Antibiotics amikacin (AMK) and clarithromycin (CLR) were purchased from Sigma-Aldrich and stock solutions were prepared as per the manufacturer’s instructions. A library of 2,093 natural-product like scaffolds found in the 50,000 BioCore (BioDesign) compounds were purchased from Asinex¹. This set of unique chemotypes were chosen as representative compounds from each cluster identified from the larger BioCore library using the cheminformatics software package Molsoft ICM Chemistry Pro² and JKlustor (ChemAxon).

Mycobacterium abscessus strains were a kind gift from Dr. Thomas Byrd from the University of New Mexico (Howard et al., 2006; Greendyke and Byrd, 2008; Brown-Elliott et al., 2012). M. abscessus 390S (smooth), and 390R (rough) were grown at 37°C in Middlebrook 7H9 supplemented with 10% OADC. Antibiotics [50 μg/ml kanamycin (KAN) or 32 μg/ml amikacin (AMK)] were added for selection when appropriate. M. abscessus 390R strain was passaged 5X with a tuberculin needle (21G) to reduce clumps before carrying out experiments.

Mycobacterium abscessus strains 390S and 390R were transformed with plasmids, episomal pVVRG (Rodrigues Felix et al., 2017) and integrative pMV306hsp+LuxG13 (Andreu et al., 2010) by electroporation. The former plasmid expresses mCherry and GFP constitutively from smyc and hsp60 promoters, respectively while the latter plasmid carries the entire LuxCDABE operon. The bioluminescent reporter plasmid is a kind gift from Brian Robertson & Siouxsie Wiles (Addgene # 26159). Transformants were selected on 7H10 plates supplemented with OADC and 50 μg/ml KAN. The resulting strains are referred to as 390S-mCh/390R-mCh and 390S-lux/390R-lux throughout this manuscript. Table 1 details the bacterial strains and plasmids used in this study.

The stability of the reporter plasmids was assessed by comparing the signal as well as cell viability (by CFU) of M. abscessus strains, 390S-lux and 390S-mCh, when grown in the presence and absence of the selection antibiotic. The strains were grown to logarithmic phase in Middlebrook 7H9 media containing KAN, centrifuged, washed 3X with phosphate buffered saline (PBS) and diluted to an optical density (OD₆₀₀) of 0.05 (∼ 6^∗10⁶ CFU/ml) in 7H9 media. The diluted culture was then added to 96-well plates at a final volume of 100 μl per well with either KAN or no KAN. White and black 96-well solid bottom plates (Corning) were used for measuring luminescence and mCherry fluorescence, respectively. The plates were incubated at 37°C and 5% CO₂. To avoid evaporation artifacts such as edge effects, the plates were kept in a humidifying box that contained reservoirs filled with water. This incubation method was followed for other experiments throughout this manuscript. Signal was read at different time intervals (0, 24, 48, and 72 h) using a Biotek Synergy H4 plate reader. The red fluorescent signal was measured at excitation/emission wavelengths of 575/620 nm and bioluminescence was measured as relative light units (RLUs) with the reading time of 1 s. After measuring the signal, an aliquot was removed from each well (both with and without KAN selection) at the indicated times, serially diluted and plated on Middlebrook 7H10 KAN agar plates.

The M. abscessus 390S strain was examined for susceptibility to three commonly used organic solvents- dimethylsulfoxide (DMSO), dimethylformamide (DMF) and acetone. A logarithmic culture of 390S-lux was diluted to an OD of 0.01 (∼2 × 10⁶ CFU/ml) and added to a range of 0.5–10% of DMSO, DMF and acetone in a white solid bottom 384 -well plate. Controls containing no organic solvent (negative control) or AMK (positive control) were included. The plates were incubated at 37°C and luminescence was read at 0, 24, 48, and 72 h using a Biotek plate reader. Susceptibility to organic solvents was calculated as percent inhibition using the formula [(negative control signal - sample signal)/negative control signal ^∗ 100].

The fluorescence and luminescence signal readouts were optimized for the 390S strain in a 384-well plate with different starting inoculum sizes and volumes. Logarithmic phase cultures of 390S-lux and 390S-mCh were diluted to three different ODs -0.01, 0.05 and 0.1 – in Middlebrook 7H9 broth. The diluted culture was then added to appropriate plates at three different volumes -30, 50, and 70 μl per well – for luminescence and mCherry fluorescent signal. The signal was read at time intervals of 0, 24, 48, and 72 h of incubation using a Biotek plate reader using the parameters described earlier. AMK (32 μg/ml) was included as a positive control. A Z-factor, a statistical parameter of HTS quality control was calculated for each time point, OD and volume as described previously (Zhang et al., 1999). A Z-factor quantifies the quality of an assay and it takes into account the assay signal dynamic range and data variation associated with the signal. A Z-factor value of 0.5 and above is indicative of a robust HTS assay.

Minimum inhibitory concentrations (MIC) were determined with CLR and AMK, clinically relevant antibiotics known to be active against M. abscessus. A two-fold serial dilution of the antibiotic in a 16-step process was carried out for CLR (128-0.0039 μg/ml) and AMK (256-0.0078 μg/ml). M. abscessus 390S-lux and 390R-lux cultures were diluted from a logarithmic-phase culture in Middlebrook 7H9 broth and added to the appropriate 384–well microtiter plates at a final OD₆₀₀ of 0.01 in a total volume of 30 μl per well. An untreated control was also included. The plates were incubated at 37°C and luminescence and fluorescence was read at 0, 24, 48, and 72 h. A modified Gompertz model was used to fit the data and to generate dose-response curves using GraphPad Prism (Lambert and Pearson, 2000). MIC, defined as the lowest drug concentration at which more than 99% of bacterial growth is inhibited, was calculated from the fitted curve as compared to the untreated control.

A library of 2,093 natural product-like compounds obtained from Asinex were screened for activity against M. abscessus. M. abscessus 390S-lux culture diluted to an OD₆₀₀ of 0.01 was added to a 384-well solid bottom white plates containing compounds at a final concentration of 34 μg/ml in 30 μl volume. The compounds and the culture were added using a liquid robotic handler (Precision XS 2000, Biotek). DMSO (2%) and amikacin controls were also included in the screening plates. The screening plates were incubated for a period of 72 h. Bioluminescence was measured after every 24 h. Percent inhibition and Z-factor were calculated as described earlier. A compound exhibiting inhibition ≥ 50% relative to the untreated control are considered potential hit candidates.

An in vitro killing assay was employed to confirm the hits obtained from the Asinex library screen. Six best hits that showed > 70% inhibition at all time points in the HTS bioluminescent screen were added at 34 μg/ml (100 μM) to the M. abscessus 390S-lux culture of OD₆₀₀ 0.01 in a 384-well solid bottom white plate and incubated at 37°C for 48 h. A DMSO vehicle control was also included. Luminescence was measured and aliquots from each well were serially diluted in PBS of which 50 μl of dilutions (10^-1–10^-8) were plated onto Middlebrook 7H10 quad-plates. Colonies were counted after 5 days and percent inhibition relative to its respective DMSO control was calculated. MIC determination of the selected hits (34.0–0.001 μg/ml) was carried out as described earlier while maintaining the final 2% DMSO throughout the dilution.

We constructed fluorescent reporter strains 390S-mCh/390R-mCh (episomal mCherry expression) and luminescent reporter strains 390S-lux/390R-lux (chromosomal expression of the entire luxCDABE operon). An episomal mCherry construct, rather than an integrated construct was utilized to achieve acceptable S/B ratios due to multiple plasmid copies per bacterial cell. Because we intended to avoid antibiotic selection for reporter plasmids during antimicrobial assays, we first examined the stability of the plasmids in the absence of selective pressure to ensure consistent signal output. To do this, we measured the fluorescent (mCherry) and luminescent (Lux) signal output over time with and without antibiotic selection (KAN). The steady signal ratio ∼1 under +KAN/-KAN conditions over a period of 72 h indicated little to no loss of plasmid in the absence of selection (Table 2). This was confirmed by plating the cultures onto KAN containing media. We observed no difference in CFU when 390S-mCh and 390S-lux cultures were grown with or without KAN selection for a 72 h period (Figure 1). This indicates that both the episomal fluorescent and integrative bioluminescent plasmids are stable in the absence of the selection antibiotic for at least 72 h and therefore, are suitable for use in testing antimicrobial activity of drugs.

We examined the tolerance of M. abscessus 390S reporter strains for organic solvents commonly used to dissolve compounds in most drug libraries. We exposed 390S-lux to DMSO, DMF, and acetone at concentrations ranging from 0 to 10% for 72 h. We found that M. abscessus was highly susceptible to DMF followed by DMSO and acetone (Supplementary Figure S1). Inhibition > 20% was taken as an indication of toxicity not suitable for drug screening purposes. DMF concentration above 0.5% showed inhibition greater than the 20% cut-off. In contrast, DMSO did not result in significant inhibition even when the concentration was increased to 2% which is four-fold higher than deleterious DMF concentrations. Acetone was very well-tolerated and did not result in inhibition greater than 20% even at the highest concentration of 10%. This information is critical to guide decisions regarding choice of solvents and appropriate concentrations to use in HTS drug assays without compromising the growth of M. abscessus.

After ensuring that the reporter plasmids are stable, we optimized the drug screening conditions in a 384-well plate for both the reporter strains (390S-lux and 390S-mCh). Several parameters such as the starting inoculum, volume and incubation time were simultaneously evaluated to achieve an optimal Z-factor > 0.5. Three different cell densities of 390S-mCh and 390S-lux (0.01, 0.05 and 0.1) were tested with different volumes (30, 50, and 70 μl) and at different times (24, 48, and 72 h). We obtained an excellent Z-factor close to 0.8 with cultures of OD₆₀₀ 0.01 irrespective of the sample volume and incubation time (Figure 2 and Supplementary Table S1). Z-factors dropped with an increase in OD above 0.01. We also achieved a robust gain of signal in the untreated control versus the AMK-treated control with both the reporters (Figure 3). The fluorescence signal reached the highest intensity after 72 h while the luminescence signal reached saturation after 24 h (Figure 3). We also determined the signal-to-background ratios (S/B) with each optimized parameter to assess the background contribution to the signal readout (Supplementary Table S2). A minimum signal of 10-fold over the background was achieved with both signal types for M. abscessus 390S strain indicating assay reliability. We observed S/B ratios with luciferase at least 70-fold higher than with fluorescent mCherry as expected. Based on the Z-factor and signal intensity, we selected the starting inoculum of OD₆₀₀ 0.01 and 30 μl volume for both assays, with incubation times of 72 and 48 h for the fluorescence and luminescence read-out assay, respectively. The choice of the smallest working volume (30 μl) was also based on maximizing miniaturization and cost-effective use of compounds. Reproducible S/B ratios > 10 and Z-factors > 0.8 with both fluorescent and luminescent reporters indicates two robust, 384-well screening assays. The bioluminescence reporter was used for further validation studies because of sensitivity.

To validate our assays for drug susceptibility testing, we determined the MIC of two antibiotics, the macrolide CLR and the aminoglycoside AMK, using bioluminescent M. abscessus strains (390S-lux and 390R-lux). These antibiotics were chosen for two reasons. First, their MIC is well documented which allows us to validate MIC determined via the reporter-based assay. Second, they are clinically relevant front-line treatment options for M. abscessus infections (Brown et al., 1992; Maurer et al., 2012; Lee et al., 2015). Dose-response curves are shown in Supplementary Figure S2 and MIC values are listed in Table 3. The range of MIC values determined with the smooth variant for the 24–72 h time period is 0.2–2 μg/ml for CLR and 4–10 μg/ml for AMK. The MIC values of amikacin are at least 5-fold higher compared to clarithromycin with both the M. abscessus morphotypes consistent with the previous reports (Greendyke and Byrd, 2008; Ferro et al., 2015). An increase in MIC values was observed with the length of incubation time as noted by others with M. abscessus, other bacteria and fungi (Kenny and Cartwright, 1993; Pujol et al., 1997; Paphitou et al., 2002; Egervarn et al., 2007; Lee et al., 2014). We did not observe differences in the MIC values for CLR and AMK between the two strains, 390S and 390R (Table 3). Slightly higher MIC exhibited by 390R in contrast to 390S was found to be statistically insignificant (one-sample t-test).

To validate the utility of our assay for M. abscessus drug discovery, a pilot screen of 2,093 diverse natural product-inspired scaffolds (Asinex) was carried out for activity against M. abscessus (390S) at a concentration of 34 μg/ml. The bioluminescence signal was used to monitor bacterial growth/inhibition after every 24 h over a period of 72 h with the intent to capture potent hits early on and less potent ones at later time-points. Loss of signal as compared to the untreated control indicates an active compound. Z-factor calculated across all the drug plates were above 0.5 over the course of 72 h indicating robustness of the screen. We obtained a total of 49 hits (hit rate of 2.3%) based on the cutoff of ≥50% inhibition relative to the DMSO control (Figure 4 and Supplementary Table S3). Five hits showed > 90% inhibition of M. abscessus with two appearing at 24 h and 3 hits at 48 h. Similarly, the number of hits with >80% inhibition were higher at 48 h than at 24 and 72 h (Table 4). We chose to validate 6 potent hits that exhibited > 70% inhibition at all time points from the primary screen by repeating the luminescence based assay and also by enumerating CFU. Five of the six selected hits showed greater than 50% percent inhibition compared to the DMSO control in both of the two assays (luminescence and CFU) after 48 h incubation (Table 5) indicating the validity of the reporter-based assay. The structures, molecular formulas and SMILES ID of the five validated compounds are shown in Figure 5 and Supplementary Table S4. We also determined MIC values (Table 5) to better rank the hits based on potency. Hit compound 33 appeared to be most potent indicated by the lowest MIC value of 4.8 μg/ml (Table 5), a reasonable starting point for hit-to-lead development.