Et₃N (15.2 mL, 0.109 mmol), oxyma (4.7 mg, 0.0328 mmol), and EDCI (6.3 mg, 0.0328 mmol) were added to a stirred solution of vancomycin (32.5 mg, 0.0219 mmol) and Boc-Gly-OH (4.6 mg, 0.0263 mmol) in DMF (0.2 mL). After being stirred for 22 h at room temperature, the reaction mixture was filtered. The filtrate was purified by reverse phase HPLC to afford Boc-Glycyl-vancomycin (85% yield). 0.1 mL of 4.0 M HCl (dioxane solution) was then added to a stirred solution of Boc-Glycyl-vancomycin (3.9 mg, 2.39 mmol) in dioxane (0.1 mL). After 4 h at room temperature, all volatiles were removed under reduced pressure to afford HCl-Glycyl-vancomycin. Et₃N (1.0 mL, 7.32 mmol) and cyanin5.5 NHS ester (0.52 mg, 0.732 mmol) in DMF (0.1 mL) were added to a stirred solution of HCl-Glycyl-vancomycin in DMF (0.1 mL). After 6 h at room temperature, all volatiles were removed under reduced pressure to afford the crude DLF-1. This was purified by reverse phase HPLC to afford DLF-1 (>95% purity).

Binding of bacterial PG peptide to vancomycin or DLF-1 was monitored using an ITC₂₀₀ (Microcal, Piscataway, NJ). Vancomycin hydrochloride from Streptomyces orientalis (32 μM final concentration, V2002, Sigma) or DLF-1 (7 μM final) was diluted in 1X PBS pH 7.4 (21-040-CM, Corning), and 204 μL was loaded into the cell. Experiments contained 1% DMSO (BP-231, Fisher) in both syringe and cell, to aid solubility. Peptide Ac-L-Lys-D-Ala-D-Ala was dissolved in the same buffer (600 or 100 μM, respectively), and 39.4 μL was loaded in the syringe. For each binding, three separate experiments were conducted at 25°C (298.15K). Injections were performed with serial injections of peptide; first, one injection of 1 μL, followed by 19 incremental injections of 2 μL, at 120 s intervals. Data from the first injection was excluded, due to pre-equilibration mixing between the contents of cell and syringe at the syringe tip. Peak areas were integrated, normalized, and then fitted by non-linear regression using the independent sites model in Origin™ (version 2.3.6, Microcal, Piscataway, NJ).

The mycobacterium strains (M. smegmatis, M. bovis BCG, and Mtb) were started from −80°C stocks. For M. bovis BCG, and Mtb, cultures growth in the presence or absence of DLF-1 were performed in 7H9 broth (Difco, Detroit, MI) supplemented with 0.5% glycerol, 10% oleic acid dextrose complex without catalase and 0.05% Tween 80 (MOAD/T medium) at 37°C without shaking. The bacterial cultures were then started in the presence of different concentrations of DLF-1 (0, 100 nM, 1 μM) and monitored OD over time in a 96-well plate with a Tecan multimode reader (Infinite 200 Pro, Tecan) at absorbance 600 nm with bandwidth 9 nm and number of flashes as 25. OD was measured daily for M. bovis BCG and Mtb to plot growth curves. For M. smegmatis, culture was set up in a 96-well plate, and the plate was kept in the reader at 37°C. OD was measured hourly with automatic shaking for 10 min before OD measurement. The Supplementary Figure 1 shows the correlation between the OD values measured with the Tecan microplate reader at 600 nm and the OD value in a cuvette measured with commonly used ThermoScientific Genesys 20 spectrophotometer (ThermoFisher Scientific) at 600 nm, which helps convert the OD values of the Tecan microplate reader into OD values in a cuvette measured with the Genesys 20.

Minimum inhibitory concentration of DLF-1 against M. smegmatis at which 90% of strains were inhibited (MIC₉₀) was determined using the standard broth microdilution method in 96-well microtiter plates as described previously with minor modifications (Swenson et al., 1982). Briefly, serial doubling dilutions of DLF-1 were prepared in MOAD broth to achieve final concentrations ranging from 8 to 0.125 μg/ml. M. smegmatis was added to each well with a final inoculum of ~10⁵ CFU/ml. The plate was incubated at 37°C. OD₆₀₀ was measured at 48 h with a Tecan microplate reader. The wells with M. smegmatis alone were controls.

Bacterial cultures were stopped when they reached the logarithmic growth phase (OD₆₀₀, 0.4~0.6). For each strain of studied bacteria, a series of dilutions were made starting from OD₆₀₀ of 1.0 measured with Genesys 20 in growth medium. Bacteria cells were centrifuged at 9000 rpm for 5 min, and cell pellets were washed with 1X PBS twice and finally resuspended in 1X PBS containing 100 nM DLF-1 for 3 h on ice (for fast-growing M. smegmatis) or at room temperature (for M. bovis BCG and Mtb CDC1551). After incubation with DLF-1, bacterial cells were centrifuged (9000 rpm, 5 min), washed once with 1X PBS (10,000 rpm, 10 min) to remove unbound DLF-1, and were finally resuspended in equal volumes of 1X PBS solutions. Each sample was added to a fluorescence 96-well microplate, and fluorescence intensity of DLF-1 was measured with a multimode fluorescence plate reader (Infinite^R 200 PRO, Tecan) at excitation 680 nm and emission 710 nm. Each strain of bacteria was also titrated to enumerate CFU. For determination of threshold of detection, we optimized the condition of the Tecan reader, with gain value as 150, excitation 670 nm and emission 700 nm. For cell infection, a series of concentrations of a tdTomato-expressing Mtb strain were incubated with 100 nM DLF-1, then were used to infect THP-1 macrophages and A549 human lung epithelial cells. Fluorescence was measured 3-h post infection by the Tecan plate reader.

A tdTomato expressing Mtb strain (Erdman-pJDC60) was applied to infect cells after labeling bacteria with DLF-1. This strain was cultured in MOAD/T medium with addition of 25 ug/ml kanamycin until it reaches the logarithmic phase (OD₆₀₀, 0.4~0.6). A549, or THP-1 cells were grown on the coverslips in 24-well cell culture plates at a concentration of 2 × 10⁵ cells/well. For THP-1, phorbol 12-myristate 13-acetate (PMA) was used to stimulate the cells for 3 days. The cells were then infected with bacteria at 37°C for 3 h. After bacterial infection, extracellular bacteria were aspirated, and the infected cells were washed with 1X PBS, treated with amikacin (200 μg/ml) for additional 2 h to eliminate extracellular bacteria as described previously (Mehta et al., 1996; Napier et al., 2011). Bacterium infected cells were then fixed with 4% freshly made paraformaldehyde for 10 min at room temperature. After washing with 1X PBS for three times, the fixed infected cells were permeabilized with 0.2% Tween-80/PBS (v/v) solution for 5 min, washed again with 1X PBS for three times, and then stained with DAPI solution (300 nM) for 5 min in the dark. After another three-time washing with 1X PBS, the coverslips containing bacterium infected cells were taken out and mounted to glass slides for fluorescence microscopy analysis with a deconvolution microscope (Carl Zeiss). For imaging of DLF-1 labeled dormant Mtb inside macrophages, a Mtb CDC1551 strain carrying a genome integrated tdTomato gene (CDC1551-pYK13) was cultured in a modified Wayne non-replicating persistence model. THP-1 cells seeded in a 24-well plate containing cover slips were stimulated with PMA for 3 days, and then infected with the DLF-1 (100 nM) labeled, non-replicating td-Tomato-expressing Mtb. After bacterial infection, the same protocols of DAPI staining, slide mounting, and microscopy imaging were conducted. Images were taken with a 63X oil lens.

We constructed a tdTomato-expressing Mtb in a more stable manner by site-specifically integrating a single copy tdTomato into the mycobacterial chromosome, using a mycobacterial plasmid containing the phage attachment site attP and the L5 integrase (int) gene from the mycobacteriophage L5 genome, as has been shown by others for bioluminescent reporters (Andreu et al., 2010, 2013). To construct this vector, we began with replacing the GFP gene in pFJS8, an E. coli-mycobacterium shuttle plasmid containing the L5 promoter (Miltner et al., 2005), with a fragment from pDONR222 (Invitrogen) containing a suicide ccdB gene and a chloramphenicol resistance gene (Cm^R) at restriction sites of HindIII/KpnI. In this way, the expressions of ccdB and Cm^R are under a L5 promoter. We then cut the L5-ccdB- Cm^R fragment from the new vector at XhoI and KpnI and ligated it into XhoI/KpnI digested pYUB412, a mycobacterial integrating cosmid vector (Bange et al., 1999). The tdTomato gene was then PCR amplified from pRSETB-tdTomato using an up-stream primer containing HindIII site at 5′ end and a down-stream primer containing KpnI site at 5′ end (F: 5′-TATAAAGCTTGGATCCATGGTGAGCAAGG-3′; R: 5′-TATAGGTACCGAATTCTTACTTGTACAGCTCGTCC-3′). The purified PCR product was then digested with HindIII and KpnI and ligated into the HindIII/KpnI digested pYUB412 containing L5-ccdB- Cm^R. The new plasmid is designated as pYK13. After transformation of the pYK13 into CDC1551, we selected colonies with hygromycin selective agar plates, and confirmed the success of integration of tdTomato into Mtb genome by PCR amplifying the tdTomato and hygromycin resistance genes, and by the presence of tdTomato specific fluorescence.

We modified the Wayne non-replicating persistence (hypoxic shift-down) model by substituting Dubos with 7H9 (containing 0.05% Tween 80 and 10% OAD) and using the Mtb strain expressing genome-integrated tdTomato (CDC1551-pYK13). Mtb cultures were set up at 10⁵–10⁶ cfu/ml in 12 × 75 mm tubes in a total volume of 3.0 mL (head space ratio, HSR = 0.5). Each tube contained a 13 × 4 mm diameter stir bar (Fisher, PA, USA). Cultures were incubated in a 37°C incubator on a stir plate with 200 rpm. Some cultures also contained Methylene Blue at a concentration of 1.5 mg/L to be used as an indicator of oxygen depletion. An additional tube of Methylene Blue without Mtb was used as a color reference. When the Methylene Blue-containing tube turned colorless, cultures were incubated for an additional week (total 18–22 days), and then were used for DLF-1 fluorescence labeling. A sample of the culture was carefully removed from the middle portion of each tube by pipette to avoid disturbing the pellicle on the surface or the sediment on the bottom. Optical density reading (600 nm) of the cultures was determined. A series of dilutions were made and then centrifuged (9000 rpm, 5 min), washed once with 1X PBS/T, and then resuspended in 1X PBS. DLF-1 labeling of the dormant cultures were performed similarly as for labeling of other bacteria before. The Mtb CDC1551 strain carrying a genome-integrated tdTomato gene was cultured in the Wayne model to reach dormant state, diluted in various concentrations, and then labeled with DLF-1. Fluorescent intensity of DLF-1 labeled dormant Mtb was plotted against the concentration of bacteria or concentration of DLF-1, respectively. All conditions were set up in triplicates. Samples of dilutions were taken for CFU enumeration on 7H10 agar plates in duplicates. Colonies were enumerated at 4–6 weeks following inoculation.

Mtb transposon mutant of plcC, CDC1551, ΔhbhA and MT103 strains were cultured in MOAD/T medium. Bacterial cells grown to middle-logarithmic phase (OD_600nm, 0.4~0.6) were harvested, labeled with 100 nM DLF-1 at room temperature for 3 h, washed and suspended in antibiotic free A549 cell culture medium. A549 cells were cultured and seeded to 96-well fluorescent microplates (5 × 10⁴/well) 1 day before experiment. DLF-1 labeled Mtb mutants and wild type control strains were used to infect A549 cells at a MOI of 50. Fluorescence of the microplates was measured to quantitate bacteria loaded initially. After 3 h of infection, infected cells were washed once with 1X PBS, then added cell culture medium containing 200 μg/ml amikacin and incubated for another 2 h to remove extracellular bacteria. After aspirating amikacin containing medium and washed with PBS solution, fluorescence of the microplates were measured to quantitate intracellular bacteria. Bacterial infection rate of each strain was calculated as the average fluorescence intensity at the end of infection divided by the average fluorescence at the beginning. Bacterial infection ratio of mutant to wild type control is calculated as the mutant infection rate divided by that of wild type strain. Bacterial samples from the initial and end of infection were taken and plated for CFU analysis.

The new fluorescent probe was designed to bind to the terminal peptide D-Ala-D-Ala of the PG precursor. DLF-1 was synthesized by conjugating fluorescent Cy5.5 with glycyl-vancomycin (Figure 1). Isothermal Titration Calorimetry (ITC) was applied to verify that DLF-1 binds to the D-Ala-D-Ala motif of PG. ITC measures the heat release (or absorption) associated with a binding event, offering an orthogonal, non-fluorescence based, validation of binding (Pierce et al., 1999). ITC shows that DLF-1 binds stoichiometrically (N = 0.99) to the peptide Ac-L-Lys-D-Ala-D-Ala (AcKAA) in PBS (Figure 2) with high affinity (K_d 2.3 μM). The binding affinity for DLF-1 is similar to parent molecule vancomycin (2.5 μM, Supplementary Table 1) and in line with prior studies (Leavitt and Freire, 2001; Rekharsky et al., 2006). These results suggest that DLF-1 has suitable biochemical target binding properties to advance to microbial profiling.

In order to label actively growing bacteria, chemical probes should not affect bacterial growth. We tested whether DLF-1 labeling inhibits growths of M. smegmatis, M. bovis bacillus Calmette-Guérin (BCG) and Mtb CDC1551. Two concentrations of DLF-1 were used to incubate with bacteria grown in liquid media. The growth curves of bacteria incubated with two concentrations of DLF-1 were compared with those without DLF-1. One μM of DLF-1 in culture media significantly inhibited growth of M. smegmatis, while having much less of effect on growth of M. bovis BCG and Mtb CDC1551. When 100 nM DLF-1 was used, no significant effect on growth was observed for Mtb (Figure 3). The MIC₉₀ for DLF-1 against M. smegmatis was determined as 1 μg/ml (~0.45 μM) by broth microdilution method. We then decided to use 100 nM of DLF-1 for the remaining bacterial labeling.

We have validated the application of DLF-1 to detection of actively replicating Mtb in vitro, in comparison with fast-growing M. smegmatis, and slow-growing M. bovis BCG. The correlation between fluorescent signal and bacterial numbers was evaluated after DLF-1 labeling of each Mycobacterium spp. Fluorescence generated from DLF-1 labeling linearly correlated with bacterial CFUs in all tested mycobacterial strains (Figure 4). The lowest number of Mtb within the linear range is 5.3 × 10⁴ (P < 0.05, Figure 4D).

We also examined the threshold for differentiation of DLF-1 labeled Mtb from the same number of Mtb without DLF-1 labeling using an optimal combination of excitation and emission wavelengths and gain value set up in the multimode reader. An Mtb strain carrying a plasmid expressing tdTomato was used (Kong et al., 2016), and the tdTomato fluorescence served as an internal control for DLF-1 labeling. When measured at the optimal excitation and emission wavelengths of the Cy5.5 fluorophore on DLF-1, DLF-1 labeled Mtb showed significantly higher fluorescence than the bacteria without labeling at each dilution of bacteria (Supplementary Figure 2A). When measured at the appropriate wavelengths for tdTomato, there was no such a difference between the two groups at any dilution of bacteria, suggesting there are the same bacterial numbers in the two groups at each dilution (Supplementary Figure 2B). The threshold of detection of Mtb for DLF-1 is ~10³ CFU using the multimode plate reader with the optimal wavelengths and gain value described in Materials and Methods (P < 0.05). Although this condition can differentiate 10³ CFU from the non-labeled control, the detected fluorescence is not correlated with CFU very well (R² = 0.75).

We examined if DLF-1 could be applied to image Mtb in cell infection models. A tdTomato-expressing Mtb strain was incubated with DLF-1 and infected THP-1 macrophages and A549 human lung epithelial cells. Fluorescence was detected from MOI = 10 in both cell lines (Figure 5A), and the fluorescence measured from DLF-1 labeled Mtb increased as MOIs rose. The fluorescent intensity of endogenous tdTomato among different MOI groups in A549 or THP-1 cells was measured in parallel with fluorescence of DLF-1. Cell infection with DLF-1 labeled Mtb was further verified by fluorescent microscopy. Slides with fixed infected cells were imaged with a fluorescent deconvolution microscope using pseudo colors: tdTomato-green, Cy5.5-red and DAPI-blue. As shown in Figure 5B, tdTomato and Cy5.5 are co-localized in cytoplasm of THP-1 and A549, which confirmed that DLF-1 can be applied to detect labeled Mtb inside of cells.

In order to study molecular mechanisms involved in the onset of latency and/or reactivation, several in vitro and in vivo models have been developed to mimic LTBI in human. The Wayne model developed in 1996 is a well-characterized in vitro model based on hypoxic growth of Mtb (Wayne and Hayes, 1996). In this model, tubercle bacilli are cultured in an undisturbed liquid medium and exposed to oxygen depletion by a way of a completely sealed culture container. Mtb undergo an orderly process of metabolic shutdown, and develop resistance to rifampicin and isoniazid, but are sensitive to the anaerobic bactericide metronidazole (Wayne and Lin, 1982; Wayne, 1994; Wayne and Hayes, 1996). To validate DLF-1 in detecting dormant Mtb, we labeled dormant Mtb generated by the Wayne model with DLF-1.

An Mtb CDC1551 strain with a genome-integrated tdTomato gene was used in the Wayne model to generate dormant Mtb. With this strain, fluorescence of tdTomato can serve as an internal positive control for DLF-1 labeling. The dormant state of Mtb from this model was confirmed by verifying drug tolerance. After Mtb reached dormant state, 0.1 μg/ml rifampin was added in the culture for 3 days. The proportion of survived dormant Mtb reached 23.5 ± 4.8% (calculated as CFU of treated/CFU of untreated), which is consistent with previously reported results (Wayne and Hayes, 1996). With the same dose of rifampin, 97.5% of actively replicating Mtb was killed (Supplementary Table 2).

After the hypoxic dormant bacilli were obtained, cultures were washed with PBS, diluted into a series of concentrations, and incubated with DLF-1. The bacteria were plated for CFU enumeration. The fluorescent signal from labeled bacteria was measured with a multimode plate reader, using Cy5.5 and tdTomato wavelengths, respectively. Correlation between DLF-1 fluorescence and DLF-1 concentrations was analyzed first to determine dose-response relationship of DLF-1 labeling (Supplementary Figure 3A). The measured fluorescence increased as the concentration of DLF-1 went higher. Correlation between fluorescence and bacterial numbers was also analyzed using DLF-1 and tdTomato wavelengths, respectively (Figure 6A and Supplementary Figure 3B). Both DLF-1 specific fluorescence and tdTomato fluorescence correlated very well with bacterial CFUs. The threshold of detection for DLF-1 labeled dormant Mtb was determined as 3.1 × 10⁵ CFU (Figure 6B). DLF-1 labeling dormant Mtb was also confirmed with fluorescent microscopy using a deconvolution microscope as shown in Figure 6C for bacterial smear and in Figure 6D for intracellular dormant Mtb in THP-1 cells. The red pseudo color from DLF-1 co-localized with the green pseudo color from tdTomato very well both in smear slides and in cell infection slides, which confirms that DLF-1 is capable of labeling dormant Mtb, and the fluorescent signal can be measured when bacteria are inside of cells.

Conventionally, identification of Mtb virulence genes relies on comparison of bacterial CFUs between mutant and wild type strains under specific conditions. DLF-1 labeling has great potential to replace CFU enumeration in vitro. As DLF-1 labeling can quantitate labeled Mtb inside cells, we set out to identify bacterial genes that are critical for cell entry. Previous studies have reported that the heparin-binding hemagglutinin (HbhA) of Mtb plays an important role in infecting epithelial cells (Pethe et al., 2001; Mueller-Ortiz et al., 2002; Parra et al., 2004; Choi et al., 2013). HbhA was initially identified in Mtb and M. bovis BCG (Menozzi et al., 1996, 1998). It is located on the surface of the mycobacterium and mediates binding of the bacillus to epithelial cells and fibroblasts (Menozzi et al., 1996). Deletion of hbhA (ΔhbhA) leads to reduced adherence of bacteria to epithelial cells but shows no effect on adherence to macrophages (Pethe et al., 2001). We evaluated entry rates of ΔhbhA and the wild type Mt103 strain in A549 human lung epithelial cells using DLF-1 labeling, in comparison with CFU enumeration. Infection ratios of ΔhbhA to the wild type strain Mt103 at a MOI of 50 was <1 based on both DLF-1 fluorescent intensity and CFU results (Table 1). The infection ratio calculated using DLF-1 fluorescent intensity was consistent with the infection ratio calculated using CFU, which suggests that DLF-1 can be applied to screen bacterial genes that are critical for cell infection, and the ΔhbhA strain can serve as a positive strain for this purpose.

Phospholipases are important bacterial virulence factors (Titball, 1998). Among them, phospholipase C (Plc) is important for replicating and cell-to-cell spread in mouse peritoneal macrophages for Listeria monocytogenes (Camilli et al., 1991). The Plc of Clostridium perfringenes is also named α toxin, which is haemolytic and suppresses neutrophil respiratory burst by disturbing host cell protein kinase C signaling (Terada et al., 1999). Mtb has four Plc genes, designated plcA, B, C, and D. The plcA, B, and C loci are clustered, while plcD is in a separated region in the chromosome of Mtb CDC1551. Expression of plc genes are upregulated at the 24-h post infection in macrophage infections, but deletions of plc genes do not impair growth in the human macrophage cell line THP-1. The plcABC triple mutant and plcABCD quadruple mutants are both attenuated in the late phase of mouse infection, indicating plc genes are important for Mtb survival in hosts (Raynaud et al., 2002).

We obtained the plcC mutant from BEI Resources (Manassas, VA) and examined its cell infection rate, in comparison to the infection rate of wild type CDC1551 in A549 cells. The bacterial infection ratio of plcC mutant to wide type CDC1551 was 0.7 ± 0.1 based on DLF-1 fluorescence intensity, and was 0.5 ± 0.2 based on CFU enumeration (Table 1). These results indicate that plcC is an important gene for Mtb entering alveolar epithelial cells. The detailed mechanism of how plcC helps bacilli infect epithelial cells remains to be explored.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.