Original approaches are needed to cope with drug-recalcitrant infections that threaten global health, including tuberculosis. Here, we implemented a drug-discovery strategy based on a customized microfluidic dynamic single-cell screening for pheno-tuning compounds (µDeSCRiPTor). Our strategy assesses dynamic phenotypic changes occurring at the level of single mycobacterial cells upon exposure to pheno-tuning compounds, via time-lapse microscopy. The goal was to identify compounds that synergize with existing anti-tubercular drugs, ultimately making the bacterial population more homogeneously susceptible to therapy. Among the four identified hits, we focused on M06, which shows promise for improving tuberculosis therapy.

Dynamic microfluidic screening videos (Dataset 1). A Mycobacterium smegmatis fluorescent reporter of DNA damage response (RecA-GFP), also constitutively expressing a red-fluorescent marker (mCherry_cyt) was analyzed by multi-phasic time-resolved microscopy using a customized multi-condition microfluidic platform. Cells were dispensed into the 32 microfluidic growth chambers and imaged at regular time intervals using a motorized inverted epifluorescence microscope, equipped with a 100X oil objective. Fluorescence imaging conditions were: GFP/FITC [Excitation (Ex) at 475/28 nm, Emission (Em) at 525/48 nm] 50% transmittance (T), 150 msec; mCherry (Ex 575/25, Em 625/45), 50% T, 150 msec. Coordinates of the location of individual bacteria or small groups of bacteria, referred to as points, were recorded and monitored throughout the three phases of the experiment: exponential growth in fresh medium; exposure to subinhibitory concentrations of 72 test compounds or control conditions; and washing of cells in fresh medium. Raw image stacks were extracted for each channel and smoothed. Binary masks of whole microcolonies were created from red-channel images, using K-means clustering to identify the regions of interest (ROI) containing only cells. All objects smaller than a single cell, separated from the main ROI, or placed on the edge of the image were omitted from the final mask. A shape complexity index was also calculated for each ROI, and images with an index lower than 14 were excluded. This process was reiterated for each stack of images (movie). Finally, binary masks were visually examined and only 54% of them were retained for further ROI analysis and statistics.

Snapshot images of M. tuberculosis expressing different NAT variants under stress (Dataset 2). Secondary cultures of M. tuberculosis H37Rv control (WT), NAT (rv3566c) mutant (MUT, Leu81_del), NAT-overexpressing (OE) were cultured in Middlebrook 7H9 broth until OD₆₀₀ 0.5, in duplicate. Secondary cultures of M. tuberculosis NAT-silencing (SI) and related control strain (CTsg) were diluted to OD₆₀₀ 0.05 and induced with anhydrotetracycline (100 ng/mL) for 3 days prior to stress exposure, in duplicate. Mid-log phase secondary cultures were treated for 24 h with DMSO (CT), H₂O₂ (30 mM), or with different drugs at 10X-MIC concentration: M06 (12.5 µg/mL); moxifloxacin (MOX, 500 ng/mL); isoniazid (INH, 250 ng/mL); rifampicin (RIF, 800 ng/mL). Next, bacilli were stained with either CellROX (5 µM) or FM464 (5 µg/mL) dyes for 30 min. Bacteria were imaged by phase-contrast and fluorescence using an inverted DeltaVision Elite Microscope (Leica) equipped with a UPLFLN100XO2/PH3/1.30 objective (Olympus). All samples were prepared by dispensing 0.6 μL of bacterial suspension between two #1.5 coverslips, sealed with glue. Exposure conditions: phase-contrast 50% T (channel 2), 100 msec; Cy5 (Ex 632/22, Em 679/34 – channel 1) 50% T, 100 msec for CellROX detection; TRITC (Ex 542/27, Em 597/45 – channel 1) 50% T, 100 msec for FM464 detection. Individual bacilli were automatically segmented in phase contrast, using Omnipose for cell segmentation in combination with a custom-trained model. The latter was obtained from training on a dataset consisting of the ground truth dataset provided as part of Omnipose and additional manually labeled images of M. tuberculosis. A Python notebook was written to extract single-cell fluorescence and cell area from snapshot images using the Omnipose-derived segmentation masks. Inverted masks were used to extract background fluorescence, which was subtracted from the fluorescence of each cell.

Time-lapse videos of M. tuberculosis RecA reporter stressed with INH and M06 (Dataset 3). Log-phase M. tuberculosis reporter of DNA damage response (RecA-mCherry), also constitutively expressing a green-fluorescent marker (GFP_cyt) reporter (OD_600nm 0.5) was diluted ten times in fresh 7H9 medium in the absence of detergents and seeded into our microfluidic hexa-device. The microfluidic assembly was mounted on a DeltaVision Microscope at 37 °C. Growth medium in the absence or presence of M06 (1.25 µg/mL, 1X MIC) and/or INH (MIC: 0.2 µ/mL, 8X MIC) was injected at 10-µl/min rate from a syringe pump, according to the different experimental phases. Images were acquired every 3 hours using a UPLFLN100XO2/PH3/1.30 objective (Olympus) and a high-speed sCMOS camera, 2,560 x 2,160 pixels, pixel size 6.5 x 6.5 μm, 15-bit, spectral range of 370–1,100 nm. Exposure conditions were phase-contrast 100 % T (channel 2), 200 msec, FITC (Ex 492/28, Em 523/23 – channel 1) 50% T, 50 msec; TRITC (Ex 556/25, Em 611/47 – channel 3) 100% T, 250 msec. Individual cells were manually segmented at specific time points using ROI Manager Macro of ImageJ 2.0.0.-rc-59/1.51n or counted over time using the Cell Counter plugin. Total fluorescence was normalized to cell area and background fluorescence was also subtracted.

Snapshot images of THP-1 macrophages infected with M. tuberculosis RecA reporter and treated with M06 (Dataset 4). THP-1 monocytes (ATCC TIB-202) were cultured in RPMI 1640 (Gibco) supplemented with 10% FBS and 1X PS. Each stock of about 10⁶ cells per mL in RPMI and 10% DMSO was defrosted at 37 °C, washed once in 10 mL, and inoculated in 30 mL of prewarmed complete RPMI using a T-175 flask. Cells were propagated at 37 °C in a humidified 5% CO₂ atmosphere until confluence, for up to four days. For final experiments, confluent cells were diluted to a concentration of 10⁵ cells/mL in complete RPMI without phenol red and without antibiotics and seeded into a 35-mm µ-Dish (ibidi) in the presence of PMA (30 ng/mL). After 36-hour incubation, cells were washed with fresh medium and re-incubated overnight before infection. Differentiated THP1 macrophages were infected with M. tuberculosis RecA-mCherry_GFP_cyt reporter diluted in RPMI without phenol red to OD₆₀₀ 0.005 (MOI 1:5). After 3 hours of incubation at 37 °C in humidified 5% CO₂ atmosphere, cells were washed five times with complete RPMI without phenol red. Infected cells were either left untreated (CT) or were treated with M06 (M06, 1.25 µg/mL, 1X MIC). Imaging was carried out at time zero (D0) and after 2 days (D2) of infection. Eight images were acquired for each field of view through a z-stack of 16 µm, starting the scan from the middle of the sample, using a 100X oil immersion objective, 1.4NA, WD 0.12mm (Olympus). Exposure conditions on the Delta Vision microscope: bright field 10% T, 50 msec (channel 0); FITC (Ex 475/28, Em 525/48 – channel 1) 50% T, 25 msec; TRITC (Ex 542/27, Em 597/45 – channel 2) 100% T, 100 msec. Image stacks were projected by maximum intensity in SoftWorx, and processed on ImageJ, by segmenting intracellular bacterial foci based on their green fluorescence. Derived masks were used to measure the size and both fluorescence channels to assess RecA induction and control the fluorescence of intracellular bacilli. Inverted masks were used to subtract background fluorescence from each ROI. The customized ImageJ macro used for analysis is also provided (Macro_Analysis_THP1_MTUB.ijm).

Open-source image analysis software, such as ImageJ, can be used to open both .tif and .dv files.