA GNAT -SeAcs CTD complex crystals were grown by hanging drop diffusion where 1 µL of concentrated protein at 10.0 mg/mL in Tris-HCl buffer (10 mM, pH 8.0) containing CoA (2 mM) was mixed with 1 µL of 11.2% (w/v) polyethylene glycol 8000, 100 mM 1,4piperazinediethanesulfonic acid pH 6.5, and 120 mM Li 2 SO 4 . Crystals formed spontaneously after 2 months and grew to a maximum dimension of 400 µm x 50 µm x 50 µm. For freezing, crystals were transferred to paratone-N and then flash frozen in liquid nitrogen.
X-Ray data collection and structural refinement. X-Ray diffraction data for the SlPatA GNAT crystals and SlPatA GNAT -SeAcs CTD complex crystals were collected at the SBC 19-ID and SBC 19-BM beamlines respectively (Advanced Photon Source, Argonne, IL). The data sets were integrated and scaled with the program HKL3000(29). X-ray data collection statistics are given in Table 4. The structure of the SlPatA GNAT -SeAcs CTD complex was solved by molecular replacement using SlPatA GNAT and SeAcs CTD (PDB # 1PG4) as search models where residues in loops were deleted. The structure of SlPatA GNAT was solved by single-wavelength anomalous diffraction (SAD) using crystals containing selenomethionine protein. The HKL3000 suite was used to build an initial model of seleno-methionine SlPatA GNAT utilizing the programs SHELX, mlphare, DM, and ARP/wARP (30)(31)(32). This initial structure was used without any further refinement as a molecular replacement model to determine the structure of the native protein using the program Phaser (32,33). The native structure of SlPatA GNAT and that of the SlPatA GNAT -SeAcs CTD complex were refined by iterative cycles of manual model building in Coot and restrained refinement in Refmac 5.6 (34,35). Data processing and refinement statistics are presented in Table 4. The native SlPatA GNAT and SlPatA GNAT -SeAcs CTD complex structures have been deposited with the accession codes PDB # 4NXY and 4U5Y, respectively.

Results
SlPatA GNAT interacts with SeAcs CTD . To address the question of how GNAT specificity is achieved, the interaction between SlPatA GNAT with SeAcs CTD was investigated. Our attempts to crystallize SePat were not met with success, so the protein acetyltransferase from S. lividans (SlPatA GNAT ) was chosen. SlPatA GNAT proved amenable to structural studies and previous studies showed that SlPatA WT acetylates SeAcs WT in vitro (10), even though the domain organization of SePat WT and SlPatA WT is reversed (Fig. 1B). SlPatA GNAT was sufficient for functionality in vivo through its ability to substitute for SePat WT during growth on 10 mM acetate. As expected, expression of SlPatA GNAT inhibited growth of a S. enterica Δpat ΔcobB strain, but allowed growth of a S. enterica Δpat cobB + strain, which retained the ability to deacetylate acetyllysine (Figs. 1A, 1E).
Interaction model for SlPatA GNAT and SeAcs CTD . Initial attempts to co-crystallize SlPatA GNAT with SeAcs CTD (contains the target K609) were unsuccessful. This problem was overcome by introducing a covalent linkage between the SlPatA GNAT and SeAcs CTD domains at a position that neither affected the enzymatic activity of SlPatA GNAT nor the formation of the SlPatA GNAT /SeAcs CTD binary complex (described below). The location for the linkage was identified by first creating a computational model for the complex from the high-resolution structures of the individual domains.
The three-dimensional structure of SlPatA GNAT was determined to 1.5 Å (Fig. 1C, 1D, Table 4). The structure of SlPatA GNAT revealed a characteristic mixed α/β GNAT fold that contained the conserved acetyl-CoA binding site including the catalytic residue E123. SlPatA GNAT residues F126 and M168 overlapped with the modeled Ac-CoA structure, thus are likely to undergo a shift upon Ac-CoA binding, as observed for the analogous residues of the Ac-CoA bound structure of MtPatA (F238 and M280).
The most similar structure to SlPatA GNAT was that of the GNAT domain from Sulfolobus solfataricus Pat (PDB # 3F8K) with an RMSD of 1.32Å over 131 residues. The major secondary structure differences between the SsPat GNAT and SlPatA GNAT exist along the protein substrate-binding surface.
The structure of SlPatA GNAT determined here was combined with the previously reported structures of SeAcs WT (PDB # 1PG3, 1PG4) (13) to generate computer models of the interaction interface using the ClusPro 2.0 server (36)(37)(38)(39). The models were evaluated by requiring that the distance between the SlPatA GNAT catalytic residue E123 and the ε-amino group of the target lysine in the PX 4 GK (where K is SeAcs residue K609) motif be similar to that of 8 Å observed in the crystal structure of the Tetrahymena Gcn5 bound to the H3 peptide substrate (PDB # 1QSN). The best computational models placed residue E123 within 15Å of the α-carbon of the target lysine (K609 side chain was not resolved in the SeAcs WT structure) (40). The best model for SeAcs WT -SlPatA GNAT interactions is shown in figure 1F. Notably, SlPatA GNAT was predicted to interact predominantly with SeAcs CTD . Figure  2A shows the best model for the interaction between SlPatA GNAT domain and the SeAcs CTD (residues D518 to S652).
SlPatA GNAT -SeAcs CTD crosslinking reveals a selective orientation for enzyme-substrate interaction. The validity of the interaction model was tested by introducing a series of crosslinks between SlPatA GNAT and SeAcs CTD in nine orientations with linkers of distinct lengths, which included one orientation that approximated the ClusPro model shown in figure 2A. This strategy allowed us to examine the transfer of the acetyl moiety within the SlPatA GNAT and SeAcs CTD complexes as a function of cross-linker length and orientation. Residues were chosen at the periphery of the predicted interaction surfaces of SlPatA GNAT and SeAcs CTD to avoid substitutions that might otherwise disturb the putative interface. A single-cysteine variant was constructed for each chosen residue, which included residues S73, A110, A164 in SlPatA GNAT and residues A538, H567, D600 in SeAcs CTD (Figs. 2A, 2B). Purified single-Cys variants were cross-linked using sulfhydryl-specific cross-linkers with reported lengths of 2.05 Å to 19.9 Å (Fig. 2C). Crosslinking in all nine configurations was successful with at least four different crosslinkers for each combination (Fig. 2E, Coomassie Blue-stained images).
The ability of the SlPatA GNAT to acetylate SeAcs CTD in each cross-linked complex was used as a measure of the biochemical relevance of the orientation allowed by the cross-linker length constraints. In the absence of a cross-link, SlPatA GNAT acetylation of SeAcs CTD was efficient only when SlPatA GNAT was in 10-fold molar excess of the SeAcs CTD (Fig. 2D). As seen in figure 2E, SlPatA GNAT acetylated SeAcs CTD in nearly all orientations tested when the spacer length was ≥8Å. The single exception was the complex between variants SlPatA GNAT A110C-SeAcs CTD A538C, in which we did not observe acetylation for any of the crosslinked complexes tested. Notably, only complex SlPatA GNAT S73C-SeAcs CTD H567C resulted in detectable acetylation when a direct disulfide bond between the cysteine residues of each protein held the complex together. This direct disulfide bond severely restricted the interactions between SlPatA GNAT and SeAcs CTD , and yet the acetylation signal was strong (Fig. 2E). Thus, we hypothesized that the protein orientations in the SlPatA S73C -SeAcs H567C complex reflected the interactions of these proteins in vivo. Significantly, residues S73 of SlPatA GNAT and H567 of SeAcs CTD were positioned near one another in the ClusPro interaction model ( Fig.  2A).
Crystal structure of the SlPatA GNAT -SeAcs CTD complex at 1.9Å resolution. To visualize how variants SlPatA GNAT S73C and SeAcs CTD H567C interacted, we crystallized the SlPatA GNAT S73C-SeAcs CTD H567C complex formed with an 8-Å linker, and the structure was determined to 1.9Å (Fig. 3A-D, Table 4). In the of SlPatA GNAT S73C-SeAcs CTD H567C structure, the catalytic E123 residue of SlPatA GNAT was 4.7 Å away from the target K609 of SeAcs CTD (Fig. 3D,E), a distance within the range observed for Tetrahymena Gcn5 bound to a lysine-containing peptide substrate (40,41).
The interface between SlPatA GNAT and SeAcs CTD included more than just the interactions between the K609-containing SeAcs WT loop and the primary active site of SlPatA GNAT , for example, it included interactions that are well separated in the primary sequence. The interface shows good shape complementarity with a shape correlation statistic (Sc) of 0.54 that is similar to 0.60 observed in the Gcn5 H3 peptide complex (1QSN) (42). These values fall within the expected range for this type of complex where an antibody/antigen complex results in an Sc value from 0.64 to 0.68 and while an aberrant interface will result in an Sc of around 0.35.
The SlPatA GNAT -SeAcs CTD interaction surface was distinct and larger than that of the Gcn5-H3 peptide complex (Fig. 3B,C,F). The interface between SlPatA GNAT and SeAcs CTD buried a total surface area of 2150 Å 2 where this was 48% polar and 52% non-polar, which is typical for recognition surfaces. The distribution of hydrophobicity in the interface is roughly the same as across the total surface area of either protein. This is consistent with a transient interface in which specificity is driven by charge-charge interactions with minimal hydrophobic contributions. The size and disposition of the residues in the binding interface was consistent with the hypothesis that Pat substrate specificity involves elements outside the simple PX 4 GK loop motif (12). However, the PX 4 GK loop does play a structural role in positioning K609 into the active site cleft of SlPatA GNAT . The carbonyl oxygen of the preceding glycine residue hydrogen bonded to residue R64 of SlPatA GNAT and facilitated a bend in the backbone loop conformation (Φ 97°, ψ 9°). Also located within the PX 4 GK loop, the carbonyl oxygens of SeAcs CTD R606 and S607 hydrogen bond with the positively charged side chains of SlPatA GNAT residues R79 and R64, respectively (Fig. 3E). Similar to the Tetrahymena Gcn5 H3 complex, hydrophobic interactions were also involved in positioning the target lysine (40). SlPatA GNAT residue F66 packed against the methylene groups of SeAcs CTD residue K609 (Fig. 3E).
In addition to the interactions with the PX 4 GK motif there were complementary ionic interactions between the protein domains, where a large negatively charged surface patch on SlPatA GNAT interacted with a complementary positive patch on SeAcs CTD (Figs. 3B, 3C). A prominent group of arginine residues in SeAcs CTD (R612, R613, and R616) lay on the surface of a short α-helix that followed the PX 4 GK motif (Figs. 3B), where these are conserved in Acs homologues from bacteria, archaea, and eukaryotes (Fig. 4A) (43). These arginines interacted with a negative patch on SlPatA GNAT that included residues F66, E160, and E184 (Figs. 3C). These interactions most likely contribute to the specificity of the SlPatA GNAT domain for its substrate.
SlPatA GNAT homologs shown to acetylate the cognate Acs from the same organism exhibit amino acid sequence conservation at several of the residues noted above (Fig. 4B).
In vitro and in vivo evidence that amino acid charge reversals at the SlPatA GNAT -SeAcs CTD interface disrupt interactions. The protein:protein interactions observed in the crystal structure were tested with a bacterial two-hybrid assay in vivo by mutating charged surface residues (Fig. 5). Introduction of an opposing charge into the interacting surface of the SlPatA D185R or SeAcs CTD (e.g., SeAcs R606E , SeAcs R613D ) significantly reduced interactions of those proteins with SeAcs CTD or SlPatA GNAT , respectively. Conversely, as a control, substitution of a residue near, but outside of the interaction interface (SlPatA GNAT E121R) did not significantly affect the SlPatA GNAT -SeAcs CTD interaction, supporting the orientation of the domains in the X-ray structural model. Importantly, the bacterial two-hybrid system results were reproduced both in vitro and in vivo when non-truncated forms of SeAcs and SlPatA were used, and when the SlPatA homologue from S. enterica (SePat) was used (Fig. 6, Fig.  7). Full-length SeAcs WT variants SeAcs R606E and SeAcs R613D retained activity in vitro despite amino acid substitutions near the catalytic K609 [(SeAcs WT , 8.3 ± 0.5 µmol AMP min -1 mg -1 ; SeAcs R606E , 5.1 ± 0.2 µmol AMP min -1 mg -1 ; and SeAcs R613D , 2.8 ± 0.2 µmol AMP min -1 mg -1 ; mean ± S.D., n=9)], however SlPatA WT and its S. enterica homologue SePat WT acetylated these proteins less efficiently (Fig. 6). Amino acid substitutions near the active site lysine of AMPforming CoA ligases have been shown to affect activity (11).
We also demonstrated that variant SlPat GNAT E121R interacted with SeAcs WT in vivo and inhibited growth of a S. enterica Δpat ΔcobB strain, but not the growth of a S. enterica Δpat cobB + strain. In contrast, variant SlPat GNAT D185R only slightly inhibited growth of the S. enterica Δpat ΔcobB strain (Fig. 7). These data were consistent with the observation that the SlPat D185R variant exhibited significantly weaker interactions with SeAcs CTD in the bacterial-twohybrid assay (Fig. 5). When higher levels of variant SlPat GNAT D185R were present, growth of S. enterica on 10 mM acetate was inhibited (Fig. 7). These results were consistent with the idea that variant SlPat GNAT D185R and SeAcs CTD interactions were weakened but not abolished.

Discussion
Mapping the interaction surface between the SlPatA GNAT and the globular protein substrate SeAcs CTD is a significant advance in our understanding of how protein lysine acetyltransferases recognize-with specificitylarge globular protein targets. To generate an interaction model of a GNAT with a protein substrate, the crystal structure of SlPatA GNAT was first solved. The structure of SlPatA GNAT and SsPat WT , the most closely related structure, differed significantly at the predicted substrate interface. SsPat WT was crystallized after limited proteolysis, and the absence of amino acids (specifically residues 42-52 that link helix α2 to strand β2) may distort the active site cleft. SsPat WT acetylates the DNA binding protein Alba, but it is unknown whether it can acetylate AMP-forming CoA ligases. Thus, we cannot rule out the possibility that the structural differences surrounding the SsPat WT substratebinding cleft may represent distinct substrate specificity.
Interestingly, the factors that contribute to specificity observed for SlPatA are different from those reported for methylmalonyl-CoA synthetase of Rhodopseudomonas palustris. In that case, a loop spanning residues 447-450 in the R. palustris methylmalonyl-CoA synthetase (RpMatB) was identified as containing elements critical for binding to, and subsequent acetylation by the R. palustris protein acetlytransferase (RpPat) enzyme (12). The equivalent loop in SeAcs CTD (residues 565-568) is not involved in the SlPatA GNAT -SeAcs CTD interface, which highlights the complexity and difficulty in predicting GNAT substrate specificity from first principles.
The structure of the acetylation complex revealed complementary electrostatic interactions between SlPatA GNAT and SeAcs CTD . The charged residues involved in the SlPatA GNAT -SeAcs CTD interface are conserved in only some species (Fig. 4A,B). The positively charged SeAcs residues R612, R613, and R616 are seen in homologues in all domains of life, yet further examination of the exceptions may reveal differences in PatA-Acs interactions among species. Likewise, few of the PatA residues involved in the PatA-Acs interaction are conserved. Considering that the GNAT family of acetyltransferases is noted for its lack of primary sequence conservation (7), predicting interacting residues of GNATs in the absence of structural data remains challenging. The extensive interaction surface observed at the SlPatA GNAT -SeAcs CTD may be a key feature of PatA-Acs interactions. A large interaction surface would facilitate evolution of distinct constellations of interactions between each GNAT and its protein substrate(s). Continued structural analysis of GNAT-substrate complexes will reveal the range of interactions that occur between GNATs and their protein substrates.
The GNAT-substrate interface identified in the SlPatA GNAT -SeAcs CTD crystal structure is remarkably distinct from the Gcn5-H3 peptide complexes reported to date (Figs. 3C, 3E), and reveals structural roles of residues within and distant from the PX 4 GK motif found in substrates of Pat-type GNATs. The structure of the SlPatA GNAT -SeAcs CTD interaction will serve as a model to further identify, validate, and engineer (11) specific globular protein targets of GNAT protein acetyltransferases.  SeAcs CTD (phosphor images labeled "[ 14 C] Acetylation"). Samples were quenched after 60 min, separated by SDS-PAGE and stained with Coomassie Blue to visualize proteins (labeled "SDS-PAGE"). Full length SlPatA was incubated with SeAcs CTD at a molar ratio of 1:3 (SlPatA WT :SeAcs CTD ) for reference. Images of Coomassie Blue-stained gels and phosphor images were cropped to bands corresponding to the SeAcs CTD . Acetylation was quantified relative to the signal obtained with SlPatA plus SeAcs CTD , and is reported as the mean (n=3). S.D. was ≤18% of the mean value. (E) Transfer of the acetyl moiety from [1-14 C]-acetyl-CoA to the SeAcs CTD was tested (phosphor images) for each of the SlPatA GNAT -SeAcs CTD complexes. Images of Coomassie Blue-stained gels (labeled "SDS-PAGE") and phosphor images (labeled "[ 14 C] Acetylation") were cropped to bands corresponding to the SlPatA GNAT -SeAcs CTD heterodimers.   4. SlPatA GNAT and SeAcs CTD residues at the interaction interface are conserved. (A) Alignment of sequences in and around the Acs CTD PX 4 GK motif (black box) from S. enterica (SeAcs, accession # NP_463140), Saccharomyces cerevisiae (Acs2p, accession # NP_013254), Halobacterium salinarum (HsAcs, accession # WP_0109027), and S. lividans (SlAcs, accession # EFD68454). Blue shaded boxes indicate conserved positively-charged residues. " * " indicates a fully conserved residue; " : " indicates residues with high similarity; " . " indicates residues with low similarity. (B) Alignment of GNAT domain from homologs of SlPatA (accession # EFD66247) from S. enterica (SePat, accession # XNP_461586), R. palustris (RpPat, accession # NP_494576), and Mycobacterium tuberculosis (MtPat, accession # WP_003906490). Notation is described as above. Red and green shaded boxes indicate negatively charged and hydrophobic residues, respectively, observed at the SlPatA GNAT -SeAcs CTD interaction interface. Sequence alignment generated in ClustalW2 (45).  Acetylation"). (B) Reaction controls lacking SePat were incubated for 90 minutes and imaged as described above. Gels and phosphor images were cropped to the SeAcs bands and labeled as described in panel A. (C) Phosphor signal associated with each band in panels A and B was quantified as described in the experimental procedures. (D) SeAcs, SeAcs R606E , and SeAcs R613D were incubated with SlPatA (white bars) or SePat (gray bars) at the ratio described above in the presence or absence of acetyl-CoA. After 90 min, SeAcs activity was measured in an NADH-consumption assay. All data points are mean ± standard deviation (n=6). ) during growth on NCE minimal medium supplemented with acetate (10 mM). Growth experiments were performed at 37°C using a microtiter plate and a microtiter plate reader (Bio-Tek Instruments). All data points represent mean value. All standard deviations <0.015 absorbance units (n=4).  Where R work refers to the R factor for the data utilized in the refinement and R free refers to the R factor for 5% of the data that were excluded from the refinement.
by guest on July 8, 2020