ApmA Is a Unique Aminoglycoside Antibiotic Acetyltransferase That Inactivates Apramycin

Apramycin is an aminoglycoside antibiotic that has been traditionally used in veterinary medicine. Recently, it has become an attractive candidate to repurpose in the fight against multidrug-resistant pathogens prioritized by the World Health Organization.

IMPORTANCE Apramycin is an aminoglycoside antibiotic that has been traditionally used in veterinary medicine. Recently, it has become an attractive candidate to repurpose in the fight against multidrug-resistant pathogens prioritized by the World Health Organization. Its atypical structure circumvents most of the clinically relevant mechanisms of resistance that impact this class of antibiotics. Prior to repurposing apramycin, it is important to understand the resistance mechanisms that could be a liability. Our study characterizes the most recently identified apramycin resistance element, apmA. We show ApmA does not belong to the protein families typically associated with aminoglycoside resistance and is responsible for modifying a different site on the molecule. The data presented will be critical in the development of apramycin derivatives that will evade apmA in the event it becomes prevalent in the clinic.
KEYWORDS aminoglycoside-modifying enzymes, antibiotic resistance, apramycin W aksman's tandem discoveries of streptomycin and neomycin over 65 years ago ushered in the clinical use of the aminoglycoside antibiotics (AGs) for the treatment of bacterial infections (1,2). Since then, a variety of AGs have found clinical success. All AGs have a six-membered aminocyclitol core that serves to distinguish subfamilies of the class. For example, the 4,6-deoxstreptamine antibiotics tobramycin, gentamicin, and amikacin are particularly effective against Gram-negative pathogens. These antibiotics offer improved oto-and nephrotoxicity profiles over the 4,5-deoxstreptamine-containing scaffolds such as neomycin (3,4). Most AGs act through noncovalent Citation Bordeleau E, Stogios PJ, Evdokimova E, Koteva K, Savchenko A, Wright GD. 2021. ApmA is a unique aminoglycoside antibiotic acetyltransferase that inactivates apramycin. mBio 12:e02705-20. https://doi.org/10.1128/ mBio. 02705-20. binding to the small ribosomal subunit in a fashion that disrupts the proofreading property of translation (5). The result is subversion of the genetic code followed by the production of aberrant proteins and peptides, resulting in cell death.
Soon after the introduction of AGs in clinical practice, acquired resistance mediated by mobile genetic elements was reported (6). Over the past decades, a plethora of AG resistance genes have been identified, many of them moving through bacterial populations by lateral gene transfer (7). In addition to the upregulation of efflux systems and mutations in outer membrane porins occurring in some bacteria such as Pseudomonas aeruginosa (8), two general mechanisms of resistance dominate in pathogens, drug inactivation and target modification. Both mechanisms result in a decreased affinity of the AG for its ribosomal target (9,10). Inactivation of AGs occurs via chemical modification by one of the following three classes of aminoglycoside-modifying enzymes (AMEs): O-phosphorylation (APHs), O-adenylylation (ANTs), or N-acetylation by aminoglycoside acetyltransferases (AACs) (11)(12)(13). Modification of the drug target is the most recent form of aminoglycoside resistance to emerge in the clinic (14). The 16S rRNA methyltransferases (RMTases) are responsible for the N7 methylation of G1405 (e.g., Escherichia coli numbering; ArmA, Rmt family) or N1 methylation of A1408 (14,15) (e.g., NpmA, KamB) within the 16S rRNA, respectively, conferring resistance to only 4,6-disubstituted or all 4,5-and 4,6-disubstituted AGs. RMTases are increasingly found in carbapenem-resistant isolates, greatly limiting therapeutic options for infections caused by these bacteria (16).
The broad mechanistic and genetic diversity of AG resistance impacts the use of existing drugs and dampens enthusiasm in the discovery of new members of this family of antibiotics, despite their highly desirable bactericidal activity toward Gram-positive and Gram-negative pathogens. In response to this challenge, the group at Achaogen embarked on an effort to deliver a next-generation semisynthetic AG that was not susceptible to common AMEs. The result of this effort is plazomicin (Zemdri) approved by the FDA in 2018, which retains antibiotic activity in the presence of most AMEs (17,18). However, plazomicin remains ineffective against isolates coexpressing many RMTases, resistance genes that were not in significant circulation when the development program was launched (19). This fact emphasizes the potential for rapid dissemination of new resistance elements in the clinic that may move more rapidly through bacterial populations than the drug development process.
Apramycin is an unusual AG where the deoxystreptamine (DOS) aminocyclitol ring is monosubstituted and is linked to an octadiose element ( Fig. 1) (20). Apramycin has been used in veterinary medicine for decades but more recently has been found to exhibit broad activity against WHO-prioritized multidrug-resistant (MDR) pathogens such as carbapenemase-producing Enterobacteriaceae and Acinetobacter baumannii (3,(21)(22)(23)(24). The unique monosubstitution of apramycin's DOS ring prevents both inactivation by a majority of common AMEs and resistance due to target alteration by N7 RMTases (24). This characteristic makes apramycin particularly attractive as a candidate next-generation AG for clinical use in humans (25).
Prior to the introduction of this antibiotic into the clinic, knowledge of the apramycin resistome is important. The N-acetyltransferase AAC(3)-IVa, a common selectable antibiotic resistance marker for molecular biology studies in actinomycetes, confers high-level apramycin resistance and is occasionally found in clinical isolates of Enterobacteriaceae (26)(27)(28). AAC (1) has been reported to be associated with apramycin resistance; however, the sequence of this gene is unavailable and the original isolates lost (A. Lovering, personal communication) (29,30). ApmA is another acetyltransferase linked to apramycin resistance. ApmA was first reported in 2011 from bovine and porcine methicillin-resistant Staphylococcus aureus (MRSA) isolates (31,32). The gene was found as the sole resistance element on a smaller plasmid as well as a larger antimicrobial multiresistance plasmid that also contained heavy metal resistance genes and potential virulence elements (32,33). AAC(3)-IVa adopts the structural fold for the more recently studied AAC(3) enzymes belonging the Antibiotic_NAT family (PDB ID 6MN4). Primary sequence alignment suggests that ApmA does not share this three-dimensional structure. Given the growing interest in apramycin as a drug candidate, we have investigated ApmA's activity toward apramycin and determined its three-dimensional structure. We identify ApmA as a member of the left-handed b-helix (LbH) superfamily, similar to chloramphenicol and streptogramin O-acetyltransferase resistance enzymes. This is the first report of an AG resistance element with this protein fold and reflects the diversity and enzymatic opportunism of antibiotic resistance.

RESULTS
The apramycin resistome is limited to four known genes. The resistome of apramycin was evaluated through susceptibility testing against our in-house antibiotic resistance platform consisting of a panel of isogenic strains of Escherichia coli BW25113 DtolC DbamB, each expressing unique aminoglycoside resistance elements (34). Apramycin susceptibility was surveyed against 27 aminoglycoside resistance elements, consisting of 11 AACs, 11 APHs, two ANTs, and four RMTases. Gene expression levels were under the control of the constitutive promoter P bla . A control strain not expressing an aminoglycoside resistance element was used as a reference for apramycin potency. We found only the four previously reported apramycin resistance elements to confer resistance (Fig. S1 in the supplemental material). The two N1-A1408-directed RMTases, NpmA and KamB, and the two acetyltransferases, AAC(3)-IVa and ApmA, each confer a high level of resistance to apramycin ($64 mg/ml; Table 1). ApmA remains the only member of these apramycin resistance elements uncharacterized with respect to its structure and function.
ApmA acetylates apramycin at 29-NH 2 . Purified recombinant ApmA was used to produce acetylated apramycin in vitro to determine the regiospecificity of acetyl group transfer. Characterization of the acetylated product was carried out using high-resolution electrospray ionization-mass spectrometry (HR-ESI-MS), which revealed a single acetylation of the apramycin core (mass increase of 42.0 Da) ( Table 2).
Further characterization of the regiospecificity of acetyl transfer was accomplished using a combination of one-and two-dimensional nuclear magnetic resonance spectroscopy (1D and 2D NMR) of purified, ApmA-inactivated apramycin. (Table S1; Fig. S4 to S8) Acetyl-apramycin NMR spectra were compared with those reported in the literature for the apramycin-free base (35). We noted significant deshielding of the 29 proton from 3.02 ppm in apramycin to 4.06 ppm in the acetylated product in the 1 H NMR. This change is a result of the effect of a carbonyl group attached to a neighboring atom (N29). In the 13 C NMR, two new carbon shifts were observed at 21.8 ppm and at 172.89 ppm, corresponding to a methyl group carbon and carbonyl carbon chemical shifts. Lastly, heteronuclear multiple bond correlation spectra (HMBC) revealed correlations between the carbonyl carbon and methyl protons, as well as their correlation with the 29 proton (Fig. 2). These data are consistent with acetylation of apramycin by ApmA at the 29-NH 2 of ring I.
ApmA is an N-acetyltransferase from the left-handed b-helix protein superfamily. Primary sequence analysis of ApmA identified a seven-hexapeptide repeat motif, suggesting it is a member of the LbH superfamily of acetyltransferases that includes the xenobiotic acetyltransferases (XAT) subclass of LbHs, responsible for the inactivation of streptogramin group A antibiotics (36, 37) (Vat) and chloramphenicol (38, 39) (CATB). Crystals of the apoenzyme and ApmA in complex with apramycin or acetyl-CoA were obtained and their structures solved to resolutions of 2.08 Å, 1.85 Å, and 2.30 Å, respectively (Table S2). Analysis of these complexes showed ApmA to be a trimeric protein and confirmed it to be a member of the LbH superfamily.
The overall structure of ApmA is consistent with the canonical XAT architecture (36): it consists of a C-terminal region comprised of three a-helices (residues 232 to 274), a central LbH domain where the hexapeptide motif is repeated seven times (residues 83 to 231), and an insert located in the center of the LbH fold, characterized by two a-helices (residues 132 to 176). Unlike other XATs, ApmA contains an additional Nterminal region, comprised of four b-sheets and two a-helices (residues 1 to 82) ( Fig. 3A and B). The LbH domains of three neighboring chains together form a tunnel to shuttle the pantothenate arm of acetyl-CoA into the apramycin binding pocket (Fig. 4A). The N-terminal region appears to play a role in distorting the first two b-strands from the 7-stranded b-sheet of the LbH domain relative to their position in VatA and XAT (PDB ID 2XAT) (Fig. S2). These b-strands are twisted nearly 90°from their position in VatA and XAT and significantly alter the shape and volume of the substrate binding pocket (Fig. S2). Three acetyl-acceptor binding pockets were identified with the positioning of the central DOS ring consistent with other AMEs, which typically are lined with aspartate and glutamate residues (40). In the ApmA-apramycin complex, the N3 atom of the  AG's aminocyclitol ring is coordinated by two aspartic acids (Asp144 and Asp145) in a helix within the C-terminal domain and one glutamic acid (Glu85) position the N1 atom. Ring III of apramycin interacts with several residues from one of the helices of the inserted loop region of the LbH domain (Gln147, Gln159, and Tyr172). The 69-OH of ring I interacts with N113 from within the LbH domain ( Fig. 4A and 4B). Consistent with the NMR structure of 29-acetyl-apramycin, we observed the 29-NH 2 of the antibiotic positioned for acetylation. Upon superimposition of the two binary complexes (root mean square deviation [RMSD], 0.29 Å), the 29-NH 2 lies between 3.1 to 3.7 Å from the carbonyl carbon of the acetyl moiety (Fig. 4B). Tyr102 participates in a network of hydrogen bonds between the 29-NH 2 and Asp144 that may be important for antibiotic binding. The imidazole side chain from His135 is within 2.8 to 3.1 Å of apramycin's 29-NH 2 . Consistent with other XAT enzymes, the N1 of this imidazole is hydrogen bonded to a carbonyl of the peptide backbone (Val142) (Fig. 4B). It is suspected that interaction increases the basicity of imidazole's N2 to help in a role as a general base (37,38). Superimposition of ApmA with VatA, in complex with their antibiotic substrates, demonstrates that the His135 of ApmA is geometrically equivalent to the catalytic histidine for VatA (His87) (36) (Fig. 5). Tyr102 of ApmA was also found in the same position of Tyr42 of VatA, a residue responsible for stabilizing the oxyanion intermediate for O-acetylation (36). To further assess the significance of His135 at the sequence level, we gathered reference sequences for Vats and CATBs from the Comprehensive Antibiotic Resistance Database (CARD) (41) to construct a multiple sequence alignment. Our analysis revealed that His135 of ApmA aligns with the conserved catalytic histidine found across all Vat and CATB sequences, essential for the O-acetylation of their respective substrates (36) (Fig. S3). We next generated the alanine mutant of His135 to evaluate the impact of this substitution on ApmA's detoxification of apramycin through cell-based assays. Upon expression of the His135Ala mutant in E. coli, resistance to apramycin remained within 2fold of when the wild-type enzyme was expressed (32 to 64mg/ml). These results suggest the role of this histidine in the acetylation of apramycin does not hold the same significance as for the function of other XATs and requires further investigation.

DISCUSSION
Apramycin's atypical structure in comparison with other AGs has garnered considerable attention for its potential as a next-generation AG antibiotic. The development of derivatives, termed apralogs, has focused on retaining apramycin's low ototoxic potential while increasing potency coupled with evading resistance to apramycin through AAC(3)-IV-mediated modification (42,43). Molecular modeling of 3-acetyl-apramycin bound to the ribosome indicates that reduced binding to the target results from a steric clash between the 16S rRNA phosphate backbone and the amide positioned at C3 of apramycin (24). We show that ApmA is an acetyltransferase that instead modifies apramycin at the N29 position of the octadiose element to confer high-level drug resistance. However, N29 does not make direct contacts with the 16S rRNA and could spatially accommodate the acetyl group. We suggest that acetylation at this position reduces the affinity of the overall molecule for its target. The N29 participates in an intramolecular interaction with O5 of the 2-DOS ring. Intramolecular interactions between AG sugar rings have been suggested to play an important role in the recognition and binding of AGs to the 16S rRNA (44). The octadiose element of apramycin participates in important hydrogen bonds with A1408 of the 16S rRNA, creating a glycoside-adenine pseudo-base pair (43,45,46). Acetylation of N29 would disrupt the orientation of the 2-DOS and octadiose ring. The carbonyl would also have the potential to create new intramolecular interactions that alter the configuration of the molecule. Unfavorable intermolecular interactions would also be introduced if the carbonyl group approaches the negatively charged phosphate backbone, creating an electronic clash. Lastly, the overall charge of apramycin will be impacted by the replacement of the primary amine with an amide. The removal of the positively charged amine would further hinder its ability to interact with the negatively charged RNA (Fig. 6).
The regiospecificity of the acetyl transfer phenotypically assigns ApmA to the AAC (29) family of AMEs. However, unlike ApmA, AAC (29) enzymes are members of the GNAT superfamily (19) and are found chromosomally encoded in Providencia stuartii and Mycobacterium tuberculosis (47,48). Our analysis of AG resistance shows that AAC (29)-Ia from P. stuartii does not confer resistance toward apramycin, making ApmA the first AAC(29) enzyme documented to do so. Furthermore, initial reports of apmA found the sequence mobilized on plasmids from Staphylococci isolates from bovine and swine (31)(32)(33). The most recent report of the gene has also identified apmA in Campylobacter isolated from pork, demonstrating an expansion of host and crossover to Gram-negative pathogens (49).
The findings of this study highlight the adaptation of a common protein fold to generate new functions in antibiotic resistance. Concerns that other XAT enzymes capable of conferring resistance to other classes of antibiotics were discussed over 20 years ago (50). Antibiotic acetyltransferases are products of convergence of function, with CoA-dependent acetylation spanning several protein scaffolds (39,(51)(52)(53)(54)(55). Our crystallographic data demonstrate ApmA belongs to the LbH superfamily of acetyltransferases, which had not previously been linked to the detoxification of AG antibiotics. The substrate specificity of the LbH scaffold has expanded to accommodate AGs, reminiscent of other sugar-containing substrates of LbH N-acetyltransferases involved in O-antigen biosynthesis (56,57). Previously, XATs were observed to only O-acetylate and their respective antibiotic substrates with the help of a conserved catalytic histidine. There are also FIG 6 Impact of 29 acetylation on ribosomal binding of apramycin. Crystal structure of apramycin bound to the ribosome (PDB ID 4AQY). Important interactions between N3 and the ribosome for recognition of the 2-DOS ring are indicated. Intramolecular interaction between N29 and O5 is highlighted. Acetylation at position N3 creates steric clash with the phosphate backbone. Acetylation of the N29 position is predicted to create an electronic clash with the negatively charged backbone of A1492 and G1491.
LbH N-acetyltransferases that contain a histidine in the active site and have been shown to be important in the acetylation of their respective nucleotide-linked sugar substrates (56,58). We show that the His135Ala ApmA mutant is still capable of N-acetylating apramycin to confer an equivalent level of resistance to when the wild-type protein is expressed. This suggests that the molecular mechanism of acetyl transfer is similar to that of the GCN5 family AAC(69) enzymes where no catalytic base is necessary. Here, acyl transfer occurs due to the alignment and proximity of acetyl-CoA and the nucleophilic receptor amine of the antibiotic (59).
Overall, this work has the potential to aid in the identification of apralogs with reduced susceptibility to ApmA. Furthermore, the distribution of apmA should be monitored for further transfer into clinical isolates. This unique AME threatens not only the success of apramycin's introduction into the clinic but may impact other AGs susceptible to its modification as well.

MATERIALS AND METHODS
Bacterial strains and apmA cloning. apmA (GenBank accession no. FN806789.3) was synthesized as a gBlock by Integrated DNA Technologies (IDT) for cloning into pGDP3 (34) and pET19b-TEV using NdeI and XhoI restriction sites. The pET19b-TEV plasmid consists of an N-terminal His 10 tag cleavable by a tobacco etch virus protease (TEV). The pGDP3 construct of apmA was transformed into hyperpermeable, efflux-deficient mutant E. coli BW25113 DtolC DbamB (for antimicrobial susceptibility testing). Construct of pET19b-TEV:apmA was transformed into BL21(DE3)-Gold competent cells (for crystallography). All constructs were verified through Sanger sequencing at the Mobix sequencing facility, McMaster University.
The final construct was verified through Sanger sequencing at the Mobix sequencing facility, McMaster University.
Antimicrobial susceptibility testing. Screening against our in-house antibiotic resistance platform was carried out as previously described (34). Apramycin susceptibility testing of E. coli BW25113 DtolC DbamB expressing apmA and apmA(H135A) was completed in triplicate following the Clinical and Laboratory Standards Institute (CLSI) protocols for the broth microdilution method (61). Strains were cultured in cation-adjusted Mueller-Hinton broth (CAMHB) in a 96-well format. Plates were incubated in a shaking incubator at 37°C for 18 h.
Protein expression and purification. For acetylated product characterization, E. coli BL21(DE3) Rosetta-gami pLysS transformed with pET19b-TEV constructs of apmA were grown in autoinduction medium supplemented with selection antibiotics for 3 days at 25°C and 180 rpm. Cells were collected by centrifugation at 6,400 Â g, 4°C, and resuspended in binding buffer (25 mM HEPES [pH 7.5], 300 mM NaCl, and 10 mM imidazole). Resuspended cells were lysed using a continuous cell disrupter at 20,000 lb/in 2 , 4°C, followed by centrifugation at 30,000 Â g to remove cell debris. ApmA was purified from the lysate by nickelnitrilotriacetic acid (Ni-NTA) affinity chromatography (Qiagen) at 4°C. A 2-ml volume of Ni-NTA resin was preequilibrated with binding buffer and incubated with the lysate for 1 h on ice prior to purification. The resin was washed with wash buffer (25 mM HEPES [pH 7.5], 300 mM NaCl, and 25 mM imidazole), and proteins were eluted with a 4-stepwise gradient (25%, 50%, 75%, and 100%) of elution buffer (25 mM HEPES [pH 7.5], 300 mM NaCl, and 250 mM imidazole). Elutions were dialyzed overnight against dialysis buffer (25 mM HEPES [pH 7.5], 300 mM NaCl). SDS-PAGE gel electrophoresis was performed to assess sample purity. To prepare stock solutions, concentrated ApmA was diluted to a final concentration ranging from 30 mM to 150mM in dialysis buffer. Protein dilutions were flash frozen in liquid nitrogen and stored at 280°C.
For crystallography studies, native ApmA was expressed in Gold competent E. coli BL21(DE3). A 3-ml overnight culture was diluted into 1 liter of LB media containing selection antibiotics and grown at 37°C with shaking. Expression of selenomethionine-derivatized ApmA was carried out in M9 minimal media following the manufacturer's instructions (Shanghai Medicilon). Expression was induced with isopropyl b-D-1-thiogalactopyranoside (IPTG) at 17°C when the optical density at 600 nm (OD 600 ) reached 0.6 to 0.8. The overnight cell culture was then collected by centrifugation at 7,000 Â g. Cells were resuspended in binding buffer (100 mM HEPES [pH 7.5], 500 mM NaCl, 5 mM imidazole, and 5% glycerol [vol/ vol]) and lysed with a sonicator, and cell debris was removed by centrifugation at 30,000 Â g. The cell lysate was loaded on a 4-ml Ni-NTA column (Qiagen) preequilibrated with 250 ml of binding buffer. The resin was washed with wash buffer (100 mM HEPES [pH 7.5], 500 mM NaCl, 30 mM imidazole, and 5% glycerol [vol/vol]), and the proteins were eluted with elution buffer (100 mM HEPES [pH 7.5], 500 mM NaCl, 250 mM imidazole, and 5% glycerol [vol/vol]). The His 10 -tagged proteins were then subjected to overnight TEV cleavage using 50 mg of TEV protease per mg of His 10 -tagged protein in binding buffer and dialyzed overnight against the binding buffer. The His 10 -tag and TEV were removed by running the protein again over the Ni-NTA column. The tag-free proteins were dialyzed against dialysis buffer (50 mM HEPES [pH 7.5], 500 mM NaCl) overnight, and the purity of the protein was analyzed by SDSpolyacrylamide gel electrophoresis. Spectroscopic characterization of ApmA-catalyzed acetylated apramycin. ApmA-catalyzed acetylated apramycin was produced from 50-ml in vitro reactions (50 mM HEPES, pH 7.5) consisting of 500 mM aminoglycoside, 500 mM acetyl-CoA, and 1 mM ApmA. Reaction mixtures were incubated at room temperature until acetylated products (mass increase of 42.0 Da) were detected by liquid chromatography (LC)/ESI-MS. Enzymes were removed by centrifugation using an Amicon Ultra-15 centrifugal filter and the flowthrough subsequently concentrated. Acetyl-apramycin was purified from the concentrate using AG50W-X8 strong cation resin. The resin was preequilibrated with 1% NH 4 OH and washed with H 2 O until a neutral pH was obtained. Fractions containing acetylated products were identified by LC/ESI-MS, followed by detailed analysis with NMR and HR-ESI-MS. LC/ESI-MS data were acquired using a QTrap 2000 (Applied Biosystems) system equipped with an Agilent 1100 LC interface. HR-ESI-MS data were acquired using an Agilent 1290 ultraperformance liquid chromatography (UPLC) separation module and quadrupole time of flight (Q-TOF) G6550A mass detector in positive ion mode. NMR analysis was completed using a Bruker AVIII 700 MHz instrument in deuterated water as the solvent. The chemical shifts are reported in parts per million.
Crystallization and structure determination. All crystals were grown at room temperature using the vapor diffusion sitting drop method with 0.5 ml of protein solution mixed with 0.5 ml of reservoir solution. Crystals were grown using the following reservoir solutions: ApmA plus apramycin complex (0.2 M CaCl, 20% polyethylene glycol 3350 [PEG 3350], and 5 mM apramycin) and ApmA plus acetyl-CoA complex (0.1 M citric acid, pH 3.6, 30% PEG 200, 5 mM apramycin, and 2 mM acetyl-CoA). All crystals were cryoprotected with Paratone-N or ethylene glycol and then flash frozen in liquid nitrogen prior to diffraction data collection. Diffraction data were collected at the Advanced Photon Source, Argonne National Laboratory, beamlines 19-ID or 21-ID. All data were processed by HKL-3000. The ApmA plus apramycin structure was solved using the single anomalous diffraction (SAD) method, and this was used to solve all additional complexes by MR. Structure refinement was performed using Phenix.refine plus manual building with Coot. The presence of substrate molecules was identified by building into the Fo-Fc difference density after the initial rounds of refinement.
Sequence and structural analysis. PyMOL (62) was used to identify potential interacting residues (#4.0 Å in distance) of ApmA with substrates apramycin and acetyl-CoA. Structural superimpositions with VatA (PDB ID 4HUS) were constructed using the cealign function in PyMol. XAT representative sequences were obtained from the CARD (41). ApmA and Vat sequences were aligned with the Expresso (63-65) function of T-Coffee (66,67) to build a profile hidden Markov model (HMMER 3.3.1) (68). All sequences were aligned with the resulting HMM profile and visualized using Jalview (69).
Data availability. PDB validation reports for crystal structures obtained in this study have been submitted with the manuscript. Accession numbers are as follows: 7JM0 (ApmA apoenzyme), 7JM1 (ApmA complex with acetyl-CoA), and 7J6M2 (ApmA complex with apramycin).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.