As previously described (Zhang et al., 2018), cultures of M. smegmatis strains were grown in Middlebrook 7H9 medium (Becton Dickinson) supplemented with 0.5% (v/v) glycerol, 0.05% (v/v) Tween 80 (Sigma) and 10% ADS enrichment (5% (w/v) bovine serum albumin fraction V, 2% (w/v) dextrose and 8.1% (w/v) NaCl). Single bacterial colonies of M. smegmatis mc²155 strain were cultured on Middlebrook 7H10 medium (Becton Dickinson) containing 10% ADS, and 0.5% (v/v) glycerol. Kanamycin, erythromycin and clarithromycin were purchased from Sigma.

To overexpress MSMEG_1954 in mycobacteria, the full-length coding gene of msmeg_1954 was amplified from M. smegmatis genomic DNA and cloned into the mycobacteria overexpression vector pMV261 (Stover et al., 1991) to yield pMV261-1954 for transformation into M. smegmatis (identified as pMV261-1954-mc²155). For a control, the empty pMV261 vector was also transformed into M. smegmatis and referred to as pMV261-mc²155.

The inhibition curves under erythromycin treatment were determined in 50 ml 7H9 medium culture of recombinant M. smegmatis strains in 100 ml flasks, by monitoring OD₆₀₀, as described previously (Li et al., 2015). Briefly, logarithmic phase M. smegmatis cultures (OD₆₀₀ of 0.5) of tested strains were diluted to an OD₆₀₀ of 0.1 and then allowed to grow in the presence or absence of 3 mg/L erythromycin at 37°C with shaking at 200 rpm. Cultures were monitored by measurement of OD₆₀₀. The growth of recombinant M. smegmatis in the presence of 0.75 mg/L clarithromycin was measured after 24 h incubation. All antibiotic resistant tests were repeated at least three times. Statistical analyses were performed using t-tests. *** p < 0.001, ** p < 0.01, and *p < 0.05.

Survival under erythromycin treatment was determined as indicated. Logarithmic phase cultures (OD600 ∼ 0.3) were treated with 25 mg/L erythromycin for 3 h, serially diluted (1:10) and spotted (3 μL) onto 7H10 medium. Photographs were taken after 3 days of incubation at 37°C. Experiments were repeated at least three times.

Logarithmic phase cultures (OD₆₀₀ of 0.8) of M. smegmatis strains were treated with different concentrations of erythromycin for 30 min and collected for RNA isolation, while untreated stains were set up as control groups. The corresponding cell pellets were resuspended into TRIzol (Invitrogen, United States) and RNA was purified according to the manufacturer’s instructions. SuperScript^TM III First-Strand Synthesis System (Invitrogen, United States) was used for cDNA synthesized. Quantitative real-time PCR (qRT-PCR) was performed using 2 × SYBR real-time PCR pre-mix (Takara Biotechnology Inc., Japan) on a Bio-Rad iCycler. M. smegmatis rpoD (the coding sequencing of RNA polymerase sigma factor SigA) was used to normalize the gene expression of msmeg_1954. The relative gene expression was calculated using the 2^–ΔΔCT method (Livak and Schmittgen, 2001). Experiments were repeated at least three times. Levels of mRNA expression of msmeg_1954 under clarithromycin treatment was carried out in the logarithmic phase cultures (OD₆₀₀ of 1.8) of M. smegmatis as described above.

The gene encoding the full length msmeg_1954 was amplified by polymerase chain reaction (PCR) from M. smegmatis genomic DNA and sub-cloned into prokaryotic expression vector pGEX-6p-1 with an N-terminal GST tag. Recombinant plasmid was transformed into E. coli BL21-Codon Plus (DE3) RIL for protein expression.

Cells were grown in Luria Bertani medium supplemented with 100 mg/L ampicillin at 37°C until OD₆₀₀ reached 0.8 and then induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 18 h at 16°C. Cells were harvested by centrifugation and lysed in 1 x PBS, pH 7.4 by ultrasonication. After centrifugation of cell lysates, supernatants containing soluble recombinant protein were loaded onto a GST-affinity chromatography (GE Healthcare) and then washed with 1 × PBS to remove non-specifically bound proteins. The proteolytic cleavage of the GST-tag was performed with PreScission Protease overnight in cold room, and the tagless MSMEG_1954 was further purified with a Superdex 200 (10/300) gel filtration column (GE Healthcare) pre-equilibrated with a buffer containing 25 mM Tris-HCl, pH 8.0, and 150 mM NaCl. Fractions containing target protein were then pooled and concentrated for crystallization. The expression and purification of MSMEG_1954 mutants used in the ATPase assay were carried out similarly as described above.

For crystallization, the purified MSMEG_1954 was concentrated to 20 mg/mL in 25 mM Tris-HCl (pH 8.0), 150 mM NaCl. The protein solution for Apo structure contained 20 mg/ml MSMEG_1954. To prepare the protein-ligand complex, MSMEG_1954 (20 mg/ml) and each ligand (protein: ligand molar ratio of 1: 5) were incubated overnight at 4°C before co-crystallization.

Initial crystallization screening was performed using sitting-drop vapor-diffusion at 293K and crystal screening kits from Hampton Research and Wizard from Rigaku. After about 1 month, crystals of apo-MSMEG_1954 (form 1) were obtained with reservoir solutions containing 0.2 M Ammonium acetate, 0.1 M Hepes (pH7.5), 25% w/v PEG3350. Crystals of apo-MSMEG_1954 (form 2) were obtained with a reservoir solution containing 0.2 M ammonium sulfate, 0.1 M Tris (pH8.5), 25% w/v PEG3350 after about 2 weeks. Crystals in complex with ADP grew in a solution containing 0.2 M calcium acetate hydrate, 0.1 M sodium cacodylate trihydrate (pH6.5), 18% PEG 8000. All crystals were soaked in the reservoir solution supplemented with 25% (v/v) glycerol before being flash-cooled in liquid nitrogen.

Diffraction data for form 1 crystals were collected on Beamline 17U of Shanghai Synchrotron Radiation Facility (SSRF) at 100K. Diffraction data for MSMEG_1954-ADP complex were collected on Beamline 18U of SSRF. Data were processed with the HKL2000 suite of programs (Otwinowski and Minor, 1997). Diffraction data for form 2 crystals were collected on beamline 19U of SSRF at 100 K and were integrated, scaled and merged using the HKL3000 (Minor et al., 2006).

The structure of form 1 and form 2 crystals was solved by molecular replacement (MR) with Phaser (McCoy et al., 2007) in CCP4 (Potterton et al., 2003) using the kinase domain of MABP-1 as a search template. The initial model was rebuilt by Phenix.AutoBuild (Adams et al., 2002). Manual adjustment of the model with Coot (Emsley et al., 2010) and refinement with Phenix located 395 residues and 293 waters in the model of form 1. In form 2, the electron density map of residues 1–97; 182–184, and 376–392 were not visible, and were therefore not included in the final structure. The MSMEG_1954-ADP complex structure was solved by molecular replacement (MR) using the form 2 crystal structure as a search template.

All structures were validated by MolProbity (Chen et al., 2010). The statistics of the phasing and refinement are summarized in Supplementary Table 1. The figures were prepared using PyMOL (DeLano Scientific) unless otherwise stated.

Site-directed mutagenesis was performed using Fast Mutagenesis System (Trans Gen Biotech) following the manufacturer’s instructions. The plasmids pMV261- msmeg_1954 and pGEX-6p-1-msmeg_1954 were used as the corresponding template for constructing a series of mutants for overexpression in M. smegmatis and for protein purification in E. coli, respectively. The pMV261-msmeg_1954 mutants were transformed into M. smegmatis mc²155, and pGEX-6P-1-msmeg_1954 mutants were transformed into E. coli BL21-Codon Plus (DE3) RIL. The N-terminal truncation of 98 residues (MSMEG_1954N^{△ 98}) was constructed into pMV261 and transformed into M. smegmatis mc²155 to obtain pMV261- msmeg_1954N^{Δ 98}-mc²155.

ATPase activity was estimated using QuantiChrom^TM ATPase/GTPase Assay Kit (BioAssay Systems). Reactions were performed in 96-well plates in a 40 μL volume with 20 μL of 2 × assay buffer (40 mM Hepes, 80 mM NaCl, 8 mM MgAc2, 1 mM EDTA, pH 7.5), 10 μg of purified MSMEG_1954 protein, or the indicated mutants. Reactions were initiated by the addition of 1 mM ATP. All measurements were performed at 25°C for 20 min for at least three repeats. To test the endogenous ATPase activity, a control with no enzyme was set up under identical conditions. In total, 200 μL of reagent was added to stop the reaction after incubation for 30 min at room temperature. The absorbance at 620 nm was measured using a microplate reader (BioTek, Synergy 4). The rate of ATP hydrolysis was then determined by assaying the liberated Pi and compared with a standard curve. Data were fitted to the standard Michaelis–Menten equation.

To investigate whether MSMEG_1954 is associated with macrolide resistance, we measured changes in the mRNA levels of MSMEG_1954 in response to macrolide treatment in mc²155. Quantitative reverse transcription PCR (qRT–PCR) analysis showed that the transcription level of MSMEG_1954 increases upon exposure to different concentrations of erythromycin or clarithromycin (Figures 1A,B).

Next, we constructed strains overexpressing MSMEG_1954 for antibiotic resistance testing. The MSMEG_1954 overexpressing strains (pMV261-1954-mc²155) have a growth advantage over M. smegmatis that has an empty pMV261 vector (pMV261-mc²155) when treated with 0.75 mg/L of clarithromycin (Figure 1C). Likewise, the pMV261-1954-mc²155 construct also showed an erythromycin resistant phenotype when compared with pMV261-mc²155 strains upon exposure to erythromycin (Figures 1D–F).

Taken together, these results indicate that MSMEG_1954 is induced in response to macrolide exposure and under the overexpression scenario tested here, appears to provide resistance in M. smegmatis just like MABP-1 in Mycobacterium tuberculosis.

Two crystal structures (form 1 and form 2) of apo-MSMEG_1954 were initially determined (Supplementary Table 1). No electron density for the N-terminal 97 residues (α1-α3 helices) in MSMEG_1954-form 2 was detected. Similar to MABP-1, MSMEG_1954-form 1 and MSMEG_1954-form 2 are both composed of two domains (Figure 2A), the kinase-like domain with a phosphate-binding loop (P-loop), and the N terminal accessory domain that consists of six α-helices: α1–α4 helices (residues 1–114) and α6–α7 helices (residues 158–199) (Figure 2B). The kinase-like domain is conserved in both structures. However, the N terminal accessory domain in MSMEG_1954-form 1 and MSMEG_1954-form 2 have different conformations. More specifically, the N terminal accessory domain (especially helices α1-α3 and α6-α7) of MSMEG_1954-form 1 has a very compact arrangement such that the domain is in direct contact with the kinase-like domain, which block the formation of substrate binding site (Figure 2C). In contrast, the remaining α6 and α7 helices of the N terminal accessory domain in MSMEG_1954-form 2 are positioned far away from the kinase-like domain (Figure 2D), in a way similar to MABP-1 (Figure 2E), which leave the substrate binding pocket open. The SDS-PAGE results of form 2 crystals showed four bands (Supplementary Figure 2), suggesting that the missing N-terminal 97 residues in MSMEG_1954-form 2 were probably degraded during crystallization.

To gain further insights into how MSMEG_1954 recognizes ATP, we solved the crystal structure of MSMEG_1954 in complex with ADP. The electron density for ADP was unequivocal (Figure 3A). The density of N-terminal 99 residues was still invisible in this structure. Moreover, structural superimpose show that MSMEG_1954-ADP is similar to MSMEG_1954-form 2 in an open state conformation (Figure 3B). The α6 and α7 helices of the N-terminal accessory domain are positioned far away from the kinase-like domain and fall in a position between the MSMEG_1954-form 2 and MABP-1 (Figure 3B).

Structural analysis confirmed that the ATP binding pocket of MSMEG_1954 is similar to that observed for MABP-1 (Figure 4A). The adenyl group is held in place by a π-stacking arrangement with the side chain of W239; the N1 atom of the adenyl group forms a hydrogen bond with the carbonyl nitrogen of M240. Hydrogen bonds are also formed between the exocyclic N6 atom of the adenyl group and the carbonyl oxygen of E238. Meanwhile, the side chain of K152 forms hydrogen bonds with the α and β-phosphate group of ADP, the backbone amide of S134 from the P-loop and its side chain form hydrogen bonds with β-phosphate group. The remaining interactions are mediated through the calcium ions which were introduced in the crystallization buffer. One calcium ion bridges the α-phosphate oxygen and the side chain of D299 and Q286, and is also linked to three adjacent water molecules. The second calcium ion is coordinated by the β-phosphate oxygen and D299, along with four adjacent water molecules.

As demonstrated in MABP-1, MSMEG_1954 also has ATPase activity with a kcat of 0.00864 s^–1 and Km of 0.231 mM (Figure 4B), values that are comparable to that for MABP-1 (kcat = 0.0137 s^–1, Km = 0.285 mM). The typical ABC transporters hydrolyses ATP with a Michaelis constant (Km) for ATP of 0.23 mM (Lin et al., 2015), which is similar to MSMEG_1954 and MABP-1. Compared with the wild-type enzyme, the K152A, E199A, E205A and W239A mutants lost 73, 74, 81, and 52% activity, respectively, emphasizing their importance in nucleotide binding and catalysis (Figure 4C). Importantly, all these mutants also have a significantly decreased erythromycin resistant phenotype in M. smegmatis, (Supplementary Figure 3), suggesting that antibiotic resistance also depends on the ATPase activity of MSMEG_1954.

Although the sequence alignment studies showed that the substrate binding pocket is conserved between MSMEG_1954 and MABP-1 (Supplementary Figure 1), we have not been able to obtain the structure of MSMEG_1954 bound with erythromycin. Previously, we have described the structure of MABP-1 bound with erythromycin (Figure 5A). The structural alignment of MABP-1- erythromycin and MSMEG_1954 provide details in substrate binding pocket (Figures 5B,C). The substate binding pocket in MSMEG_1954-form 1 is completely blocked by the N-terminal accessory domain. Meanwhile, the antibiotic binding pocket appears to be open in MSMEG_1954-ADP.

A series of mutants containing point mutations in the substrate binding pocket of MSMEG_1954 were made to test their ability to confer resistance to erythromycin. The K67A, E191A, R195A, M198G, D281A, I402A, R405A mutants all significantly reduced the ability of MSMEG_1954 to confer erythromycin resistance when compared with M. smegmatis strains overexpressing wild type MSMEG_1954 (Figure 5D), emphasizing their critical roles in binding and in erythromycin resistance.

In order to verify the physiological function of N-terminal domain, the construct with N-terminal 98 residues deleted (MSMEG_1954N^{△ 98}) was made. The antibiotic resistance testing showed MSMEG_1954N^{△ 98} completely abolished erythromycin resistance in M. smegmatis (Figure 5D), which demonstrates the indispensable function of the accessory domain in substrate binding.