Structural Insights into the Substrate Specificity of the Rhodopseudomonas palustris Protein Acetyltransferase RpPat

Background: RpPat acetylates many acyl-CoA synthetase enzymes. Results: RpPat does not acetylate RpMatB but can acetylate chimeric versions of it that differ from the wild-type enzyme in a loop region >20 Å away from the acetylated lysine. Conclusion: RpPat likely interacts with a large surface area of its substrates. Significance: RpPat substrates cannot be predicted by a short acetylation motif alone. Lysine acetylation is a post-translational modification that is important for the regulation of metabolism in both prokaryotes and eukaryotes. In bacteria, the best studied protein acetyltransferase is Pat. In the purple photosynthetic bacterium Rhodopseudomonas palustris, at least 10 AMP-forming acyl-CoA synthetase enzymes are acetylated by the Pat homologue RpPat. All bona fide RpPat substrates contain the conserved motif PX4GK. Here, we show that the presence of such a motif is necessary but not sufficient for recognition by RpPat. RpPat failed to acetylate the methylmalonyl-CoA synthetase of this bacterium (hereafter RpMatB) in vivo and in vitro, despite the homology of RpMatB to known RpPat substrates. We used RpMatB to identify structural determinants that are recognized by RpPat. To do this, we constructed a series of RpMatB chimeras that became substrates of RpPat. In such chimeras, a short region (11–25 residues) of RpMatB located >20 residues N-terminal to the acetylation site was replaced with the corresponding sequences from other AMP-forming acyl-CoA synthetases that were known RpPat substrates. Strikingly, the enzymatic activity of RpMatB chimeras was regulated by acetylation both in vitro and in vivo. Crystal structures of two of these chimeras showed that the major difference between them and wild-type RpMatB was within a loop region ∼23 Å from the acetylation site. On the basis of these results, we suggest that RpPat likely interacts with a relatively large surface of its substrates, in addition to the PX4GK motif, and that RpPat probably has relatively narrow substrate specificity.

Lysine acetylation is a post-translational modification that occurs in all three domains of life (1,2). In eukaryotes, histone tails are subject to many types of modifications, including lysine acetylation, and such modifications play an important role in gene expression and chromatin stability (3,4). Because histone acetylation has been extensively studied, the modifying enzymes are often referred to as histone acetyltransferases, despite the fact that they can acetylate non-histone proteins as well. There are three groups of histone acetyltransferases as follows: the Gcn5-related N-acetyltransferases, the MYST family, and the p300 family (5,6). Only Gcn5-related N-acetyltransferases are found in bacteria, and different family members acetylate a wide range of substrates containing primary amines, including small molecules, protein N-terminal amino groups, and lysine side chains (7).
The bacterial protein acetyltransferase Pat 4 was first discovered as a regulator of the AMP-forming acetyl-CoA synthetase (Acs) enzyme in Salmonella enterica (8). In Escherichia coli, the homologue of SePat is known as Pka or YfiQ (9). Acs (EC 6.2.1.1; pfam 00501) activates acetate to acetyl-CoA (10). In S. enterica, Acs activity is abolished when residue Lys-609 is acetylated. Acs acetylation is reversed by the action of a sirtuin-type NAD ϩ -dependent protein deacetylase called CobB (11,12). In other bacteria, Pat homologues have been shown to acetylate different acyl-CoA synthetases and to propionylate the enzyme propionyl-CoA synthetase (13)(14)(15). A feature shared by all these protein substrates is the acetylation site, which is located within a conserved PX 4 GK motif located near the C terminus of these proteins. At present, it is not known which components of this motif are important for recognition or whether this motif is sufficient for acetylation by Pat.
There are reports of protein substrates of Pat that are not acyl-CoA synthetases and do not contain the PX 4 GK motif. For example, the E. coli Pat homologue Pka acetylates RNase R, and acetylation affects the stability of the protein rather than its enzymatic activity (9,16). S. enterica Pat has been reported to acetylate a wide range of metabolic enzymes in addition to Acs, including glyceraldehyde-3-phosphate dehydrogenase (GapA), isocitrate dehydrogenase phosphatase/kinase (AceK), and isocitrate lyase (AceA) (17). We note that our group was unable to repeat these findings (14). Thus, whether Pat has broad or narrow substrate specificity remains an open question. Because it has been suggested that Pat may be a master regulator of bacterial metabolism (17), it is important to define the substrate range of Pat.
In this work, we investigated the substrate specificity of RpPat, the Pat homologue in the purple photosynthetic ␣-proteobacterium Rhodopseudomonas palustris. In this bacterium, RpPat acetylates 10 different acyl-CoA synthetases involved in the activation of short, medium, and long chain fatty acids and aromatic acids (13,14). To date, all known substrates of RpPat are acyl-CoA synthetases, and all contain the PX 4 GK motif at the site of acetylation. R. palustris has two protein deacetylases, RpSrtN and RpLdaA, both of which can deacetylate and thus reactivate acyl-CoA synthetases (13,14). We examined two acyl-CoA synthetases involved in the catabolism of dicarboxylic acids in R. palustris, namely pimeloyl-CoA synthetase (RpPimA) and methylmalonyl-CoA synthetase (RpMatB). For reasons explained below, we also studied benzoate:CoA ligase (BclM) from Burkholderia xenovorans LB400 (BxBclM).
RpPimA activates medium chain mono-and dicarboxylic acids (18), and RpMatB activates malonate (a C 3 dicarboxylate) and methyl malonate (a related C 4 dicarboxylate) (19). Here, we show that wild-type RpMatB is not a substrate of RpPat, despite the fact that RpMatB contains the catalytic lysine residue located within the PX 4 GK motif.
Using a combination of in vivo, in vitro, and structural approaches, we determined that in RpMatB the sequence of a structural loop in the C-terminal domain recognized by RpPat in its bona fide substrates is different. As a result of relatively minor changes in this sequence, RpPat does not modify RpMatB. In substrates of RpPat, the alluded region is surprisingly far away from the acetylation site, strongly suggesting that it may be difficult to identify RpPat substrates exclusively on the presence of the PX 4 GK motif. It is noteworthy that RpPat (and probably its homologues in other organisms) does not acetylate all AMP-forming acyl-CoA synthetases. These results and others reported elsewhere (14) continue to question the idea proposed by others that S. enterica Pat is not a very specific enzyme (17).

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-All strains and plasmids used in this study are listed in supplemental Tables S1 and S2. E. coli strains were grown at 37°C in lysogenic broth (LB, Difco) (18,19). B. xenovorans LB400 was obtained from the Agricultural Research Service culture collection, United States Department of Agriculture, and was cultured at 30°C on LB devoid of NaCl. All R. palustris strains were derivatives of R. palustris CGA009 (20) and were cultured at 30°C in photosynthetic medium (PM) (21) supplemented with succinate (10 mM), benzoate (3 mM), and NaHCO 3 (10 mM) or methyl malonate (10 mM) and vitamin B 12 (150 nM). When used, ampicillin was at 100 g ml Ϫ1 ; chloramphenicol was at 20 g ml Ϫ1 , and kanamycin was at 50 g ml Ϫ1 (E. coli) or 75 g ml Ϫ1 (R. palustris). Radiolabeled [1-14 C]acetyl-CoA (54 mCi/mmol) was purchased from Moravek, and all other chemicals were from Sigma. For growth curves, R. palustris starter cultures were grown to full density photosynthetically on PM containing succinate and kanamycin in Balch tubes (22) flushed with N 2 . Cultures were diluted 1:16 into fresh medium supplemented with the appropriate carbon source and kanamycin. Cultures were incubated at 30°C in light without shaking, and growth was monitored at 660 nm using a Spectronic Spec20D (Thermo). Three biological replicates of each strain were monitored, and each experiment was repeated twice.
Molecular Techniques-DNA manipulations were performed using standard techniques (23). Restriction endonucleases were purchased from Fermentas. DNA was amplified using PfuUltra II Fusion DNA polymerase (Stratagene), and site-directed mutagenesis was performed using the QuikChange kit (Stratagene). Plasmid DNA was purified using the Wizard Plus SV miniprep kit (Promega), and PCR products were purified using the Wizard SV gel and PCR clean-up kit (Promega). DNA sequencing reactions were resolved and analyzed at the University of Wisconsin-Madison Biotechnology Center. Oligonucleotide primer sequences are listed in supplemental Table S3. Plasmids were moved into R. palustris by electroporation.
Plasmid Construction-R. palustris pat, pimA, and matB were amplified from R. palustris genomic DNA using the primers listed in supplemental Tables S2 and S3. The PCR products were cut with the specified restriction enzymes and ligated into the overexpression vector pTEV5 cut with the same enzymes (24). The resulting plasmids directed the synthesis of proteins with an N-terminal hexahistidine (His 6 ) tag, which was removed using recombinant tobacco etch virus (rTEV) protease (25). B. xenovorans bclM (locus tag Bxe_A1419) was amplified from a single colony and cloned into pTEV5 using PIPE cloning (26). The chimeric genes were generated using overlapextension PCR (27) or site-directed mutagenesis and were subsequently cloned into pTEV5. A detailed description of chimera construction can be found in the supplemental material.
Protein Overexpression-The plasmid encoding RpPat (pRp-PAT13) was transformed into strain BL21/pLysS. The resulting strain was grown until early stationary phase and subcultured 1:100 in 3 liters of super broth supplemented with ampicillin (100 g ml Ϫ1 ) and chloramphenicol (10 g ml Ϫ1 ). Cultures were grown with shaking to an A 600 ϳ0.5, and protein expression was induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside. Cultures were grown overnight at 30°C; cells were harvested by centrifugation at 8000 ϫ g for 12 min in a Beckman Coulter Avanti J-20 XOI refrigerated centrifuge with a JLA-8.1000 rotor, and cell pellets were frozen at Ϫ80°C for later use. All other overexpression plasmids were transformed into strain JE9314, a pka-deficient derivative of E. coli C41(DE3) (28), to ensure that proteins were purified in their nonacetylated state. The resulting strains were grown until early stationary phase and subcultured 1:100 in 2 liters of lysogenic broth supplemented with ampicillin (100 g ml Ϫ1 ). Protein expression was induced with isopropyl 1-thio-␤-D-galactopyranoside as described above, and cell pellets were frozen at Ϫ80°C for later use.
Protein Purification-Protein concentrations were determined using a NanoDrop 1000 spectrophotometer (Thermo-Fisher) and A 280 molar extinction coefficients for each protein.
Extinction coefficients were obtained from the Integrated Microbial Genomes website (29) or calculated using the Prot-Param tool (30).
R. palustris Pat-RpPat was overproduced as an rTEV-cleavable His 6 N-terminal fusion protein. The cell pellet was resuspended in 30 ml of RpPat binding buffer (sodium phosphate buffer (50 mM, pH 7.5), NaCl (500 mM), imidazole (20 mM), tris(2-carboxyethyl)phosphine (TCEP, 0.5 mM)) and supplemented with lysozyme and deoxyribonuclease I (1 mg ml Ϫ1 each). Cells were lysed using a French pressure cell (two passes, 10 MPa). Debris was removed by centrifugation at 39,000 ϫ g for 30 min at 4°C, and the soluble fraction was passed through a 0.45-m filter. His 6 -RpPat was purified by nickel-affinity purification using an FPLC (ÄKTA) equipped with a 5-ml His-Trap FF column. After binding to the column, the column was washed with 6% RpPat elution buffer (sodium phosphate buffer (50 mM, pH 7.5), NaCl (500 mM), imidazole (500 mM), TCEP (0.5 mM)) and eluted in a linear gradient up to 100% RpPat elution buffer. Fractions containing RpPat were pooled and incubated with rTEV (1:10 rTEV to protein ratio) for 3 h at room temperature to cleave the tag. RpPat was dialyzed three times against RpPat binding buffer lacking imidazole; the first dialysis buffer contained EDTA (1 mM). The His 6 tag and His 7tagged rTEV protease were removed by passage over the His-Trap column again, using the same buffers as the first purification except that the binding buffer did not contain imidazole. Untagged RpPat eluted in the wash step containing 6% RpPat elution buffer. Fractions were pooled and dialyzed three times against RpPat storage buffer containing HEPES (50 mM, pH 7.5), glycerol (20%, v/v), and NaCl (500, 300, and 150 mM, successively), before drop freezing in liquid nitrogen and storage at Ϫ80°C.
Chimeras-The purification protocols for chimeras RpMatB-BxBclM B1 K491A and B3 K488A for crystallography were similar to the ones described above, except that the purified, untagged protein was dialyzed into HEPES buffer (50 mM, pH 7.5) before concentration to 10 -12 mg ml Ϫ1 using Ultracel 10,000 molecular weight cutoff centrifugal filters (Amicon Ultra). Accessible lysine side chains were reductively methylated using formaldehyde and dimethylamine borane complex as described previously (31). The reaction was quenched by addition of Tris-HCl buffer (pH 7.5) to a final concentration of 500 mM, and the protein was dialyzed overnight against 2 liters of Tris-HCl buffer (50 mM, pH 7.5) containing dithiothreitol (DTT, 2 mM), before dialysis into Tris-HCl buffer (10 mM, pH 7.5). The protein concentration was adjusted to 10 mg ml Ϫ1 before drop-freezing in liquid nitrogen.
In Vitro Protein Acetylation Assay-Proteins were acetylated using radiolabeled acetyl-CoA as described previously (8). . Reactions (25 l total volume) were incubated for 60 min at 30°C. Aliquots (10 l each) were resolved using SDS-PAGE and visualized by staining with Coomassie Blue R-250. Gels were dried and exposed overnight to a MultiPurpose Phosphor Screen (Packard Instrument Co.). Radioactivity associated with protein was quantified using a Typhoon FLA9000 Biomolecular Imager (GE Healthcare) and Opti-Quant version 04.00 software (Packard Instrument Co.).
X-Ray Data Collection and Structural Refinement-X-ray diffraction data for all crystals were collected at the SBC 19-ID beam line (Advanced Photon Source, Argonne, IL). The data sets were integrated and scaled with the program HKL2000 (33). X-ray data collection statistics are given in Table 1.
Structural Determination-The structures of both methylated RpMatB chimeras were solved by molecular replacement using the program Phaser (34, 35) and the methylated RpMatB K488A structure (PDB accession number 4FUT) (36) as the search model. This was followed by iterative cycles of manual model building in Coot (37) and TLS and restrained refinement in Refmac 5.6 and Phenix.refine: 1.6_289 (38,39). For both structures, TLS groups were defined such that each polypeptide chain within the asymmetric unit was divided into the N-terminal (residues 1-400) and C-terminal domains (residues 401-503 or 506). These two domains were naturally flexible during the catalytic cycle, and introduction of these TLS groups led to a 2% decrease in R free for both structures (

RESULTS
In R. palustris, MatB Is Not Regulated by Acetylation-We tested whether RpPat could acetylate the acyl-CoA synthetases RpMatB, RpPimA, and BxBclM. RpMatB is 39% identical to RpPimA over 376 residues and 28% identical to BxBclM over 485 residues, suggesting that all three enzymes share many structural features, although they do exhibit different substrate specificities. RpPimA, RpMatB, and BxBclM contain the conserved lysine residue and surrounding motif PX 4 GK found in other AMP-forming acyl-CoA synthetases that are acetylated, such as the R. palustris benzoate:CoA ligase (RpBadA, Fig. 1A). We purified each enzyme, and each was tested as a substrate for RpPat using 14 C-labeled acetyl-CoA as co-substrate (Fig. 1B). RpPat efficiently acetylated RpPimA and RpBclM, and when the conserved lysine residues (Lys-534 in RpPimA or Lys-520 in BxBclM) were changed to alanine, acetylation was no longer detectable, consistent with the expectation that this lysine was modified in both proteins. Surprisingly, RpPat did not acetylate RpMatB, despite its similarity to other acyl-CoA synthetases that are substrates of RpPat.
Growth of R. palustris on Methyl Malonate Is Not under Acetylation Control-We recently showed that matB is required for photoheterotrophic growth of R. palustris on methyl malonate as a carbon source (36). If RpMatB activity were regulated by acetylation in vivo, we would expect that a deacetylase-deficient strain (⌬srtN ⌬ldaA) would have a growth defect, as is observed during growth on benzoate (Fig.  1C). This growth defect on benzoate occurs because RpBadA (benzoate:CoA ligase) is acetylated and inactivated by RpPat, and the deacetylases RpSrtN and RpLdaA are not present to reactivate RpBadA. As shown in Fig. 1D, an R. palustris ⌬srtN ⌬ldaA strain (JE11616, supplemental Table S1) grew as well as the wild-type strain on methyl malonate, a result that was consistent with our in vitro data, which suggested that RpMatB was not a substrate of RpPat.
We further confirmed that RpPat did not acetylate RpMatB. To do this, we quantified RpMatB activity after incubation with RpPat and acetyl-CoA ( Table 2). Results of positive control  (41,42). These enzymes activate fatty acids in two steps, an adenylation halfreaction and a thioester-forming half-reaction; to accomplish this, the C-terminal domain rotates ϳ140°between two catalytic conformations (43,44). The catalytic lysine residue that is acetylated in members of this family of enzymes is located within the C-terminal domain ( Fig. 2A). When the enzyme is in the adenylation conformation, the catalytic lysine residue is buried in the active site in the cleft between the N-and C-terminal domains, although this lysine residue is surface-exposed when the enzyme rotates into the thioester-forming conformation. Thus, this lysine is likely only available for modification when the enzyme adopts the thioester-forming conformation.
To determine why RpPat could not acetylate RpMatB, we made a series of fusion proteins in which a portion of the C-terminal domain of RpMatB was replaced with the corresponding sequence of RpPimA (Fig. 2C). RpPat acetylated RpMatB-Rp-PimA chimeras that contained at least the last 62 residues of RpPimA (Fig. 2B, annotated as chimeras P1, P2, and P3). These chimeras included at least 43 RpPimA-derived residues upstream of the acetylated lysine, demonstrating that acetylation depended on more than just the motif immediately surrounding the acetylation site.
To determine whether this entire region upstream of the acetylated lysine was required to turn RpMatB into an RpPat substrate, we made a second set of chimeras in which an internal piece of RpMatB was replaced with the corresponding sequence from RpPimA (Fig. 2E). Strong acetylation was observed when residues 440 -465 of RpMatB were replaced with the corresponding region of RpPimA (Fig. 2D, chimera P9). Acetylation still occurred at the conserved lysine residue (Lys-489 in the P9 chimera) because substitution of an alanine at this position abolished acetylation (Fig. 2F). These results showed that the primary sequence surrounding Lys-488 in RpMatB (Fig. 2H) would not block the modification of Lys-488 by RpPat, but a region spanning 49 to 23 residues N-terminal to this lysine was critical for recognition by RpPat.
RpPat also acetylated a similar RpMatB-BxBclM chimera (Fig. 2, F and G), although the segment of RpMatB replaced by BxBclM had to be adjusted slightly to yield a stable protein. In this chimera, referred to as RpMatB-BxBclM B1, residues 444 -465 of RpMatB were replaced with the corresponding sequence of BxBclM (Fig. 2H). Substitution of the conserved active site lysine residue, Lys-491 in the RpMatB-BxBclM B1 chimera, abolished acetylation (Fig. 2F), demonstrating that acetylation occurred at the expected site.
RpMatB Chimeras Are Regulated by Acetylation in Vitro and in Vivo-Both the RpMatB-RpPimA P9 and RpMatB-BxBclM B1 chimeras retained enzymatic activity with methyl malonate, although the RpMatB-RpPimA P9 chimera was only ϳ40% as active as wild-type RpMatB, and the RpMatB-BxBclM B1 chimera was ϳ70% as active as wild-type RpMatB ( Table 2). The RpMatB-RpPimA P9 chimera was efficiently acetylated by Pat and lost ϳ99% of its activity upon acetylation, whereas the  RpMatB-BxBclM B1 chimera lost ϳ70% of its activity upon acetylation, under our assay conditions ( Table 2). The RpMatB-RpPimA P9 chimera, which retained 40% of the wild-type RpMatB activity, did not complement a ⌬matB strain growing photoheterotrophically on methyl malonate as a carbon source (Fig. 3). However, the same chimera did complement a ⌬matB ⌬pat strain (Fig. 3, circles), suggesting that in the ⌬matB pat ϩ strain the chimera was acetylated and rendered inactive. If this were true, we predicted that a ⌬matB strain co-expressing the genes encoding the chimera and the RpLdaA deacetylase would also grow, and this was the case (Fig. 3, open  diamonds). These results indicated that the RpMatB-RpPimA P9 chimera could be regulated by acetylation in vivo.
Three-dimensional Crystal Structure of RpMatB-RpBclM Chimera B1-Because acyl-CoA synthetases show little sequence conservation in the region between residues 440 and 465 (RpMatB numbering, Fig. 2H), it was unclear why this region was important for recognition or acetylation by RpPat. We hypothesized that there was a structural difference between wild-type RpMatB and its chimeras, and we attempted to crystallize both the RpMatB-RpPimA P9 and RpMatB-BxBclM B1 chimeras so that we could compare them with RpMatB and BxBclM, whose structures have been previously published (36,45). Attempts to crystallize either RpPimA or the RpMatB-Rp-PimA P9 chimera were not successful; therefore, only the RpMatB-BxBclM B1 chimera was pursued.
We obtained crystals of the RpMatB-BxBclM B1 chimera with ATP bound only after we chemically methylated the surface-exposed lysine residues. Methylation of lysine residues is known to affect packing within the crystal lattice without significantly changing the three-dimensional structure of the protein (46 -48). This method has been used to obtain crystals of recalcitrant proteins, including RpMatB (36, 49 -53). To avoid predictable problems stemming from chemical methylation of the active site lysine residue, this lysine (Lys-491) was changed to alanine.
The structure of the RpMatB-BxBclM B1 chimera was solved by molecular replacement, using RpMatB K488A with ATP bound as a search model. Three RpMatB-BxBclM B1 chimera monomers were found within the unit cell each having substantially the same structure, including that of the chimera region. Within the RpMatB-BxBclM B1 chimera structure, chain A compared with chain B and chain A compared with chain C have root mean square deviations (r.m.s.d.) of 0.46 Å over 3948 eq atoms and 0.50 Å over 3949 eq atoms, respectively. However, chain C does show some disorder in the C-terminal domain due to crystal packing. Fig. 4 shows an alignment of the C-terminal domains of the RpMatB-BxBclM B1 chimera, RpMatB K488A , and BxBclM. All three structures are very similar, having an r.m.s.d. of 0.46 Å over 501 eq ␣-carbons and 1.88 Å over 456 eq ␣-carbons when comparing RpMatB-BxBclM B1 chimera and BxBclM, respectively, with RpMatB K488A and differ significantly only within the chimeric region.
The BxBclM-derived residues in the RpMatB-BxBclM B1 chimera encompass the end of a ␤-strand, a connecting loop (referred to here as the chimera loop), and an ␣-helix that is located on the opposite side of the domain from the acetylated lysine residue and active site loop (Fig. 4). Electron density maps for this region are well defined as shown in Fig. 5. Although this chimeric region is very similar to the BxBclM structure from which it is derived (having an r.m.s.d. of 1.56 Å over 207 eq atoms), it is distinct from the corresponding region of RpMatB. The most notable difference is that there is one extra turn in the ␣-helix of BxBclM and the chimera compared with RpMatB (four turns instead of three), and the chimera loop has a different spatial orientation. Based on this structural comparison, it seemed possible that RpPat made crucial contacts with either the extra turn in the ␣-helix and/or the chimera loop.
We generated a sequence alignment of RpMatB and the RpMatB-BxBclM B1 chimera based on the structural alignment shown in Fig. 4, which included a three-amino acid gap in the RpMatB sequence where the extra turn in the ␣-helix would be (Fig. 6A). To identify which parts of this chimera were important for acetylation by RpPat, we first focused on the last three turns in the ␣-helix, which were structurally very similar in RpMatB and the chimera. We constructed chimera RpMatB-BxBclM B2, in which the last 11 residues in the chimeric region were changed back to the corresponding sequence in RpMatB (Fig. 6A). Chimera RpMatB-BxBclM B2 was acetylated as well as the RpMatB-BxBclM B1 chimera (Fig. 6, B and C; Table 2), suggesting that the amino acid side chains on the last three turns of the ␣-helix of the chimera were not important for acetylation. Next, we eliminated the three residues (Glu-Ile-Leu) that comprised the extra turn of the ␣-helix to generate chimera RpMatB-BxBclM B3 (Fig. 6A). The latter was also acetylated by RpPat (Fig. 6, B and C), albeit not as efficiently as the RpMatB-BxBclM B1 chimera (40% decrease in activity upon acetylation compared with a 71% decrease in activity for the RpMatB-BxBclM B1 chimera, Table 2). These data suggested that RpPat recognized proteins that have either a threeturn or a four-turn ␣-helix in this position.
Crystal Structure of RpMatB-RpBclM Chimera B3-We determined the crystal structure of the RpMatB-BxBclM B3 chimera to confirm that it did have a shortened ␣-helix. The RpMatB-BxBclM B3 chimera K488A variant was methylated prior to crystallization and crystallized with the same unit cell and space group as the previously reported ATP-RpMatB K488A (19). As expected, the RpMatB-BxBclM B3 chimera did have a shorter ␣-helix within the chimeric region, similar to that of wild-type RpMatB (Fig. 6D). Again electron density for this region was well defined as shown in Fig. 5. The chimera loop had a different conformation than both wild-type RpMatB and the RpMatB-BxBclM B1 chimera suggesting that it was not the FIGURE 3. Activity of the RpMatB-RpPimA P9 chimera is controlled by acetylation in vivo. Absorbance was monitored during photosynthetic growth of R. palustris on methyl malonate (10 mM). Data points are averages of four replicates, and error bars represent standard deviations. Squares, wildtype (CGA009) harboring plasmid pBBR1MCS-2 as a vector control; filled triangles, strain JE13568 (⌬matB/pBBR1MCS-2); filled diamonds, strain JE13912 (⌬matB/pRpMATB25 (encodes the RpMatB-RpPimA P9 chimera)); circles, strain JE14288 (⌬matB ⌬pat/pRpMATB25 (encodes the RpMatB-RpPimA P9 chimera)); empty diamonds, strain JE14957 (⌬matB/pRpMATB45 (encodes the RpMatB-RpPimA P9 chimera and ldaA ϩ ). backbone conformation of the loop but rather the disposition of amino acid side chains that is important for RpPat recognition.
The RpMatB-BxBclM B3 chimera has nine residues that differ from wild-type RpMatB. The first is a phenylalanine in place of valine at position 444, within the ␤-strand immediately preceding the chimera loop, which appears to be required for stability of the chimera. Two conserved valine residues and eight nonconserved residues follow this phenylalanine within the chimera loop that differs between RpMatB and the RpMatB-BxBclM B3 chimera (Fig. 6A). When the eight nonconserved residues were changed, two at a time, to the corresponding RpMatB sequence, the amount of acetylation decreased substantially (Fig. 6, B and C). The largest effect was seen when the first two pairs of amino acids were changed (chimeras RpMatB-BxBclM B4 and B5, changes in residues Leu-447 to Glu-450 of the chimera), which were each acetylated about 5-fold less than the RpMatB-BxBclM B1 chimera. These results showed that changes within the chimera loop strongly affected how well RpPat recognized the RpMatB-BxBclM chimeras.
Both the RpMatB-BxBclM B1 and B3 chimeras crystallized in the adenylation conformation, in which the acetylated lysine residue (Lys-488) is buried in the active site cleft between the Nand C-terminal domains. We have previously reported structures of RpMatB in the adenylation and apo conformations, and we generated a model of the thioester-forming conformation based on the structure of Streptomyces coelicolor MatB (36,54). Fig. 7 shows a stereo view of the B3 chimera modeled in all three conformations, with the Lys-488 side chain modeled from the MatB apo structure. It is likely that RpPat interacts with acyl-CoA synthetases when they are in the thioester-forming conformation, as the active site lysine residue is most accessible in this state; however, modification may also be possible in the apo conformation.
From the model of the RpMatB-BxBclM B3 chimera in the thioester-forming conformation (Fig. 8A), it seems plausible that RpPat would interact with both Lys-488 in the active site loop and the tip of the chimera loop, because both loops protrude from the same face of the C-terminal domain. This information also suggested that RpPat recognized a large section of RpMatB as the two most extreme points of these loops (the hydroxyl of Ser-454 and methylene group of Met-486) were 35 Å apart within the crystal structure.  Interestingly, the RpMatB-BxBclM B3 chimera has another pronounced positively charged region consisting of Lys-448 and Arg-449 within the chimera loop, whereas the wild-type RpMatB structure is predominantly electronegative throughout the chimera loop region. These surface representations show substantial differences in both surface shape and charge in the chimera loop that may explain how RpPat can recognize the RpMatB-BxBclM B3 chimera but not wild-type RpMatB.

DISCUSSION
A Loop ϳ20 Å Away from the Acetylation Site Is Critical for RpPat Recognition of Its Protein Substrate-Previous studies have shown that RpPat acetylates at least 10 different acyl-CoA synthetases in R. palustris, all of which contain a conserved PX 4 GK motif within the active site loop (13,14). Notably, this motif is not sufficient for acetylation. For example, RpMatB is not a substrate of RpPat even though it shares extensive homology with other acyl-CoA synthetases and does contain the PX 4 GK motif preceding the catalytic lysine residue. The key difference in RpMatB appears to be the shape and electrostatic potential of a loop region located ϳ20 Å away from the conserved active site lysine residue and Ͼ20 amino acids away in the protein sequence. Such a structure appears crucial for acetylation (Figs. 2 and 6). This loop region found in bona fide substrates of RpPat can replace the corresponding sequence of RpMatB to yield catalytically active RpMatB variants that are now subject to acetylation control (Figs. 2 and 3).
At present, the three-dimensional structure of RpPat or its homologues has not been reported. Hence, we can only speculate how RpPat may interact with its substrates. Results reported herein suggest that RpPat interacts with a large surface area of its substrates (Fig. 8). It seems possible that Pat simultaneously interacts with the active site lysine residue and the loop region identified by this work and that RpPat may require these two positively charged loop regions for substrate recognition. These results demonstrate that RpPat, and perhaps other Pat homologues, shows a great deal of specificity among acyl-CoA synthetases. These results further support recently reported data (14) that challenged suggestions by others that Pat homologues in E. coli and S. enterica acetylate a wide range of structurally diverse substrates (17,55).
Insights into the Physiological Role of RpPat-RpMatB appears to function in methyl malonate metabolism in R. palustris, where the ultimate product of methyl malonate catabolism is likely the tricarboxylic acid cycle intermediate succinyl-CoA (36). Many, if not most, of the AMP-forming acyl-CoA synthetases that are regulated by RpPat either activate acetate or propionate directly to their corresponding acyl-CoA derivative or initiate the breakdown of longer chain fatty acids or aromatic acids to acetyl-CoA. It is plausible that RpPat is a sensor of acetyl-CoA or propionyl-CoA concentrations in the cell and uses these compounds to turn off the acyl-CoA synthetases at the beginning of these catabolic pathways. If this were the case, then we would hypothesize that RpPat only modifies acyl-CoA synthetases involved in acetyl-CoA or propionyl-CoA generation. One example of an acyl-CoA synthetase not involved in acetyl-or propionyl-CoA generation in R. palustris is RpMatB, which we have shown is not acetylated. Similarly, acyl-CoA synthetases involved in biosynthesis of natural products such as siderophores and antibiotics would not be regulated by RpPat. This hypothesis awaits further experimental work.
In R. palustris MatB May Have Evolved to Evade Inactivation by RpPat-RpMatB is the first example to date of an acyl-CoA synthetase that is not regulated by RpPat in R. palustris. Changes in a relatively nonconserved loop region distant from the acetylation site greatly affect is susceptibility to acetylation, The arrows indicate the direction and magnitude of rotation of the N-terminal domain relative to the C-terminal domain during the transition from one conformation to the next. In all cases, the C-terminal domain is from the RpMatB-BxBclM B3 chimera structure, and the N-terminal domains are either from wild-type MatB (apo and thioesterification, PDB accession code 4FUQ) or MatB K488A (adenylation, PDB accession code 4FUT). In each case the N-and C-terminal domains are aligned to RpMatB structures (apo, 4FUQ; adenylation, 4FUT) or to the S. coelicolor MatB structure (thioesterification, 3NYQ (54)). Figure was prepared with PyMOL (60).
suggesting that RpMatB evolved to escape regulation by RpPat. A variant of S. enterica acetyl-CoA synthetase that is not susceptible to acetylation has also been generated by changing one residue near the C terminus of the enzyme from leucine to proline (56). Together, these findings suggest that there may be many ways for these enzymes to evolve such that they retain their enzymatic activity but cannot be modified by Pat. In addition, some microbes do not have Pat homologues, suggesting that they either have alternative acetyltransferases that have the same role (such as Bacillus subtilis (57), Mycobacterium tuberculosis (58), and Sulfolobus solfataricus (59)) or that these microbes do not regulate acyl-CoA synthetases by acetylation.
Closing Remarks-From a practical perspective, our results indicate that it is difficult to predict, on the basis of primary sequence alone, whether or not an AMP-forming acyl-CoA synthetase is regulated by acetylation. RpMatB is the first example of an acyl-CoA synthetase that contains the PX 4 GK motif within the active site loop that is not acetylated by RpPat, although there may be many others. To make the computational identification of RpPat substrates more difficult, there appears to be little sequence conservation within the "chimera loop" identified herein as needed for acetylation. Thus, the newly discovered region does not provide needed homology to facilitate the identification of acyl-CoA synthetases that are under acetylation control. As shown by this work, more research is needed to define the true extent of the bacterial lysine acetylome and to establish how many proteins are true substrates of RpPat and its homologues in other bacteria. Evidence is mounting that RpPat specializes in acetylating a subset of acyl-CoA synthetases in R. palus-tris, and it will be interesting to see if this holds true for Pat homologues in other organisms.