Kanglemycin A Can Overcome Rifamycin Resistance Caused by ADP-Ribosylation by Arr Protein

ABSTRACT Rifamycins, such as rifampicin (Rif), are potent inhibitors of bacterial RNA polymerase (RNAP) and are widely used antibiotics. Rifamycin resistance is usually associated with mutations in RNAP that preclude rifamycin binding. However, some bacteria have a type of ADP-ribosyl transferases, Arr, which ADP-ribosylate rifamycin molecules, thus inactivating their antimicrobial activity. Here, we directly show that ADP-ribosylation abolishes inhibition of transcription by rifampicin, the most widely used rifamycin antibiotic. We also show that a natural rifamycin, kanglemycin A (KglA), which has a unique sugar moiety at the ansa chain close to the Arr modification site, does not bind to Arr from Mycobacterium smegmatis and thus is not susceptible to inactivation. We, found, however, that kanglemycin A can still be ADP-ribosylated by the Arr of an emerging pathogen, Mycobacterium abscessus. Interestingly, the only part of Arr that exhibits no homology between the species is the part that sterically clashes with the sugar moiety of kanglemycin A in M. smegmatis Arr. This suggests that M. abscessus has encountered KglA or rifamycin with a similar sugar modification in the course of evolution. The results show that KglA could be an effective antimicrobial against some of the Arr-encoding bacteria.

mucosal infections in humans. Because of resistance to most classes of antibiotics, including rifamycins, infections caused by M. abscessus remain extremely difficult to treat (9). The genome of M. abscessus encodes a homologue of Arr Ms (Arr Mab ) that confers innate highlevel Rif resistance. Deletion of Arr Mab significantly increases not only the sensitivity of M. abscessus to Rif but also that to other rifamycins, i.e., rifabutin, rifaximin, and rifapentine (7,10,11). Therefore, rifamycins that could evade the activity of Arr Mab could be promising treatment candidates. The C-25-modified rifamycins have, similarly to Rif, high MICs against wild-type M. abscessus but considerably lower MICs against a mutant lacking Arr Mab (11), suggesting that the putative Arr Mab may have a different substrate specificity to Arr Ms .
The natural product kanglemycin A (KglA) is an ansamycin antibiotic that, like other rifamycins, inhibits transcription by binding within the rifampicin-binding pocket on the b-subunit of RNAP (12). However, KglA has unique substituents present on the ansa bridge, namely, a 2,2-dimethylsuccinic acid chain at C-20 and a b-O-3,4-O,O9 methylene digitoxose moiety at C-27. When bound to RNAP, these substituents afford additional binding contacts in the rifampicin-binding pocket and produce an altered binding conformation that leads to overcoming rifampicin-resistant amino acid substitutions (12).
We hypothesized that these large, bulky substituents may also prevent KglA binding to Arr or that the ADP-ribosylated KglA may be able to bind to RNAP due to its different mode of binding compared to that of Rif. In this study, we show that M. smegmatis Arr indeed cannot bind and modify KglA. In contrast, a homologue of Arr Ms from M. abscessus, which we show is indeed a functional Rif ADP-ribosyl transferase, binds and modifies KglA with the same efficiency as that of Rif. We also, for the first time, directly show that ADP-ribosylated Rif does not inhibit bacterial RNAP.

RESULTS
M. smegmatis Arr (Arr Ms ) ADP-ribosylates Rif but not KglA. In order to characterize Arr Ms in vitro, we cloned the Arr gene from M. smegmatis into a pET28 expression vector and expressed and purified the protein from Escherichia coli. To investigate the inactivation of antibiotics by Arr Ms , we performed in vitro inactivation reactions with a tandem disk assay ( Fig. 2A). Increasing concentrations of Arr Ms were incubated with NAD 1 and antibiotic and then spotted onto paper disks which were placed onto a lawn of Staphylococcus aureus RN4220. Apparent decreases in the sizes of the zones of inhibition were interpreted as functional inactivation of the antibiotic. As shown in Fig. 2A, the control antibiotic carbenicillin is not inactivated by Arr Ms . Rif is inactivated by Arr Ms , resulting in decreased sizes of zones of inhibition when Arr Ms concentration is increased. However, KglA appears to be resistant to inactivation by Arr Ms , as the size of the zone of inhibition does not decrease even at very high concentrations of Arr Ms . The results of the disc assay were corroborated by measurement of MIC for the drugs treated with Arr Ms or left untreated ( Table 1). Note that the smaller zones of inhibition and higher MICs seen with KglA, compared to those with Rif, is thought to be due to poorer penetration of the compound through the cell envelope. These data indicate that KglA is not a substrate for Arr Ms .
To directly analyze ADP-ribosylation by Arr Ms , we utilized a tandem liquid chromatography-mass spectrometry (LC-MS)-based assay in which antibiotic and NAD 1 substrates were incubated in the absence or presence of Arr Ms and then the reaction products were separated by high-performance liquid chromatography (HPLC) and identified by mass spectrometry. When Rif is incubated with NAD 1 in the absence of Arr Ms , Rif and its oxidized form rifampicin quinone (Rifq) are resolved with no identifiable ADP-ribosylated product (Fig. 2B). This indicates that the ADP ribosylation reaction requires enzymatic catalysis. Indeed, when incubated in the presence of NAD 1 and Arr Ms , both Rif and Rifq are ADPribosylated, confirming the activity of purified Arr Ms (Fig. 2B). Interestingly, however, incubation of KglA with NAD 1 and Arr Ms failed to ADP-ribosylate KglA, indicating that KglA is not a substrate of Arr Ms (Fig. 2C), which is consistent with results of the above-described in vivo assay.
M. abscessus Arr (Arr Mab ) ADP-ribosylates both Rif and KglA. Prior genetic experiments have indicated the substrate specificity of a putative ADP-ribosyl transferase from M. abscessus, Arr Mab , may differ from that of Arr Ms (11). Therefore, we aimed to determine the function and activity of Arr Mab in vitro. First, a purified Arr Mab was analyzed in the disk assay, as described above. As can be seen from Fig. 3A (middle row), Rif activity is almost completely eliminated by both concentrations of Arr Mab , as judged by disappearance of the zone of inhibition when Rif is incubated with Arr Mab . Unexpectedly, unlike Arr Ms , Arr Mab was also capable of inactivating KglA (Fig. 3A, bottom row). Measurement of MIC confirmed this observation (Table 1).
To resolve this apparent controversy, we directly analyzed activity of Arr Mab using a tandem LC-MS-based assay, as we did for Arr Ms . Incubation of Rif and Rifq with NAD 1 and Arr Mab resulted in their ADP-ribosylation (Fig. 3B). However, unlike in the case with Arr Ms , incubation of KglA with Arr Mab and NAD 1 also resulted in ADP-ribosylation of the antibiotic (the exact mass of the product was 1,523.40 Da, the predicted mass of ADPribosyl KglA; Fig. 3C). This finding corroborates the results of the above-described disk assay. The results also confirm that the homologue of Arr Ms from M. abscessus indeed encodes a functional rifamycin ADP-ribosyl transferase that has broader substrate specificity than that of Arr Ms .
Rif and KglA binding affinities at Arr Mab and Arr Ms . The inability of Arr Ms to ADPribosylate KglA can be explained by either inability to bind the KglA molecule or inability to perform the catalysis. To distinguish between these possibilities, we investigated the binding of Rif and KglA to Arr Ms and Arr Mab using microscale thermophoresis (MST). Apparent dissociation constants (K d ) were determined by titrating serial dilutions of antibiotics against Arr Ms and Arr Mab . The affinity of Arr Ms to KglA was much lower than its affinity to Rif ( Fig. 4A and B), suggesting that the bulky 2,2-dimethylsuccinic acid chain at C-20 and/or b-O-3,4-O,O9 methylene digitoxose at C-27 of KglA prevents binding of the compound to the Arr Ms active site. In contrast, Arr Mab bound both antibiotics with the same K d . The result indicates that the inactivity of Arr Ms toward KglA is explained by the reduced binding affinity to the antibiotic. Notably, Rif binds with greater affinity to Arr Ms than to Arr Mab (Fig. 4A), suggesting that the broader substrate specificity of Arr Mab comes at a cost of reduced binding affinity.
ADP-ribosylation renders rifamycins inactive against RNA polymerase. It is not known how ADP-ribosylation of rifamycins by Arr enzymes inactivates the antibiotics. Therefore, we purified ADP-ribosyl-Rif and ADP-ribosyl-KglA from the reactions catalyzed by Arr Ms and Arr Mab , respectively, and analyzed their effects on in vitro transcription by E. coli RNAP on a linear DNA template containing the T7A1 promoter. As can be seen in Fig. 5, consistently with our previous study (12), both Rif and KglA efficiently inhibited transcription with submicromolar 50% inhibitory concentrations (IC 50 ). In contrast, ADP-ribosyl-Rif and ADP-ribosyl-KglA were inactive even at high concentrations (100 mM). Because part of the mechanism of inhibition by all rifamycins is the sterical occlusion of the pathway of the nascent RNA (13), this result suggests that ADP-ribosylation prevents rifamycins binding to RNAP.

DISCUSSION
Rifamycins inhibit bacterial transcription by targeting the b-subunit of RNAP (13). One of the findings of this work is that ADP-ribosylation of Rif by Arr proteins completely abolishes its activity against RNAP in vitro. Addition of the ADP-ribosyl at C-23 abolishes a critical hydrogen bond between the Rif molecule and the RNAP and also orientates the bulky substituent toward the surface of the Rif-binding pocket on RNAP, likely causing a severe steric clash.
Importantly, however, we show that KglA, a rifamycin with unique bulky substituent present on the ansa bridge, is not a substrate for Arr Ms , as a consequence of reduced binding affinity to the enzyme. If KglA adopts a similar conformation to that of Rif within the Rif-binding pocket of Arr Ms , a strong steric clash occurs between the C-27 b-O-3,4-O,O9 methylene digitoxose moiety of KglA and an a-helix (residues 54 to 65) that constitutes one half of the Rif-binding cleft of Arr Ms (Fig. 6A). This a-helix, termed a1, is implicated in Rif-binding interactions; residue D55 makes polar interactions with O-11 of Rif, while residues A56, W59, G60, and L63 form nonpolar interactions with the carbon backbone of the ansa bridge (3). Arr homologues are widely distributed across various pathogenic and environmental bacteria (E. coli, Pseudomonas aeruginosa, Klebsiella pneumoniae, Acinetobacter baumannii, Stenotrophomonas maltophilia, Burkholderia cenocepacia, many anaerobic bacteria, and different actinomycetes species, such as Streptomyces coelicolor). Considering that KglA also retains activity against Rif-resistant RNAP and bacteria (12,14), the antibiotic appears to be a promising lead compound with which to target some (but not all) bacteria possessing Arr enzymes.
We also characterized a homologue of Arr Ms encoded by the M. abscessus genome and confirmed that it is a true ADP-ribosylase of Rif. Like Arr Ms , Arr Mab utilizes NAD 1 to catalyze the ADP-ribosylation of Rif. However, unlike Arr Ms , KglA remains a substrate for   ADP-ribosylation by Arr Mab . The amino acid sequences between Arr Ms and Arr Mab are very similar, exhibiting a 66% sequence identity and 76% homology (Fig. 6B). However, Arr Mab a1 bears almost no sequence similarity with that of Arr Ms . This lack of homology in a1 may underlie the broader substrate specificity exhibited by Arr Mab . However, the variation of a1 residues involved in Rif binding may also explain the reduced affinity with which Arr Mab binds Rif (Fig. 4).
There is strong genomic evidence that most innate resistance mechanisms seen in clinical pathogens are of environmental origin (15). The often-communal existence of bacteria generates pressure to gain a competitive edge over neighboring microorganisms. Consequently, bacteria have evolved a number of mechanisms to counter antibiotics secreted by other microbes in their surroundings. Bulky ansamycin compounds, such as KglA, are produced by actinomycetes that occupy the same natural niches as nontuberculous mycobacteria, such as M. abscessus (14,16). It is tempting to speculate that in its evolution, M. abscessus has encountered KglA or a similar natural rifamycin with a modified ansa bridge.  It should be noted that the precise physiological role of Arr enzymes remains unclear, as Arr has been shown to play roles in cellular pathways even in the absence of rifamycins. For example, Arr is upregulated in response to different stresses (e.g., double-stranded DNA breaks, starvation, oxidative stress, and ciprofloxacin treatment) (17). In M. smegmatis, Arr contributes to biofilm formation and confers a growth fitness advantage. Although its expression has been shown to be associated with the stringent response, the mechanism of stress-correlated gene induction is unclear (18). It is possible that the main function of Arr is still inactivation of rifamycins, but the presence of rifamycins causes similar cell stresses (such as double-strand breaks caused by collisions of RNAP with replication machinery) that signal for production of Arr.

MATERIALS AND METHODS
Reagents, antibiotics, and DNA templates. All chemicals, reagents, and antibiotics were purchased from Sigma unless otherwise stated. The T7A1 promoter fragment was produced by PCR with the primers 59-CGACGTTGTAAAACGACGGCCAGTG-39 and 59-GGTCGACTCTAGAGGATCGCT-39 (IDT) from the template GGTCGACTCTAGAGGATCGCTATAACAGGCCTGCTGGTAATCGCAGGCCTTTTTATTTGGATCCAGATCCCGAAAA-TTTATCAAAAAGAGTATTGACTTAAAGTCTAACCTATAGGATACTTACAGCCATCGAGAGGGACACGGCGAATAGCC-ATCCCAATCGACACCGGGGTCCGGGATCTGGATCTGGATCGCTAATAACAGGCCTGCTGGTAATCGCAGGCCTTTTT-ATTTGGATCCCCGGGTACCGAGCTCGAATTCACTGGCCGTCGTTTTACAACGTCG and was purified by agarose gel electrophoresis. Kanglemycin A was purified as described previously (12). Protein expression and purification. E. coli RNAP core and s 70 were purified exactly as described previously (19). M. smegmatis and M. abscessus Arr were expressed in T7 express cells (New England Biolabs) transformed with pET28 expression vector encoding N-terminal 6ÂHis-tagged M. smegmatis Arr or M. abscessus Arr. Expression was induced with 0.4 mM isopropyl-b-D-thiogalactopyranoside (IPTG) in exponentially growing cells, which were then incubated overnight at room temperature on an orbital shaker (150 rpm). Cells were then harvested by centrifugation and resuspended in grinding buffer (50 mm Tris-HCl [pH 7.9], 10% glycerol, 200 mM NaCl, and protease inhibitor mixture [Roche]). Cells were then lysed by sonication and debris cleared by centrifugation. Arr enzymes were then purified by HisTrap HP (Cytiva) nickel affinity chromatography, concentrated, and dialyzed into storage buffer (50 mM Tris-HCl [pH 7.9], 50% glycerol, 200 mM NaCl, and 2 mM b-mercaptoethanol).
In vitro rifampicin ADP-ribosyl transferase activity assay. Reactions were performed in a 100-ml final volume of Arr buffer (20 mM Tris-HCl [pH 7.9], 40 mM KCl, and 0.5 mM MgCl 2 ). M. smegmatis or M. abscessus Arr (10 mM) was mixed with Rif or KglA (100 mM) in 80 ml of Arr buffer at 37°C for 10 min. A 20-ml aliquot of NAD 1 in Arr buffer was added (250 mM final) and incubated for 1 h at 37°C. The reaction was quenched with 500 ml of methanol (MeOH). Methanol was then evaporated under negative pressure and the reaction mixture analyzed by LC-MS. All analytical separations were performed on an Agilent 1260 HPLC instrument by injection of 1 to 5 ml of sample onto a Raptor ARC-18 column (150 mm by 2.1 mm) (Restek) or an Ultra C 4 column (150 mm by 2.1 mm) operated at 0.2 ml/min and then eluted using a 30-min linear gradient from 5% to 100% of acetonitrile. The mobile phase was supplemented with 0.1% formic acid. Mass spectra were recorded in positive-ion mode on a Bruker MicrOTOF II time-of-flight mass spectrometer.
Rifampicin ADP-ribosyl transferase disk assay. Reactions were performed in a 10-ml final volume of Arr buffer. M. smegmatis or M. abscessus Arr at an indicated concentration (0 mM, 20 mM, or 200 mM) was mixed with antibiotic (1 mg/ml final) in 8 ml Arr buffer at 37°C for 5 min. A 2-ml aliquot of NAD 1 in water was added (10 mM final) and incubated for 1 h at 37°C. The reaction was quenched with an equal volume of methanol, the mixture spotted onto paper disks, and a disk assay performed with an embedded lawn of S. aureus RN4220 as described previously (20).
Rifampicin ADP-ribosyl transferase MIC assay. MICs were quantified by broth microdilution assay using the S. aureus RN4220 strain at a starting cell density of 2 Â 10 5 to 5 Â 10 5 CFU/ml and tryptic soy broth (Oxoid). The compounds were dissolved in dimethyl sulfoxide (DMSO). Cultures were incubated for 24 h at 37°C, and MIC was recorded as the lowest concentration of each drug leading to inhibition of visible growth. Reproducibility was ensured by repeating all tests at least twice on separate occasions.
Purification of ADP-ribosyl rifampicin. The reaction was performed in a 2,000-ml volume of Arr buffer containing 25 mM M. smegmatis Arr, 5 mg rifampicin, and 20 mM NAD 1 . The reaction mixture was incubated at 37°C for 24 h and cleaned up on a 25 ml HyperSep C 8 solid-phase extraction (SPE) cartridge. ADP-ribosylated Rif was eluted with 30% MeOH and dried under negative pressure to yield 4.8 mg of ADP-ribosyl-Rif. Sample homogeneity was confirmed by tandem LC-MS, as described above.
In vitro transcription. Transcription from promoter DNA fragments was performed essentially as described previously (12,19). Reactions were performed in 10 ml of transcription buffer TB (20 mM Tris HCl [pH 7.9], 40 mM KCl, and 10 mM MgCl 2 ) containing 1 pmol of coli RNAP core with 3 pmols of s 70 and 10% DMSO with or without inhibitors. Transcription was initiated by the addition of a mixture of 25 mM CpA dinucleotide as a primer, 0.2 ml a-[ 32 P] UTP (10 mCi/ml; Hartmann Analytic), 10 mM UTP, 100 mM ATP, 100 mM CTP, and 100 mM GTP, and 10 nM promoter DNA. Reactions were stopped after 10-min incubation at 37°C by the addition of equal volume of formamide-containing loading buffer. Products were resolved in denaturing polyacrylamide gels, revealed by phosphorimaging (Cytiva), and analyzed using ImageQuant software (Cytiva).
Microscale thermophoresis. Binding affinity experiments were carried out on a Monolith NT.115 series instrument (Nano Temper Technologies GMBH). M. smegmatis and M. abscessus Arr were labeled with Monolith Protein labeling kit RED-NHS second-generation amine (Nano Temper Technologies GMBH) according to the manufacturer's guidelines. Roughly 5 ml of sample in MST buffer (20 mM HEPES [pH 7.9], 40 mM KCl, and 10 mM MgCl 2 ) were loaded into Monolith NT.115 Premium capillaries, and thermophoresis was measured for 30 s. Analysis was performed with Monolith software. K d was quantified by analyzing the change in normalized fluorescence (F norm ; fluorescence after thermophoresis/initial fluorescence) as a function of inhibitor concentration. Curves for K d data were fitted to a 4-parameter logistic equation using nonlinear regression in SigmaPlot software.